Bird Surveys-Expedition Field Techniques.

Expedition Field Techniques
BIRD SURVEYS
Colin Bibby, Martin Jones and Stuart Marsden
Published by the Expedition Advisory Centre

Royal Geographical Society
(with The Institute of British Geographers)
1 Kensington Gore
London SW7 2AR
Tel +44 (0)171 591 3030
Email [email protected]
Fax +44 (0)171 591 3031

Website www.rgs.org
October 1998
ISBN 0-907649-79-3
Front Cover: Chinese Grouse Bonasa sewerzowi, near-threatened and endemic to the Himalayan
foothills of Western China. Artist: Richard Thewlis.
BIRD SURVEYS
CONTENTS
Acknowledgements
Foreword
Introduction
1
2
3
Section One: Why Count Birds?

1.1
Species
1.2
Sites
1.3
Important Bird Areas
1.4
Habitats
1.5
Choice of methods
5
7
8
10
11
Section Two: Study Design

2.1
Introduction
2.1.1 Total counts
2.1.2 Sampling and bias
2.1.3 Sampling, precision and accuracy
2.1.4 Relative and absolute estimates
2.1.5 Measuring and increasing precision
2.2
Study Design
2.2.1 Choosing the right time and conditions
2.2.2 Observer bias
2.2.3 Sample sizes and replication
2.2.4 Positioning your sampling effort
2.3
Pilot surveys and training
2.4
Safety, team size and logistics
2.4.1 Health and safety
2.4.2 Logistics
2.4.3 What size team?
2.5
Which methods to use
2.5.1 Introduction
2.5.2 Bird densities
2.5.3 Point counts and line transects
15
15
15
16
17
18
19
19
21
24
26
29
31
31
31
31
32
32
33
34
Section Three: Estimating bird densities using distance sampling

3.1
Introduction
3.2
Distance sampling using line transects
3.2.1 Introduction
3.2.2 Positioning of transects
3.2.3 How many transects and how long should they be?
3.2.4 Collection of data
3.2.5 Double counting
3.2.6 Variable distances and distance bands
3.2.7 Distance estimations to groups
3.2.8 Sample sizes
3.3
Distance sampling using point counts
3.3.1 Introduction
3.3.2 Location of census stations
3.3.3 How many census stations?
3.3.4 How long should the count period be?
3.3.5 Data collection
3.3.7 Sample sizes
3.4
Examining the data
3.4.1 Bird detectability and the detection curve
3.4.2 Shape criteria
3.5
Using the DISTANCE software
3.5.1 The basic model
3.5.2 Alternative models
3.5.3 Testing model fit
3.5.4 Inputting data
3.5.5 Understanding the output
Section Four: Counting single-species
4.1
Is a single-species study appropriate?
4.2
Narrowing down the search: information is the key
4.3
The nature of the beast: rarity
4.4
Problem species and difficult habitats
4.4.1 Bird colonies/aggregations
4.4.2 Cryptic and understorey birds
4.4.3 Canopy species
4.4.4 Mixed-species flocks
4.4.5 Aerial birds
4.4.6 Nocturnal and crepuscular birds
4.4.7 Other problem birds (waterbirds,
35
36
36
36
37
37
40
40
41
41
41
41
42
43
43
43
44
45
45
45
47
50
50
50
51
51
52
53
53
57
60
60
61
61
62
62
62
Bird Surveys
4.5
4.6
4.7
birds of prey, migrants, etc.)

4.4.8 Special habitats/niches
Tailoring distance sampling methods for
individual situations
Specific techniques for special cases
4.6.1 Spot mapping
4.6.2 Counting birds at or near aggregations
4.6.3 Look down methods from vantage points
4.6.4 Nest searching/counting limited resources
4.6.5 Interviews with local people
4.6.6 Mark-recapture/banding/home ranges
4.6.7 Scientific birding
Interpreting and presenting results of specific studies
Section Five: Assessment of sites: measurement of species richness

diversity
5.1
Introduction
5.2
Compiling a species list
5.3
Standardising recording methods
5.3.1 Species discovery curves
5.3.2 Encounter rates
5.3.3 Mackinnon lists
5.3.4 Timed species-counts (TSCs)
5.3.5 Mist-netting
5.4
Analysis of data
5.4.1 Predicting total number of species from
species discovery curves
5.4.2 Analysis of encounter rate data
5.4.3 Analysis of Mackinnon list data
5.4.4 Analysis of TSC data
5.4.5 Analysis of mist-net data
5.5
Discussion
5.6
Sources of information for the recording of bird sounds
Section Six: Bird-habitat studies
6.1
Why study habitats?
6.2
Broad scale habitat studies
6.2.1 What features to map?
6.2.2 Sources of data for mapping
6.2.3 Verification of a map and sampling
6.3
Fine scale bird-habitat studies
63
63
64
67
67
68
70
71
73
73
74
75
and
76
76
81
81
83
84
86
86
87
87
89
91
91
92
94
97
99
100
100
101
102
102
6.4
6.3.1 Different approaches/survey designs

6.3.2 What habitat features to record and how?
6.3.3 Preparation of data for analysis
Analytical approaches
6.4.1 Summary statistics
6.4.2 Indices
6.4.3 Graphical and linear regression approaches
6.4.4 Logistic regression
6.4.5 Reducing the dimensions
6.4.6 Interpreting and testing the results
102
103
108
110
110
110
111
113
113
114
Section Seven: Maximising the impact of the work

7.1
Basic communications
7.2
Culture, politics and diplomacy
7.3
Summary reporting
7.4
Scientific reporting
7.5
Archiving the data
115
116
118
122
124
Section Eight: References and further reading
125
Section Nine: Sample input and output

files for DISTANCE Program
127
The BP Conservation Programme
134
Bird Surveys 1
Acknowledgements
This book has been assembled as a collaborative effort. Thanks are due to
authors of individual sections who additionally contributed to discussions
about the whole project and on sections other than their own.
We are particularly grateful to the following who read and helpfully criticised
large parts of the text, bringing to bear their experience from around the
world: Rod Hay, Sam Kanyamibwa, Borja Mila, Michael Poulsen, Richard
Thomas and Hazell Thompson. Individual sections have benefited from the
input of the following: Leon Bennun, S. Buckland (for technical help with
DISTANCE), Paul Dolman, Bill Sutherland, David Whitacre and Rob
Williams.
Ian Burfield, John Pilgrim and Rob Williams helped in many ways in coordinating and bringing the project to fruition. Katharine Gotto from the BP
Conservation Programme helped and encouraged throughout. We appreciate
the help of Shane Winser and Timothy Jones at the Expedition Advisory
Centre of the Royal Geographical Society for overseeing the final production
process. Financial support from BP has allowed us to keep down the cost of
the publication so that it can be distributed more widely.
Author Profiles:
Colin Bibby is the director of Science and Policy at BirdLife Internationals UK
headquarters. His interests are in the collection and use of scientific information for
the conservation of birds and their habitats.
Martin Jones is a senior lecturer in the Behavioural and Environmental Biology
Group in the Department of Biological Sciences at Manchester Metropolitan
University. His research interests are the ecology and conservation of tropical birds
and the effects of habitat change on island birds and butterflies.
Stuart Marsden is a lecturer in Behavioural and Environmental Biology in the
Department of Biological Sciences at Manchester Metropolitan University. His main
research interests are in the ecology and conservation of parrots and other tropical
birds.
*Figure 24 has been kindly reproduced with the permission of Oxford University
Press. Taken from: Mackinnon, J. and Phillips, K. (1993) A Field Guide to the Birds
of Sumatra, Java and Bali.
2 Expedition Field Techniques
FOREWORD
This book is a vital tool for everyone wishing to contribute to our knowledge
of the world's birds and to bird conservation. Effective conservation planning
can only be based on a sound knowledge of the species, sites and habitats in
need of protection.
Despite birds being the best known class of living organisms there are still
substantial gaps in our knowledge of the distributions, abundances and
densities of species. Birds have been demonstrated to serve as good indicators
of biodiversity and environmental change and as such can be used to make
strategic conservation planning decisions for the wider environment.
BirdLife International is delighted to have been able to collaborate with
the Expedition Advisory Centre of the Royal Geographical Society (with the
Institute of British Geographers) to produce this much needed volume. The
editors and authors are all experienced in their subjects and the book has been
reviewed and refined by specialists from around the world.
To make the best decisions, it is most important that the information on
which such decisions are made is as accurate, systematic and representative as
possible. The methods in this book will enable the user to survey birds simply
and effectively.
This book will be much used by conservationists, researchers and birders,
both amateur and professional throughout the world. I expect it will make a
significant contribution towards the furthering of knowledge about the
avifauna of the world and towards safeguarding biodiversity.
Dr Michael Rands
Director and Chief Executive, BirdLife International
Bird Surveys 3
INTRODUCTION
There are many reasons for counting birds and a large and rather forbidding
literature on the subject. Birds are among the best known parts of the Earths
biodiversity. But nevertheless soundly quantified knowledge is far from
complete for most species and regions. We believe that this is an obstacle to
conservation of birds which ornithologists can help to rectify. Birds are
relatively easier to count than most other wildlife and ornithologists have a
distinct contribution to make to biodiversity conservation by improving our
understanding of the planet, the location of biodiversity and threats it faces
from non-sustainable practices.
Unashamedly we have focused strongly on the application of counting
methods for conservation. For this reason, we have to a degree biased
coverage to forests and the tropics where so much biodiversity resides. We
have tried to impart general principles and some practical techniques in a
clear and simple manner. We may be criticised for citing few references but
have done this to help readers get going without feeling that there is a
challenge to read a huge literature first. We hope that we have illustrated
enough of the principles behind bird counting to enable the reader to take a
critical attitude to their planned study. Many of the principles are common to
any method applied to any species or habitat. There are other important kinds
of study of birds, such as ecology or population dynamics which we have not
covered at all.
This book is intended to help conservation professionals or students plan
field surveys at home or abroad. It is not possible to count birds without a
good field knowledge and ability to identify them, but this aside, we have
tried to make no prior assumptions about the skill of the reader. We have tried
to write both for internationally travelling students who have contributed so
much in the past and for biologists in developing countries who have so much
more to contribute in the future.
We will judge the success of this book by the extent to which we see it
cited in studies of important areas or threatened species which come through
to influence conservation decisions at local and national level around the
world. Ultimate success will be judged by the impact that you, the reader, can
have for conservation by collecting new and important data and supporting
the development of more effective conservation work wherever you live.
Bird Surveys 5
Section 1
WHY COUNT BIRDS?

Colin Bibby
There are many good reasons for counting birds but this guide aims to
promote better knowledge to help conservation. A recurring theme will be
that well designed field studies start with a clear purpose. The sharp definition
of purpose is probably one of the more difficult steps in designing a good
study. Once a purpose is clear, it becomes much more obvious whether any
particular study design has a reasonable chance of working and whether there
are variants which would be better.
Most surveys target a particular species or a particular place. Important
questions may arise about the use a species makes of habitats or, at a site, the
condition of different habitats and the species that occur in them. The division
between species, sites and habitats will recur throughout the book.
1.1 Species
There is an urgent need to know more about the worlds most threatened bird
species. These are officially listed in the BirdLife International publication,
Birds to Watch 2 (Collar et al. 1994). The definition of globally threatened
species has been agreed by the Species Survival Commission of the World
Conservation Union (IUCN, 1994). The system puts different species into
categories according to a set of criteria (Figure 1). The most important data
are population size and range, and trends in one or other of these. Trends are
impossible to measure unless some baseline has previously been set. For most
species this has yet to be done. In addition, threat might be measured as
known or inferred change of extent or condition of habitat. This can only be
applied if the habitat requirements of the target species are reasonably well
known. For the majority of bird species, and especially for many threatened
species, these most basic parameters are simply unknown. By 1992 for
instance, less than a quarter of threatened species in the Neotropics had been
subject to any formal counting.
One way of telling how effective this book has been will be the rate at
which successive editions of Birds to Watch show development of
quantitative knowledge on threatened species. Viewed the other way round,
perusal of Birds to Watch offers anyone who is interested a clear challenge to
get out into the field and collect some new data of real value. Such
information will not only help to ensure that threatened species are correctly
recognised, but will also help planning for their conservation. Birds to Watch
also has a category of near-threatened. This covers species given a
precautionary listing until sufficient data have been collected and analysed to
decide how their conservation status should be described.
Figure 1. Some of the criteria used for identification of IUCN Red List
Categories (from IUCN, 1994).
Population
decline
Extent of
occurrence *
Area of
occupancy *
Population
level *
Population level
Critical
>80% in 10 yrs
Endangered
>50% in 10 yrs
Vulnerable
>20% in 10 yrs
<100 km2
<5000 km2
<20,000 km2
<10 km2
<500 km2
<2,000 km2
<250
individuals
<50
individuals
<2,500
individuals
<250
individuals
<10,000
individuals
<1,000
individuals
* These criteria do not categorise species alone but have to be met in

combination with other factors indicating declines, or fragmentation of
populations. Data may be known, estimated, inferred or suspected but need
to be documented. Extent of occurrence is overall range size the area of a
polygon embracing localities. Area of occupancy is the total area of habitats
occupied, so may be much smaller but depends on a knowledge of habitat
preferences and extent of suitable habitat.
Many countries have official lists of species of national priority. These are
often based on similar ideas about range, numbers and trends but with lower
thresholds. Other species attract attention because they may be potential
environmental indicators, or simply because they are popular so called
flagship species. Given the pressing need for information on globally
threatened species, we would urge species-oriented work in remote places to
concentrate on these. There is merit in collecting quantitative data on as many
species as possible at the same time. This is often a sensible approach because
looking for threatened birds can be time consuming with rather little data in
return. Better then, to collect some systematic information on other species
while looking for your threatened target species. In other circumstances,
good data can often be collected by focusing on a single-species especially if
using a technique like play-back.
Bird Surveys 7
1.2 Sites
Species may be common currency to biologists but they are not very practical
targets for conservation. Put simply, there are just too many to be treated one
by one it is important to remember that most species are not birds and that
the vast majority are not even known to science. A more practical unit for
conservation is the protected area. This might be strictly protected for nature
conservation or, particularly in the developing world, might include
utilisation by humans.
BirdLife International has demonstrated (ICBP, 1992) the location of 218
Endemic Bird Areas (EBAs) to which one quarter of bird species on Earth are
confined (Figure 2). The Albertine Rift Mountains in central Africa, for
2
example, are about 56,000 km in extent and have 36 endemic bird species.
These EBAs, which occupy just 5% of the Earth's land surface, embrace some
three quarters of all threatened species. They are therefore critical regions for
conservation. Within the EBAs, there is a pressing need to narrow down
relatively large regions into sites of a size that are already protected or may
become so in the future. Full documentation of the EBAs is available in
Stattersfield et al. (1998). This adds a geographic dimension to the inspiration
available from Birds to Watch 2 and indicates many areas in need of
ornithological exploration.
To contribute helpfully to the conservation of sites, one needs to know
where they are and what occurs within them. In remote areas with poor
access, even locating the boundaries of a site may be practically (and
conceptually) very difficult. Well designed fieldwork needs to be clear about
its geographic boundaries.
Pioneering new areas is obviously exciting but there are also large gaps in
our knowledge of the birds of existing protected areas. Filling these can have
practical use to local managers and conservationists. The information helps us
to understand which species are most important, which might require special
management because of their poor status and which might be so rare as to be
in need of further protection elsewhere within their ranges. Baseline counts
will come to be greatly valued when repeated in the future. They will show
which species have declined in numbers and hence need additional
management if they are not to disappear.
Figure 2. Location of 218 Endemic Bird Areas (taken from ICBP, 1992;
Stattersfield et al. 1998). Each of these areas has at least two bird species
solely confined to it. In total they occupy about 5% of the Earth's land
surface but one quarter of all bird species and about three quarters of all
threatened species are confined to them. As a result, these areas are of very
high priority for further exploration and conservation.
1.3 Important Bird Areas

BirdLife International is in the process of documenting sites of global
importance for conservation in its Important Bird Areas (IBA) programme
(Figure 3).
Figure 3. BirdLife International Important Bird Areas programme.
IBAs :
are places of international significance for the conservation of birds at
the global, regional or sub-regional level;
are practical tools for conservation;
are chosen using standardised, agreed criteria applied with common
sense;
must, wherever possible, be large enough to support self-sustaining
populations of those species for which they are important;
must be amenable to conservation and, as far as possible, be
delimitable from surrounding areas;
will preferably include, where appropriate, existing protected areas;
should form part of a wider, integrated approach to conservation that
embraces sites, species and habitats.
The function of this programme is to identify and protect a network of
sites, at a biogeographic scale, that are critical for the long-term viability of
naturally occurring bird populations, across the range of those bird species for
Bird Surveys 9
which a sites-based approach is appropriate. Sites in Europe (Grimmett and

Jones, 1989) and the Middle East (Evans, 1994) have already been
documented and work is developing in the rest of the world. IBAs are
identified by a set of globally-agreed criteria. This is important to ensure the
credibility of the whole set a site cannot just be an IBA because someone
feels it is important; there has to be some supporting data. Criteria fall into
four groups (see Figure 4):
globally threatened species

restricted-range species
biome restricted assemblages
congregatory species
Population estimates are required to identify IBAs for globally threatened

species and for congregatory species (generally waterfowl or seabirds) and
appropriate techniques are discussed in Sections 3 and 4. Restricted range
species and assemblages may be sufficiently measured by one of the methods
described in Section 5. Biomes and their definitive birds have been described
for the IBA programmes for the major continents and more information on
the IBA programme is available from the partner organisations or offices of
BirdLife.
Figure 4. Criteria for globally Important Bird Areas (part).
Category
A1 Globally
threatened species
A2 Restricted-range
species
A3 Biome-restricted
assemblages
A4 Congregations
Criterion
The site regularly holds significant numbers of a
globally threatened species, or other species of global
conservation concern.
The site is known or thought to hold a significant
component of a group of species whose breeding
distributions define an Endemic Bird Area (EBA)
The site is known or thought to hold a significant
component of the group of species whose distributions
are largely or wholly confined to one biome.
i) Site known or thought to hold, on a regular basis,
1% of a biogeographic population of a congregatory
waterbird species. Or:
ii) Site known or thought to hold, on a regular basis,
1% of the global population of a congregatory seabird
or terrestrial species. Or:
iii) Site known or thought to hold, on a regular basis,
20,000 waterbirds or 10,000 pairs of seabirds of one
or more species. Or:
iv) Site known or thought to exceed thresholds set for

migratory species at bottleneck sites.
Determining boundaries of IBAs can be tricky. Sites should, as far as

possible:
be different in character or habitat or ornithological importance from the
surrounding area;
exist as an actual or potential protected area, with or without buffer zones,
or be an area which can be managed in some way, as a unit, for nature
conservation;
alone or with other sites, be a self-sufficient area which provides all the
requirements of the birds (that it is important for) which use it during the
time they are present.
Simple, conspicuous boundaries such as roads, rivers, railway lines, etc.
may be used to delimit site margins while features such as watersheds and
hilltops may help in places where there are no obvious discontinuities in
habitat (transitions of vegetation or substrate). Boundaries of ownership may
also be relevant.
1.4 Habitats
Within sites, it is fairly evident that habitat is likely to be an important
determinant of the distribution and number of birds. For sites which are not
protected, habitats might be changing, for instance as a result of logging.
Adequate management obviously depends on understanding the relationship
between birds and their habitats. If a study is oriented to a particular species,
it is also evident that questions about its distribution, ecology and threats to
its status, will partly be answered with an understanding of its habitat
requirements. While much about a bird's ecology might be studied directly in
terms of its diet, foraging behaviour or population dynamics, important
knowledge of habitats can be gleaned from good census studies.
Explicit questions about habitats are likely to take a certain form, e.g.
what are the major variations of habitats around here, and how does the
abundance of birds vary with them? Variations might be of natural origin, for
instance by soil type, along a gradient of rainfall or by altitude. Important
variations might have human origin, such as the degree of impact of logging
on forest structure, from mature to selectively-logged to clear-felled and
regrowing secondary stands.
Questions of this kind need a method for recognising and describing
variations in habitats. They also call for a well designed study capable of
collecting sufficient data across the range of habitats involved. Section 6 has
Bird Surveys 11
been written to indicate some methods for measuring and describing

vegetation and habitats because this is such an important part of answering
questions about sites and species.
1.5 Choice of methods

A simple checklist of questions (Figure 5) should help to ensure good design
of a study by pointing out problem areas where methods might not be
properly linked to the original question.
Figure 5. Eleven questions to answer in designing a study.
what is the question?

who will use the results?
who are the appropriate contacts?
where are the boundaries of the study?
how is the effort going to be distributed?
what methods will be used in the field?
is the method good enough for the purpose?
is the study realistic?
what preparations are needed?
how will the data be analysed?
how will the results be disseminated?
What is the question? The more simply one or more questions can be
posed the better. What is the status of the regionally endemic bird species in
area A? What are the effects of logging on birds in place B? What is the
world population of bird C?
Who will use the results? Protected area managers might have a very
clear idea of why they want particular information and thus what kind of field
data will be needed. Even in the same place, a research study on the
population viability of a particular species will need a different approach.
Who are the appropriate contacts? Section 7 elaborates the point that the
impact of your study will be greatly influenced by the quality of your local
contacts and diplomacy. In preparation you need local points of contact to
help plan a study that will be helpful to local or national authorities in a
position to use the information you discover. Obviously you need access to
the most up to date intelligence on what surveys have recently been done or
are planned and where the gaps are. Obvious starting points are BirdLife
Partners or offices, other established bird clubs, natural history societies,
conservation organisations or universities. Government departments might

help if you can match your interests with theirs.
Where are the boundaries of the study? This is a more difficult question
than it might seem. The status of a globally threatened species needs to be
assessed across its whole range, which might be large and poorly known.
Even a single protected area in the tropics might well be too large to be easily
covered in one visit. If the study cannot embrace such a large area, then it is
important to define the smaller area that it will cover properly. Otherwise, the
results might be very difficult for anyone else to use subsequently.
How is the effort going to be distributed? If the boundaries of the study
area are too large to allow complete coverage, then the study design must
include sampling. Sampling can be a powerful way of inferring general
patterns if it is done correctly. If it is ignored, or done badly, it can be very
misleading. Sampling might be random, which is good in theory but tough in
practice. Stratification may often be appropriate, see Section 2. With different
habitats sampled at appropriate levels the effort going into different plots or
strata should ideally be planned, but if, as is often the case, this is not
realistic, then it at least needs to be measured and documented.
What methods will be used in the field? There are only half a dozen
essentially different things that bird counters do in the field. They all require
some discipline slightly greater than pure bird-watching, (tempting though
this might be in a new place rich in exciting species). The simplest methods
(Section 5) add very little more than some basic note taking to bird-watching.
If one or two methods are selected in advance, there will be merit in
designing a data recording sheet. This has the advantage of reminding people
what data they need to record, promotes standardisation and allows a daily
check of how things are going. It is also good for data security because if the
previous days results can be safely stored away, there is no risk of loss when
going into the field again with a note-book.
Is the method good enough for the purpose? It is not easy to describe,
for all circumstances, how this test might be passed or failed. As a rule of
thumb, you need a minimum of about ten records of a species to make a
reasonable estimate of its abundance. To describe a forest bird community or
the birds of a habitat type requires a minimum of about 50 point counts or
10km of transect (see Section 3). With less formal methods, this might be 10
20 species lists, or one hour lists depending on the richness of the habitat (see
Section 5). An ideal study would not only cover several habitat types but
Bird Surveys 13
would also provide two to four replications of each in order to see whether or
not there is consistency in the generality of the results.
Unappreciated bias is a problem with the most devastating potential to
make the results of a study useless. If a key part of the range of habitats in an
area has not been sampled then it is not possible to make any inferences as to
what might be there. If this is appreciated it is no problem the results of the
study simply apply to a smaller known area. If it is not appreciated, then it is
not possible to say how applicable the results are, or to what area. There are
many other causes of bias (Section 2) which can be problematic if left to
themselves. What happens if the observer who is not actually a very
experienced ornithologist is the only person collecting data in one habitat
type, while a very observant and experienced recorder gathers the data in
another? What happens if key elements of the method are allowed to vary in
just one part of the study?
There is a common tendency to believe that any results other than the very
precise are not much use. It can be difficult to get precise results, particularly
in circumstances where fieldwork is arduous. This does not mean that it is
difficult to get any worthwhile results from such places. For many possible
questions, even quite imprecise data are enough and certainly much better
than no data at all. The handling of bias and questions about precision are
discussed in Section 2.
What preparations are needed? Time spent in preparation is rarely
wasted. Have you planned what you need to prepare for? Have you got all the
relevant background information on the area and its birds? How are you
going to learn to recognise birds or habitats in the area? Are there any key
people who might be able to help? You really need to talk with or involve
people who have used the planned methods before and who know about the
study area. Would vegetation maps, air photos or satellite images help?
Where will you find them?
For a trip to an unfamiliar foreign country, it might take most of the
preceding year to prepare fully. Even on a project to a remote part of ones
own country, preparations might take several months. Having arrived on site,
it is a good idea to practice and check methods before getting going. This
particularly includes identification of birds or trees (if needed) and measures
of distances. The better the prior preparation, the quicker this stage will be.
Is the study realistic? It is very common to be over-ambitious in
designing a study. It may actually be more useful to set and achieve a modest
objective than to half accomplish something grander and end up with an
unfinished job of limited value to anyone. As broad guidance, consider that

arranging permissions and local diplomacy might take as much as two weeks.
Even when out of the city, expect to use about half your days for local travel,
domestic maintenance or illness. Depending on where you are and the season,
days will sometimes be lost to the weather. In a new area it might take two
weeks to become familiar with the birds and design the practical details of a
study. So an eight week trip to a remote area might allow as little as 15 days
of fieldwork. During those days, one person can realistically make ten point
counts or walk a 4km transect per day, though in ideal circumstances these
figures might be doubled. These approximate limits are set by confining
fieldwork to the best time of day, by energy consumed in access and by the
rate at which it is possible to sustain fieldwork day after day. For safety
reasons it might be better to work in pairs and you might need people
guarding the camp or running domestic chores.
How will the data be analysed? The benefit of thinking about analysis
before collecting any data is that it reduces the likelihood of anything
important slipping past unmeasured or unconsidered. Are there any
considerations to make numerical data available for computer entry and
analysis? Is all the field data properly coded for location, altitude and habitat
measurements? Does anyone know how to use the software for estimating
densities?
How will the results be disseminated? There is no point in analysing the
data if the results are not going to be communicated to somebody (see Section
7). The more carefully you think in advance about what your report will be
like, the more likely that you will end up collecting appropriate data in a
suitable way. Another important thought is whether the study will be repeated
by yourselves or someone else. If it is important, one would certainly hope so.
Would it be possible for someone else to be able to repeat what you plan to
do?
If you can give clear answers to these questions your study deserves to
work and it is time for you to go into the field. The clarity of your prior
thinking will repay you well. Indeed, it might even pay you in advance since
evidence of careful planning and prior thought is very attractive to funders. In
truth, realities in the field will intrude on the best laid plans and you will need
the flexibility to change things as you learn more.
Bird Surveys 15
Section 2
STUDY DESIGN
Martin Jones
2.1 Introduction
The best studies are the ones where the participants not only have a very clear
idea of their aims but also understand the methods they are going to adopt and
crucially know how they are going to analyse the data. Once your aims
have been formulated, and with the help of this book, you should be able to
identify the appropriate methods and analysis for your study. At least one
team member should then become fully conversant with all aspects of data
analysis techniques before field work commences. Prior to beginning
fieldwork it is possible to plan the study in broad terms, but fine tuning will
always depend upon local knowledge, results of pilot studies and initial
analysis of the results as they come in. If it is apparent that the aims of the
study will not be adequately met you have two options either redesign the
work or modify the aims!
2.1.1 Total counts
In a few cases, it may be possible to make a total and accurate count of a
species, either within its world range or within a defined habitat or protected
area. A more likely situation is that total counts are impossible and some sort
of sampling is required. Sampling is always needed for establishing habitat
associations, for multi-species surveys, and for diversity studies.
2.1.2 Sampling and bias
The basic idea which underpins sampling is that because we cannot count a
whole population or bird community, we take samples and extrapolate our
results to provide estimates of the true population sizes or species diversities.
In the same way, we might sample a variety of habitats to try to build up a
true picture of what a species habitat requirements really are. The problem
with any sort of sampling is that there are many ways in which the sampling
regime could be biased. For example, many birds are more active and vocal
early in the morning, so if two forest areas are censused, one between 0600
and 0800h and the other between 1300 and 1500h, the results cannot be
compared; the first area may seem to have more birds but is this because of a
real difference in the bird populations, or just because the birds were easier to
see and hear? The sampling regime was obviously biased, and there are many
other ways in which bias can affect the outcome of any bird counting
exercise. Another example of bias is comparing results from a noisy

environment (e.g. riparian forest) with a quiet habitat. Understanding the
causes of bias and dealing with it in the appropriate way is the most important
part of study design and is dealt with in Section 2.2.
2.1.3 Sampling, precision and accuracy
If we are estimating a population, assessing species diversity or studying
habitat associations, we would never usually take just one sample. Even if we
could eliminate all sources of sampling bias, natural variation in habitats and
bird distribution will mean that samples are different. The term precision
describes the closeness of repeated sample estimates to each other, while
accuracy describes how close the estimates are to the real value.
For example, if we wanted to estimate the population density of a
particular bird species in an area of forest, we could use some equally-sized
sample plots and count the individual birds in each plot. If we had five plots
the results could be 1, 3, 12, 9 and 15, with a mean of 8 birds per plot.
However, we could also have a mean of 8 birds with results of 5, 10, 7, 8 and
10. There is a smaller spread of values around the mean in the second set of
data, which allows us to say that the results are more precise than the first set.
We may have a precise answer, but is it accurate? Unfortunately, in most
bird censusing work we can never know the answer to this question. In the
example above, perhaps some of the birds in the census plots were missed by
the observer; some individuals may not have been moving or calling and were
therefore cryptic (difficult to detect). If this holds for all of the plots, we may
still have a precise mean estimate (all the samples are close to the mean) but it
is actually a biased estimate it is an underestimate of the real density and is
not accurate. The relationship between precision and accuracy is further
explained in Figure 6.
Bird Surveys 17
Figure 6. The relationship between
precision and accuracy.

a) Imprecise and inaccurate
c) Precise and inaccurate
b) Imprecise and accurate
d) Precise and accurate
The graphs shown are estimates of the density of bird species. The real
density, unknown to those collecting the data, is indicated by a P. In 6a the
results from the separate plots cover a wide range which does not encompass
the true density - the estimates are imprecise and they are inaccurate. In 6b
the estimates cover a similarly wide range but this time the true value is
within that range - the estimates are imprecise but accurate. In 6c there is a
narrow range of estimates which do not encompass the real value - precise
but inaccurate and in 6d there is a narrow range which encompasses the real
value - precise and accurate.
As we rarely know whether the answer is accurate or not, all we can do is
get as precise and therefore a reliable answer as possible. If we have
recognised and tried to minimise the bias in our sampling methods, we would
also hope that the answer was an accurate one.
2.1.4 Relative and absolute estimates
In some cases, the accuracy of the estimate is of secondary importance or may
not even be relevant. For example, if you want to know if numbers of a
particular species are increasing or decreasing you could set up some census
routes, record bird contacts and use these data as a baseline to compare with
data collected in exactly the same way in the future. This is a relative
estimate. The actual density of birds is not important; all that matters is how
one estimate relates to another. With relative estimates you may even accept
some types of bias, provided the same bias is present when the census is
repeated. Thus, referring to Figure 6, for studies of population change (and
relative estimates in general), the data presented in 6c are as useful as those in
6d. If relative estimates do satisfy the aims of the study, it is particularly
important that the methods are recorded well enough to be repeatable.
If the aims of the study require us to know the actual density of birds, then
what we attempt is an absolute rather than a relative estimate, and now the
elimination of bias is the most important consideration. More information on
the choice between relative and absolute estimates is given in Section 2.5.
2.1.5 Measuring and increasing precision
Whether we are attempting relative or absolute population estimates, a major
goal of study design is to provide as precise an estimate as possible; we
therefore need some way of measuring precision. There are a number of
different statistics that could be used, but perhaps the most useful thing to do
is to calculate the 95% confidence limits of your estimate (the DISTANCE
software discussed in Section 3 automatically calculates the 95% confidence
limits of any estimate it produces). One way of defining the 95% confidence
Bird Surveys 19
limits (although not absolutely correct in statistical terms) is to say that there
is a 95% chance that the true estimate lies between the upper and lower limits.
For example, a population density estimate of 250 birds per km2 might have
limits of 50 and 450, i.e. you are 95% sure that the true density is between
these two limits.
Having calculated the confidence limits for an estimate, it might be
apparent that the estimate is so imprecise as to be virtually useless, so how
can we increase precision? One way is to increase our sample size the more
samples you take, the more precise (and more reliable) the estimate will
become. Unfortunately, improvements in precision are proportional to the
square root of the sample size, so to double the precision you need to increase
the sample sizes fourfold.
A biased sampling procedure may also contribute significantly to
imprecision of the estimates, and this also needs to be recognised and
addressed. For example, if half the sampling sessions were in the morning and
half in the afternoon when perhaps the birds were less active and more likely
to be missed, combining data from morning and afternoon sessions would
give lower density estimates but with wider confidence intervals a less
precise estimate as a result of a biased sampling procedure (more information
in Section 2.2.1).
2.2 Study Design

In order to get as accurate estimates as possible, or at least to know why
estimates may not be accurate, we need to identify and address any causes of
bias in our sampling regime. In the unlikely event of us being able to
eliminate all causes of bias, natural variation in habitats and bird distribution
will still reduce the precision of the estimates. In designing a study, we
therefore need to consider both the problems of bias and how we monitor and
cope with natural variation.
2.2.1 Choosing the right time and conditions
Many factors will affect bird activity and behaviour, and these in turn affect
your chances of actually recording the birds. Among the more important
factors are time of day, the season and the weather.
Time of day
Figure 7 illustrates some of the effects of time of day on bird activity. These
data on parrots and a hornbill species were collected from a vantage point
overlooking a small patch of forest on the island of Sumba, Indonesia. There
are morning and late evening peaks of activity with many fewer movements in
the middle of the day. Many forest birds will show similar trends, and singing
and calling can be even more strongly biased towards the early morning
activity peak. The aim of a census may be to record as many as possible of
the birds that are actually present, and usually as quickly as possible, so
collecting data at the peak of bird activity can be fundamental to good study
design. However, birds can be so vocal and active at dawn that it may be
impossible to record all bird contacts correctly and there can be rapid changes
in conspicuousness over a short time. A common study design, therefore, is to
begin data collection about 30 minutes after dawn and continue to midmorning, when bird activity declines. There may be another censusing period
before dusk. As part of a pilot study (section 2.3), you could determine
empirically when the best time for your own surveying would be.
Bird Surveys 21
Figure 7. Changes in bird activity with time of day. Shown are the
frequencies of flights of Indonesian parrots and hornbills from data collected
100
during long watches overlooking forest patches. Flight frequencies are
expressed as percentages
of the maxima (0600 to 0700h for parrots and 1700
80
to 1800h for hornbills).
60
Parrots
Flight
Even though censusing can be restricted to periods of higher bird activity,
frequency
Hornbills
there
is bound 40
to be some variation of activity within the restricted
period.
This can be an important cause of bias. For instance, if all censusing starts at
20
a field base and moves into the surrounding forest, all adjacent areas of forest
will be censused0early in the day and all distant areas later in the day. If more
bird contacts are made earlier in the morning, then the adjacent forest areas
6
8 10 12 14 16 18
will erroneously appear to have higher bird densities and diversities than the
other areas. A good study design
reduce this bias by ensuring, for
Hourlywould
periods
example, that alternate censuses were begun at the 'far end' of the routes. If
census routes are being repeated, then the same route should be walked from
both ends.
Season
Seasonal effects can be more difficult to cope with. Bird conspicuousness will
probably change with season, and in tropical forests there may not be
synchronisation of breeding cycles between or even within species. In a
species which is breeding, the males may be singing and calling to defend a
territory and so may be easy to record, whereas the females incubating eggs
may be the opposite. There can be no hard and fast rules about whether
studies are better designed to avoid or coincide with the peaks of breeding
activity, as this is better determined by the aims of the study (e.g. do you want
to get information on the breeding population, or perhaps the non-breeders
and migrants?). The best approach to reducing bias is to collect as much
information as possible on breeding activity as you progress through the
study. If, at the end, you discover that all contacts were with singing males,
you might be able to assume that females were incubating eggs and that your
population estimate should be doubled. However, in some cases sex ratios
may be unequal, and this can be an erroneous assumption. Furthermore, in a
population where there are many unpaired males, there may be more song
than in a healthier population where all males are paired. In some cases where
you do record both males and females, but you realise they are behaving very
differently, it may be appropriate to calculate densities for the sexes
separately and then add the estimates together.
Weather conditions
Adverse weather conditions such as low cloud, high winds, rainfall and even
very high temperatures can affect census results in three ways. Firstly, bird
activity can be directly affected (usually reduced), which will affect the
efficiency and reliability of your data collection. Secondly, the conditions
could reduce your chances of actually seeing or hearing the birds. Thirdly,
you cannot pay adequate attention to counting if you are too hot, too cold or
wet. Census results can also be affected by conditions underfoot (during dry
periods, fallen leaves may become very noisy to walk on), or by the noise of
cicadas (whose activity is influenced, amongst other things, by temperature
and humidity).
In order to reduce bias, all censusing should be carried out under a
standard set of conditions, e.g. light winds and no precipitation. It is also a
good idea to record weather conditions such as cloud cover, wind strength
and temperature even when they do conform to your 'standard' conditions,
since you might want to analyse their effects later.
2.2.2 Observer bias
Species identification
Being able to identify your target species is an obvious prerequisite for any
study. Assigning contacts to the wrong species can cause under- and overestimation of densities, as well as bias in species diversity estimates. For
many forest studies, difficult identification problems are compounded by the
fact that many contacts are through songs and calls. In a recent study on the
Indonesian island of Sumba, the percentage of contacts for different species
that were through calls rather than sightings varied from 0% to 99%, but the
mean value was just over 70% (Jones, unpubl.) It may require weeks of
practice to learn the calls and be able to recognise the majority of contacts.
It will have become obvious by now (if not during the planning stages of
the work) that it is impossible for an inexperienced team to visit an area of
tropical forest and expect to census the entire bird fauna. An exception to this
might be a project on a small island with a small number of species, but
normally you will need to restrict the scope of the project. It may be better to
have precise population estimates of a few key species or diversity estimates
for an important guild of species than unreliable data on the whole fauna.
The magnitude of the species recognition problem is partially dependent
upon how much information is already available. If your species and area are
covered by a field guide and bird call tapes, you have a good basis on which
to build up your knowledge. If not, you will need to collect all the available
Bird Surveys 23
information and prepare your own version of a guide. It is often worthwhile

visiting a museum and taking photographs and notes on lesser known species.
Once in the field, it will be useful to involve local guides/scientists who know
the birds and their calls.
Once you are at the study site, there is no substitute for good fieldcraft and
bird-watching skills, taking careful notes and discussing identification
problems with fellow recorders. A useful technique to employ during a pilot
survey (see Section 2.3) and even during the main data collection is to plot
the proportion of unknown contacts over time. Figure 8 shows such a plot for
the Sumba study.
Figure 8. The decline in proportion of unidentified contacts with increasing
field experience.
These data are taken from two visits to Sumba Island, Indonesia in 1989 and
1992. The 1989 0.16
visit was the first contact with the fauna by the survey team
but the same recorders were also on the 1992 expedition. In 1989 after 4
days of experience in the field, the proportion of unidentified contacts was
0.12
down to less than 0.1 (10%) and after ten days it was standing at 0.04 (4%).
Proportion
1989
In 1992, because of the previous experience in 1989, fewer birds were
0.08
of contacts
unidentified
at the beginning of the study. Many of the unidentified contacts
1992
unidentified
were
later identified from notes on their calls and behaviour.
0.04
Estimating distances
0 require observers to estimate distances to bird contacts;
Some census methods
these estimates can be1 a 2major
3 cause
4 5of 6bias.7 Small
8 9random errors are
acceptable, but large or systematic over- or under-estimates of distance are
very serious. There are two ways to
reduce days
these errors. The first is just to
Census
practise by selecting an object, estimating the distance to it and then checking
the estimate with a tape measure. This practising can begin at home and can
be finished off at the field site. When data collection is underway, it is very
important to check some distances regularly to make sure there is no drift in
your estimates. Everyone can improve their estimates with practice, but allow
at least a week (two would be better) of regular practice and monitor how
well different team members are performing (see Section 2.3). This training
period is not wasted time: you need it to get unbiased density estimates and it
is probably also the time when you are learning the bird fauna. The second
way to reduce errors is not a substitute for the practice but makes estimation
easier: under some circumstances it may be possible to use an optical range
finder (not usually much use in forests) or, if you are using point counts, you
could position reference markers at known distances from your census points.
It is, of course, much more difficult to estimate distances to bird calls (and
most of the contacts may well be calls) and here even more practice is
required. Ideally, one team member could play calls of various species from a
tape recorder at measured distances, but out of sight, from the rest of the
team, who then make their practice estimates.
It may seem a difficult problem to estimate distances reliably, but there
are three important things to remember. The first is that, in forests, density
estimates are usually based upon contacts over fairly short distances (for
many flycatchers, warblers and sunbirds it is often the contacts up to 20m
which are important) and these are likely to suffer from smaller errors. The
second is that although it is better to estimate distances to individual bird
contacts, it is perfectly acceptable to classify contacts into distance bands or
even within or outside a specified distance (see Section 3). If distance
estimates are particularly error-prone, it is obviously easier and also
statistically better to allocate the distances in this way. The third point is that
in spite of the problems, there is more information in a census with distance
estimates than in one without.
Inter-observer differences
Even after a lot of training, there may still be some differences between team
members in their recognition of bird species and in their estimates of
distances. There will also be differences in visual and aural acuity and in
powers of concentration. It is very important to be aware of, and to try and
accommodate, these differences even if you cannot eliminate the bias, it is
often better for all team members to be making the same errors, rather than
each member making a different error.
It is important to discover and continually monitor what the inter-observer
differences might be. The best way to do this is to carry out a pilot study as
part of the initial training period, and to build in further monitoring
throughout the data collection period (see Section 2.3). Once the differences
have been identified, a number of different options are available to deal with
them. Firstly, the differences could be eliminated by further practice and/or
negotiation during the training period; secondly, you could allocate different
duties to different team members, e.g. the best distance estimators should be
censusing birds rather than measuring tree girths; thirdly, you could organise
the fieldwork so that bias is hopefully cancelled out. As an illustration of the
third point, if you have two main bird recorders in the team and you know or
suspect that there are differences between them, you can do two things; either
make sure that all censuses are carried out separately by both recorders and
the data pooled, or ensure that they carry out the censuses together and all
Bird Surveys 25
distance estimations and species identifications are agreed between them. The
wrong thing to do is to continually send one recorder to habitat A and the
other to habitat B, as you will then be unsure whether any differences found
were real.
2.2.3 Sample sizes and replication
Questions about sample size (numbers of contacts for a species, or number of
sites sampled) relate more to coping with natural variability than observer
bias. As a general rule, the more natural variability there is, the larger sample
sizes you will need to get reasonably precise estimates. Birds which are
clumped in distribution or usually occur in flocks will need more sampling
effort as there is more natural variation in their distributions (see Section 4).
In the majority of bird conservation studies, sample sizes are too low. This
is not necessarily because of a lack of effort, but because threatened species
are usually rare. In most cases it is wise to collect as much data as possible;
but remember that initial increases in sample size have a relatively large
effect on precision while the continued increase in sample sizes has less and
less effect (see Section 2.1.5). The question of when to stop collecting data
may never arise if the target species are particularly rare. It may arise for
commoner species and/or long periods in the field, or for diversity studies.
There are two strategies for determining your required sample sizes. The
first is that some of the methods for estimating bird densities (Section 3) and
examining habitat associations (Section 6) actually recommend minimum
sample sizes. The second is that a pilot study and further examination of the
data when they come in can be invaluable in determining how the project
develops. For instance, by plotting the rate of increase of sample size for one
of your key species, you can estimate how long it would take to get the
minimum recommended sample size. If this time is beyond the study period,
you could either accept that you will never get the optimum sample and
concentrate on the study's other aims, or redesign your fieldwork to try and
get bigger samples, e.g. perhaps you could spread your sampling over a wider
area. The sample sizes you aim for and ultimately accept as being adequate
will depend upon the aims of your project. Sometimes, an order of magnitude
for a density or diversity estimate will satisfy your aims so you may only need
a small sample and a fairly imprecise estimate before moving on to another
area. As mentioned in Section 1, as few as ten contacts with a species can be
enough to make some sort of estimate of its abundance. It may not be a
particularly precise estimate, but it may be adequate to fulfill your particular
aims.
Replication of your sampling is a way of increasing the reliability and

general applicability of your results. There are two sorts of replication: one
involves resampling the same sites; the other replicates the whole study at
another site. Resampling the same sites would involve repeating the censuses
you have carried out at particular point counts or line transects. These
replicates have to be treated as such, rather than as independent samples. This
type of replication can be a good way of getting more precise information in
restricted areas and increasing sample sizes for density estimation (see
Section 3). It can also be organised in such a way as to allow you to check for
bias and consistency. If you have time, it is always a good idea to do at least
some replicates.
Replication in its other sense can be illustrated by the following example.
One of the aims of your project might be to compare the species diversities of
three forest habitat types in a protected area. You design and carry out the
study in the appropriate way, and you may even be employing the type of
sample replication outlined above. After analysing the data, you may be able
to say that the avifauna of habitat x is more diverse than that of habitat y or z.
This might well satisfy your particular aims relating to the management of
that particular area. If your aims are actually more general (e.g. is habitat x
always more diverse than y or z?), then it would be unwise to base your
conclusions on the results from just one site. Replicating the whole study at
other places where habitats x, y and z occur together would confirm whether
there are consistent differences between the habitat diversities, or whether any
differences are site-dependent. The decision to replicate individual samples or
the whole study in different areas depends upon your particular aims, but such
replication is a powerful tool and one which is too often neglected in
conservation studies.
2.2.4 Positioning your sampling effort
Habitat heterogeneity
In the discussion of sample sizes above, it has almost been assumed that it is
the total sample size which is important. What is more important is the
sample size within each sampling unit or habitat. Combining data from
different habitat types will provide bigger sample sizes, but if birds are not
distributed similarly between them, you will get biased estimates and
problems with precision and accuracy.
The first step is to establish how many broad habitat types you have in
your study area. You can never do this in an entirely satisfactory way, as you
will not know the important habitat divisions and gradients as far as the birds
are concerned. Nevertheless, even the broadest distinctions will reduce bias
Bird Surveys 27
and hopefully increase the precision of your final estimates. The main habitat
types can be identified from standard maps, aerial photos, satellite maps, pilot
surveys and local knowledge. Since the main habitat types are not necessarily
the smallest sampling units, it is probably wise to treat areas with the same
habitat, but which are geographically distinct, as separate sampling units. It is
always possible to examine the data from the two areas at a later date, and if
there are no obvious differences, combine them for further analysis.
Once the smallest sampling units have been identified, you can then
sample adequately within each unit. What constitutes an adequate sample
depends on the aims of the project, the degree of natural variability present
and the methods you are employing, but a rough guide would be to aim for
about 50 point counts or 10km of line transect within each sampling unit.
If a main aim of the project is to establish habitat associations and
preferred sites, it is essential to sample over as complete a range of habitats
(and usually altitudes) as possible, and to include areas where the target
species may be rare or even absent.
Positioning of sampling sites
If your sampling sites are specific points, then the best way of positioning
those sampling sites within the sampling units is probably through a stratified
random technique. This involves dividing up the study site with a grid, either
on a map or actually on the ground with markers, and then using random
coordinates to position the sampling site within each grid square (see Figure
9). Unfortunately, in many cases there will neither be an adequate map nor the
time or resources to set up a grid on the ground (although the latter is
undoubtedly the best option for longer-term studies). It may be possible to
position sampling sites randomly in other ways, but for safety reasons this
may not be a good idea it is easy to get lost if observers are leaving paths
and searching for randomly selected points.
If you are going to sample by continuously walking along transects, you
could start the walks from points which are determined randomly or
systematically, and the direction you walk could be random or systematically
organised. If you adopted either of these approaches in a forest habitat it
would take a lot of time, effort and habitat destruction to cut the trails you
needed. Although this might be the best approach for a longer term study, in
practice you may have to compromise and follow existing paths, stream beds,
etc. You will obviously save time doing this, but you are almost certain to get
biased data. The clearance and continued use of paths and the presence of a
stream or river is bound to have an impact on the surrounding vegetation. The
initial siting of a path is also not likely to be random with respect to

topography and vegetation.
By just collecting data along existing paths, bird and plant communities
characteristic of forest edges will be over-represented in the data collected.
One compromise would be to census along existing paths, but to make short
forays (for line or point counts) away from the paths at randomly determined
intervals. You could then compare the data on and off paths and assess how
biased your total data set might be.
If you are adopting a non-random approach to positioning sampling sites
you must be fully aware of the potential bias in the results. A useful approach
is to identify the highest environmental gradients within your study area (e.g.
low to high altitude or open to closed canopy) and deliberately sample across
those gradients. Observing and analysing the trends along the gradients will
help you understand the biases within your whole data set.
Figure 9. Positioning point count sites or the beginning of transects.
a) Along paths or rivers
b) Random
c) Stratifiedrandom method
In (a) point count sites have been positioned along paths or rivers. This has
the advantage of easy access and relocation but only parts of the study area
have been sampled and edge habitats are likely to be disproportionately
sampled. In (b) the point count sites have been chosen randomly which has
the advantage that the study area will be more evenly sampled. However, it is
Bird Surveys 29
possible that a completely random choice might leave some areas undersampled (indicated by the shading in (b)). The best method is to use a
stratified-random method to place the point count sites. The first step is to
superimpose a grid onto the study area. This could be done on a large-scale
map and interpreted on the ground with a Global Positioning System, or
actually marked on the ground itself. For the latter you only need to mark
one corner and decide upon the compass orientation of the grid. The distance
between the grid lines will depend upon the distance over which birds can be
detected but you will probably need at least 500m. Within each grid square
one or more positions can be selected using random numbers. These
positions could be used as point count sites or as the starting points for line
transects (which can then all proceed in the same direction or a randomly
determined compass direction). If a particular grid square encompasses
ground outside of the study area, keep taking pairs of random numbers until
the site indicated is within the study area. Similarly, if sites chosen in
adjacent squares are very close (and the same birds could be recorded from
two points) only 'accept' the random numbers if they place the sites more
than a set distance apart. This distance depends on the distance at which
birds are detected but in forest this may be 150 to 250m.
2.3 Pilot surveys and training

It is obvious from the previous section that, in order to get reliable estimates,
you need a carefully designed study and well trained personnel. Many aspects
of study design and part of the training can be accomplished before beginning
the fieldwork, but much still needs to be done when you arrive at the field
site. You may need to allow at least two weeks for further training and a pilot
survey to refine the study design (and even the aims of the project). At this
stage, liasing with or actually employing local experts becomes particularly
important.
The first stage of the pilot study is to familiarise yourself with the bird
fauna and the habitats:
obtain as much local knowledge as possible on the distribution of key

species and habitat types;
make detailed notes on sightings and calls, comparing them to taped calls
if available;
compare and discuss identification problems between observers;
start a daily log of bird records;
begin to practise distance estimates (if these are appropriate for your
study);
start tree species identification or classification (if appropriate);

start to plot habitat boundaries, likely census routes and other important
features on available maps, or begin to construct your own maps;
decide upon the smallest sampling/habitat units.
For the next stage it is a good idea to set up one or two census routes
(either for line transects or point counts) and select one person to organise
and monitor the performance of the rest of the team. If it was not obvious
before the fieldwork, it should by now be obvious who the best bird
identifiers are, so agree a division of tasks among the group:
get the bird recorders to repeat the same census routes are they
recording similar numbers of contacts per species? If not, get them to
census together to sort out any problems;
monitor the proportion of unidentified bird contacts over time is this
proportion declining fast enough? If not, put more work into species
identification problems or redefine the aims;
compare the performance of different team members in estimating the
distance to known objects who is the most accurate, who are over- or
under-estimating? Is more practice required? Should only certain
individuals be allowed to make the estimates?
similarly, compare abilities to estimate tree heights or identify important
habitat features;
identify and agree reference points for particular habitat variables, such
as canopy cover;
repeat your practice census routes at different times of day, or monitor
activity for long periods from vantage points, and decide upon the times
of day when data collection will take place;
for the key species, monitor the initial census results and predict how
long it will take to reach the required sample sizes;
for each key species, calculate the mean and standard deviation of the
number of contacts per point count or section of line transects. For the
species with the largest standard deviations (the less precise ones), plan
to get larger sample sizes (likely for flocking species or those with
uneven distributions within the same habitat);
calculate some species discovery curves (see Section 5) to estimate how
long you need to stay at sampling stations and within sampling units.
At the end of the pilot project, it should now be possible to decide the
following:
the positioning of the sampling sites (point counts or line transects);
Bird Surveys 31
the make-up of sampling teams and the division of responsibilities within

the whole group;
whether any of the project aims have to be refined (e.g. you may have
discovered that you do not have enough time to get reasonable data from
six sites, so perhaps plan to visit the four most important ones).
It is also a good idea at this stage to produce a standard sheet for

recording the data. A generalised data sheet should have been designed and
copied before getting to the field site but you may need to refine it now. Do
not allow team members to use their own notebooks for collecting census
data. Without the appropriate headings (and reminders), you always lose
information.
Once serious data collection has started, the procedures adopted during
the training period should not be completely abandoned. You should still
monitor things like the proportion of unknown contacts, distance estimating
abilities and how sample sizes and levels of precision are developing.
Monitoring sample sizes is an important factor in deciding if and when to
move to a new field site.
2.4 Safety, team size and logistics

Having designed the study and trained your team appropriately, you may still
have to make concessions for health, safety and logistical reasons.
2.4.1 Health and safety
For safety reasons (and for the division of data collection tasks see What
size team? below) data collection teams should always comprise of at least
two, preferably three, people. Potentially dangerous areas should be avoided,
however interesting they look you cannot concentrate on looking for birds if
you are watching every step or hanging onto a steep slope. Everyone needs
time to rest, so do not plan for everyone to collect data every day; you have to
be fit and alert during data collection. Be prepared for the fact that most team
members may be ill for at least some of the time, so build plenty of 'slack' into
your study design.
2.4.2 Logistics
Generally, there is a trade off between the amount of time spent travelling and
collecting data. There may be lots of potential places to visit, but good data
from a restricted area may be better than poor data from a wide area. What
you choose to do depends upon the aims of the study: looking for a rare and
little-known species may require you to cover a lot of ground; getting precise
density or diversity estimates, or detailed habitat association data, usually
means more time spent in fewer places. If you can afford to do so, hire people
to do as much of the ancillary work as possible. If much of your equipment is
carried for you, and you have cooks to buy and prepare food, you will be able
to put more of your effort into the data collection. Such local collaboration
will pay other dividends (see Section 7).
2.4.3 What size team?
There are two aspects to this question. The first concerns the size of the data
collection teams who are actually carrying out the censusing, and the second
concerns the size of the whole project team. The advantages of having a large
group are that you can collect more data; you are covered for illness; you can
involve more local collaborators in the project; it is probably cheaper per
individual and, having arranged visas and permits and transport, why not take
as many people as you can? The main disadvantages are that it is logistically
more complicated to move and feed large groups, you may have a larger
negative effect on the local environment and you may need particular
experience and skills for organising a big group.
For data collection itself, teams of three (two observers and a data
recorder) are probably the best for the following reasons:
the two observers can concentrate solely on identification and distance

estimation;
identification and distances can be compared between the two, so
individual differences can be evened out;
the recorder can check that all the information for each contact has been
provided;
the recorder can be another check on the distance estimates and can make
sure the observers are concentrating!
there are safety advantages of having three following an accident one
person can go for help whilst another can provide first aid.
The disadvantages of having three people per team, rather than one or
two, is that larger groups make more disturbance and it is obviously less
efficient than having teams of two potentially you will be collecting fewer
data. Obtaining complete and less biased data in a safe way is probably the
more important consideration, but if you do have teams of two, make sure that
an experienced person is combining the surveying and recording jobs.
Whatever size you do start off with, you should maintain it for the whole
study period.
Although three is arguably the ideal size team for data collection, it is too
low for the group as whole. Having only three would leave little time for
Bird Surveys 33
collecting habitat and other useful data, with no scope for domestic logistics
and health problems. Four or five would be a more realistic minimum team
size.
2.5 Which methods to use?

2.5.1 Introduction
The methods adopted will depend upon the aims of the project. As a general
rule, adopt the simplest methods which satisfy those aims. The more complex
methods will usually be more demanding in time and in the statistical criteria
you have to satisfy. It is better to get reliable data using a simple method than
unreliable data from a complex one, even if the latter (potentially at least)
could provide more information. Another reason for adopting simple methods
is that these are more likely to be repeatable. If you hope that others will
repeat your work in the future, perhaps as part of a long-term monitoring
programme, do not assume that they will have the same level of training or
put in the same amount of effort as you. Local conservation workers often do
not have the time to repeat complicated surveys, so keep the aims and
methods as simple as possible.
Different methods apply if you are interested in bird diversities, species
densities or habitat associations. For bird diversity and bird/habitat methods,
go straight to Sections 5 and 6 respectively. For bird densities, a number of
decisions about study design have to be made before moving on to Section 3.
2.5.2 Bird densities
The basic decision to be made first is whether you want relative or absolute
density estimates (the difference between them was outlined in Section 2.1.4).
If you want to know if numbers of a particular species are increasing or
decreasing, or if you want to compare the birds in two areas of similar habitat,
then relative estimates may satisfy your aims. In generating these relative
estimates, you would need to standardise your methods and get as precise
estimates as possible. Potential causes of bias should be identified, but you
may decide to ignore some of them as long as the same bias is present in all
the areas you might be comparing.
Relative estimates do not allow you to make comparisons between
species. This is because different species have different levels of
conspicuousness (or call output). The same problem exists when comparing
the same species between different habitats the species may be more
conspicuous and therefore appear to be commoner in one habitat than
another, when in fact the only real difference is that it is easier to record in
one of the habitats. It is also in the nature of relative estimates that you cannot
derive population density or sizes from them.
Distance sampling (Section 3) involves estimating distances to bird
contacts and, theoretically at least, provides absolute density estimates from
which you can derive population sizes for particular areas. In practice,
because of all the sources of bias which can affect accuracy, we can not easily
know how close our 'absolute' estimate is to the real figure. However, a very
important extra reason for using a distance sampling method is that it allows
for different levels of conspicuousness between species, and between
different habitats occupied by the same species. At least you will then be able
to say that one species is probably more common than another, or that a
species is commoner in habitat x than it is in habitat y.
2.5.3 Point counts and line transects
Throughout this section two different methods of censusing have been
mentioned point counts and line transects. The former involves walking to,
and usually marking, a particular spot, and then recording all bird contacts for
a pre-determined period (often 5 to 10 minutes) before moving on to the next
point. Line transects involve the observer continually walking and recording
all contacts either side of the track walked. The precise details including
how long you should stay at a point, how far the points should be apart and
exactly how you collect the data on line transects, etc., are discussed in
Section 3. Whether you adopt point counts or line transects depends upon a
number of factors. The advantages of each method and, implicitly, the
disadvantages of the other method, are given below.
Point Counts:
concentrate fully on the birds and habitats without having to watch where
you walk;
more time available to identify contacts;
more likely to detect the cryptic and skulking species;
easy to relate bird occurrence to habitat features.
Line transects:
cover ground more quickly and record more birds;
less chance of double recording the same bird;
good for more mobile, more conspicuous species and those which flush
easily;
errors in distance estimation are less serious than for point counts (see
Section 3 for explanation).
Bird Surveys 35
If you are targeting a few species which are relatively easy to identify but
which may be mobile and occur at low densities (usually larger species, such
as parrots), line transects are undoubtedly better. If you are censusing a larger
element of the bird fauna and especially if the species are small, flocking and
difficult to identify, then point counts are better. There are many studies, of
course, for which the choice is not straightforward and perhaps neither
method is ideal. Section 4 gives more information on what to do with these
difficult species.
Section 3
ESTIMATING BIRD DENSITIES USING

DISTANCE SAMPLING
Huw Lloyd, Alexis Cahill, Martin Jones and Stuart Marsden
3.1 Introduction
In Section 2, the distinction was made between censuses that provide a
relative measure of bird abundance (e.g. numbers encountered per hour or per
km) and those that produce an estimate of bird density (number of birds per
unit of area). Of course, estimates of actual bird density are only needed if the
aim of the study is to produce them, to use density data to calculate total
population estimates, or to relate your figures to those of past surveys where
density estimates were calculated. However, these are usually very good
reasons for using distance sampling and it should be remembered that it often
takes little longer to collect distance data than it does to collect data using
other methods. What does take time is the planning and practice needed to
collect reliable and meaningful data.
The general way of producing density estimates is through distance
sampling (other methods are outlined in Section 4) and this can take place on
point counts or line transects. The crucial part of the method is that an
estimate is made of the distance from the bird contact to the centre of the
point count site or to the line which a transect walk is following. These
distance estimates are used to calculate bird densities and, of particular
importance, they take account of the fact that some birds are detectable over
much greater distances than others, and that a species may be more easily
detected in one habitat than another. Thus, even if calculating total population
sizes is not the main aim of the project, collecting the distance data will allow
you to make direct comparisons between species and between the same
species in different habitats. These are comparisons which may not be
possible with encounter rate or other relative density estimation methods.
There are four basic assumptions of distance sampling that should be
adhered to if an unbiased density estimate is to be obtained:
transects or points are representatively placed with respect to bird

density;
objects (birds) directly on the line or at each point are always detected;
objects are detected at their initial location prior to natural movement or
movement in response to the observers presence;
Bird Surveys 37
distances should be accurately measured (or at least estimated with small

and random error).
This section aims to provide a basic understanding of distance sampling

methods using both line transects and point counts, and also to show how a
study using distance sampling should be designed to meet the four critical
assumptions listed above. Once the relevant data have been collected, it is
possible to calculate approximate density estimates with a calculator but
recently a computer program has become available which produces the
estimates in a more sophisticated way. This program is called DISTANCE
and is freely available. The program has a companion book called Distance
Sampling (Buckland et al. 1993). In this section we will explain how to use
the software to analyse your data in what we suggest is the most appropriate
way. Following this introduction, and certainly if you intend to publish your
findings in a scientific journal, we strongly recommend that you get hold of
the distance sampling book and the manual (Laake et al. 1994) which
accompanies the DISTANCE software.
First, we introduce the two main methods of collecting distance sampling
data, using line transects and point counts and discuss some of the options
available to minimise problems which can arise during data collection. Next,
we discuss some of the problems of calculating density estimates from the
data you have collected. We introduce the workings of the DISTANCE
program itself and include some example syntax and output from the distance
program (Section 9).
3.2 Distance sampling using line transects

3.2.1 Introduction
The choice between line transects and point counts has already been
considered in Section 2. To summarise, line transects may be better for lower
density, more mobile species in fairly even habitats. Point counts are better
for skulking species or for censusing larger numbers of species and for work
in fine-grained habitats.
3.2.2 Positioning of transects
It is best to site the start of the transects randomly or through a stratified
random technique (see Section 2). This is one of the four basic assumptions
of distance sampling: line transects that are randomly placed with respect to
the distribution of birds are more likely to produce unbiased density estimates
which can be extrapolated to other areas of the same habitat type. If the
location of line transects is chosen subjectively, or for the observers
convenience (e.g. along trails or in an area which appears to contain high

numbers of birds), the sample obtained is only strictly representative of the
area surveyed. Usually, for logistical and safety reasons, transects are not
randomly situated and it is important to be aware of how this may bias the
results. Walking transects along large rivers or wide trails/roads may be a
particular problem as the vegetation to each side of the transect route may be
highly uncharacteristic of the forest as a whole (see Section 2).
Sometimes transects are laid out in grids which are oriented to a contour
or obvious feature in the landscape, such as a road or a river. Using such grids
may not provide a random sample but it may be fairly easy to identify and test
for causes of bias (e.g. transects near to rivers can be compared to those
further away). They may also be very useful for long term studies where
population changes are monitored at one site.
3.2.3 How many transects and how long should they be?
The total length of line transect in a study depends upon how long it takes to
get an adequate sample size for the target species and how many habitats are
to be sampled. At the end of your pilot study, you should be able to predict
how long it will take you to collect enough data and therefore how many
kilometres of transect will have to be walked. The longest transect walked in
any one day is not likely to be more than 10km. This is because censusing is
often restricted to periods of high bird activity, and the quality of the data
collected will decline as the observers begin to tire. If you need precise
estimates in well defined areas or habitats, it might be better to do many short
transects of, perhaps, around 4km. It then becomes easier to avoid some of
the bias related to time of day.
Each transect can be partitioned into distance intervals along its length.
For example, markers every 50m along a transect can help the observers to
follow the correct track and also allow habitat information to be collected for
specific sections of the transect. The habitat data can then be related to the
occurrence of bird species at particular sections of the transect (see Section
6).
3.2.4 Collection of data
Once transects have been selected, data collection can begin. The design of
the study and the methods employed should now be relatively clear. A poorly
designed study will not only lead to unreliable results but also problems with
using the DISTANCE computer program.
On each transect, the observers walk at a fairly constant speed, looking
either side of the line walked, and estimate the perpendicular distance from
Bird Surveys 39
the line to each bird contact. There are two ways of estimating the distance: 1)
you can make a direct estimate of the distance between the bird and the line,
or 2) you can estimate the distance between the observer and the bird, and the
angle of the sighting away from the line. These methods are illustrated in
Figure 10.
Figure 10. Distance estimation/measurement along transects. Either the

perpendicular distance (d1) from transect line to object is estimated or
measured, or d1 is calculated using d2 and the sighting angle (d1 = d2 . sin
()).
A critical assumption of the method is that all birds at distance 0m are

detected. This can be a problem if there is a dense and high forest canopy and
under these conditions perhaps one of the observers should concentrate solely
on the canopy. It is also important that the observer does not flush birds from
or onto the line transect ahead. Although this is an important assumption of
distance sampling, it can sometimes be a difficult problem to overcome in the
field. More commonly, birds will be flushed away from you, so keep an eye
on the line of travel ahead of you and try to record the positions from which
the birds are flushed.
Distance sampling methods aim to produce a snap-shot of all the birds
recordable from the transect line. This creates a problem for the recording of
flying birds (i.e. those not seen to leave the immediate area of the line
transect), as it is impossible to know if those birds are normally part of the
population of that area. Although it is worthwhile recording these
observations, they should not be used in the calculations as they would
produce overestimates. Leaving them out might actually cause an
underestimate but the error will almost always be much smaller. Remember
that if birds are seen to take to the air, then these birds should be included in
the count and an estimate of distance is made from the take-off point
perpendicular to the line transect.
A distance estimate and a count of the number of birds in each contact are
all that are required to calculate density but it is also useful to record the
following:
the sex of the individual birds (if possible);
the type of contact, e.g. was it a visual sighting or was the bird singing,
calling, or flying?
the time of day of each contact;
the height of the bird e.g. ground, low, mid-strata or canopy.
This information can often throw light on the biology of target species,
and is also useful when it comes to analysing and interpreting the results. For
example, if for one species all the males are singing contacts and the females
sight-only contacts, it is probably a good idea to carry out separate density
estimates for each sex.
Example data collection forms for the variable distance line transect
(VDLT) method are shown in Figure 11.
Figure 11. Example data collection forms transect methods.
Bird Surveys 41
3.2.5 Double counting

Counting the same bird twice can have important consequences, but as long
as the detection is not within the same sampling effort (i.e. along the same
line transect), then double counting will have a minimal effect on density
estimates. Also, there is no problem if a particular bird is stationary and is
detected from two different line transects. It only becomes problematic if that
bird moves from one line transect to another after it has been initially
recorded. It is obviously important to keep a mental note of bird movements
and try to avoid double counting, particularly within the same line transect.
The method assumes that the distances to bird contacts are accurately
measured or that they are estimated with only small and random errors. It is
particularly important for contacts near the line to be estimated correctly.
Large errors or consistent over- or under-estimates will seriously bias the
estimates produced by distance sampling. The importance of adequate
training in distance estimation has already been emphasised in Section 2.
Estimating exact distances to individual bird contacts perpendicular to the
line transect is, statistically, the most robust approach for distance sampling
along transects and it is the one we would recommend. However, estimating
exact distances to bird contacts can be difficult particularly for bird calls in
dense habitats. An alternative method is to employ fixed-width transects,
where birds are recorded within just two or three belts of fixed distance either
side of the transect. Using the Fixed-width Line Transect method (FWLT), all
birds are counted along the guidelines of the normal line transect method, but
each detection is attributed to a distance belt. With this method, errors in
distance estimation will only have an effect if the contact is assigned to the
wrong band, whereas with the VDLT method all errors have an effect.
Another potential source of bias with the VDLT method is that when trying to
estimate exact distances there is often a tendency to round off estimates to the
nearest five or ten metres. A quick examination of data collected in a pilot
study will show if this is happening. If it is, you can either try to be more
exact in your estimations or adopt the FWLT method.
Two distance belts is the minimum required for density estimation but it is
better to have more and it is usual to vary the widths so the bands closer to the
line are narrower. You could, for example, have 5, 10, 15, 20, 30, 40, 60,
100, 200+ metre limits. In dense habitats where most of the bird contacts will
be close to you, it is better to have narrower bands. The more bands you have,
the better for the analysis, but the greater the problem of assigning the bird
contacts correctly. If you have only two bands, the inner band should include
at least 50% of all contacts in order for you to get reasonable estimates.
3.2.7 Distance estimations to groups
Sometimes it may not be possible to estimate distances to all individual birds.
Populations of many species naturally aggregate into flocks or clusters. If this
is the case, and you are using the VDLT method, distance estimations should
be made to the geometric centre of the cluster. If using the FWLT method, all
members of a cluster are assigned to the distance band which encompasses
the centre of the cluster. If a species is known to always occur in flocks at the
time of the census there is an extra problem: a flock may be contacted through
calls but not seen and the observer may not know how many birds are present.
In these cases, it is normal to substitute the mean flock size for the visually
recorded flocks. See Section 4 for more information on problems associated
with flocking.
3.2.8 Sample sizes
Sample sizes for line transect distance sampling data have to be quite large.
Small sample sizes contain little information about density and their precision
is poor, regardless of the analytical method used. An ideal minimum should
be approximately 6080 records but an estimate (albeit an imprecise one) can
be calculated with fewer observations. If birds are clustered, the sample sizes
would have to be even larger.
3.3 Distance sampling using point counts

3.3.1 Introduction
The difference between line transects and point counts is that, for the latter,
an observer stands still in one particular location (a census station) recording
all the birds seen and heard during a fixed count period. Point counts are
often preferred to line transects when surveying less mobile bird species, and
in more fine-grained habitats. This is because a randomly placed transect
route might only pass through two or three habitat types in an area which has
many more. Census stations which are randomly or systematically allocated in
the same area are more likely to sample a wider range of the habitats present.
Also, if detailed habitat associations of bird species are an objective of the
study, habitat data can be recorded around each census station and can be
easily associated with the presence/absence of individual bird species (see
Section 6).
Point counts are also preferred to line transects in closed forest habitats
with high canopies, particularly rainforests. This is because by standing in
Bird Surveys 43
one location over a fixed period of time, an observer has a better chance of
detecting birds than if he or she is just walking through the area. Laying out
point counts systematically along transects also overcomes the problem of
trying to walk and survey birds in very difficult and uneven terrain: at the
census stations, time is spent trying to find birds rather than watching the path
of travel (although the time walking between census stations is then lost).
3.3.2 Location of census stations
As with the siting of transect routes, point count sites should be positioned
randomly within your sampling units or habitat types. In order to get adequate
coverage in each unit, you could adopt a stratified random technique as
outlined in section 2.2.3. The problems with a random placement of sites are
logistics and safety. In some areas and habitats, it might be difficult and time
consuming to get to all of the sites and there is a danger of becoming lost. If
point count sites are positioned along transect routes, time is used more
efficiently, but you must be aware of the bias that might be caused by
sampling sites in a particular order and along habitat edges (see section
2.2.3). A practical way to position point count sites is to set them out along
transect routes which follow trails or streams but to place each site at some
distance perpendicular to the transect route itself. Jones et al. (1995) used this
type of procedure: every other census station was placed 50m to alternate
sides of the transect route.
Another important consideration is the spacing of census stations. If
census stations are too close together, birds can be recorded from one station
and then have a good chance of flying the short distance to the next census
station. If stations are too far apart, then too much time is wasted walking
between them. As an approximate compromise, the minimum distance
between stations in dense forests should be 200 to 250m. If the study focuses
upon small, fairly sedentary and inconspicuous birds, the distance can be
smaller (e.g. 150m). For larger, more conspicuous and more mobile species
and particularly for studies in open habitats, the distances should be greater
350 to 400m is not unusual. A final decision on the spacing of stations should
be made at the end of the pilot study, once you have experience of the
distances over which individual birds can be recorded.
Spacing out point counts is easy if you are siting them along transect
routes; it is more difficult if you are aiming for a random distribution as, by
chance, two stations could be sited very close to each other. In these
circumstances it is best to constrain the randomisation process, such that
stations are sited randomly but, if any two are within the minimum set
distance, a new set of coordinates is allocated for one of the stations. This is
repeated until all stations are more than the minimum distance apart.
3.3.3 How many census stations?
This will depend upon the sample sizes required for each target species, and
can be predicted from a pilot survey (section 2.3). You will need a minimum
of about 50 point counts to sample the commoner species within a sampling
unit (a single habitat type at one site) and to describe the bird community of
the habitat. For rare species, very many point counts are needed to amass
enough bird records to produce precise estimates, simply because the species
is not recorded at the great majority of the point counts. The precision of the
density estimates can be increased by repeated data collection at census
stations (see section 2.2.3), but this is obviously at the expense of the area
that could be covered during a survey.
3.3.4 How long should the count period be?
This is a difficult problem. The ideal scenario is to have an instant picture of
all of the birds at or near the station. In reality it takes time to detect and take
details of all of the birds at the station. Even large and colourful birds may
only be detected if they move or call, while cryptic birds and those high up in
the canopy, may take even longer to be detected. It is a critical assumption of
distance sampling that all birds at distance 0m should be detected, and it helps
if there is a near-certainty of detection for some distance from the census
station. Staying longer at a station should increase the chance of detecting
birds but we then come up against another important assumption of the
method, namely that individual birds are not counted twice (at least during the
same point count). The longer the count period, the greater the chance that a
bird would be counted twice or, just as importantly, a bird could move
undetected into the sampling area from outside. Both of these circumstances
would lead to an overestimate of the number of birds using an area at an
instant in time.
Most studies use a count period of between five and ten minutes; the more
mobile and conspicuous your target species, the shorter time you should use.
For multi-species surveys, where different periods would be appropriate for
particular groups of species, you could adopt a longer period (e.g. ten
minutes), but record the time each bird contact is made. This enables you to
use, for example, the first five or six minutes for the more mobile species (for
which double counting could be a problem), and the whole ten minutes for the
more cryptic and sedentary species. Using more than ten minutes will not
usually be necessary. More information on counting periods appropriate for
different types of species is given in Section 4.
Bird Surveys 45
3.3.5 Data collection

The variables recorded are almost identical to those recorded for line
transects (see Section 3.2.4). Before the observers begin to record birds at a
census station, it is a good idea to wait a few minutes so the resident birds can
settle down after the disturbance produced by your arrival. Once this period is
over, the observers stand still at each station, record their start time and then
estimate distances to all bird contacts. Remember that you should estimate the
distance of each contact to a designated point and not to the observers, who
may not be standing on that exact point. It is often useful to record the exact
time of each contact or to assign them to a one or two minute block of time.
Information on sex, type of contact, height of contact in the foliage and group
size can be recorded in the same way as with line transects. Birds that fly
away from the immediate area are recorded and a distance estimate made to
their point of departure. This also goes for birds flushed as you arrive at the
station. Birds that fly into the area and land, or fly over the area, can be noted
but should be excluded from the data analysis. Double counting of birds has
the same consequences as that stated for line transects in section 3.2.5.
As with line transects, it is important that distances should be estimated
accurately or with small and random error. Density estimates generated from
point counts are even more susceptible to bias arising from inaccurate
distance estimations than those calculated from line transect data. This is
because the total area surveyed using point counts is proportional to the
square of the distance from the observer. With line transects, the area
surveyed is only linearly proportional to distance from the observer. This
places even greater importance on the need for accurate distance estimation
and adequate training before the real survey begins.
The best point count distance sampling method involves estimating the
actual distance to each bird contact, and this is often called the Variable
Circular Plot (VCP) method. Contacts can also be assigned to distance bands
in the same way as outlined for line transects in section 3.2.6, and the
procedure adopted for recording distances to flocks of birds is the same as
that outlined in Section 3.2.7. An example data collection form for the VCP
method is shown in Figure 12.
Figure 12. Example data collection form variable circular plot method. In
this example, the total count period (10 minutes) has been divided into five
two-minute periods.
3.3.7 Sample sizes
Sample sizes for point count data must be larger than corresponding ones for
line transect data to get the same degree of precision. Ideally you should aim
to accumulate 80100 contacts for each species in each sampling unit. It is
possible to calculate estimates from much smaller samples, but these will be
less precise. Again, it is important to define the level of precision you need
from your density estimates before you start the survey. This will help you to
ensure that your estimates are precise enough to detect density differences
with confidence, but to avoid spending time in collecting extra distance data,
when you could be collecting other data.
3.4 Examining the data

3.4.1 Bird detectability and the detection curve
Whether one walks around a forest or stands at particular locations, an
assumption of distance sampling is that all birds at a distance of 0m are
recorded. Usually all birds at some distance away are also recorded, but as
distance increases there is an increasing likelihood that birds will be missed.
A typical fall-off in detection with distance is shown in Figures 13ab (for
an African hornbill). These are data from a line transect where the numbers of
bird contacts in each distance band have been totalled (distances either side of
the line have been combined). Note that the two histograms are the same
shape (each distance band has the same area). The birds could have been
assigned to these bands during the fieldwork, or distances to each bird contact
could have been recorded exactly, but allocated to the distance bands
afterwards (this is done for the sake of the demonstration; they are kept
separate for the actual analysis). As we would expect, more contacts are made
closer to the observers and the numbers tail off with distance. Figure 13b is
equivalent to the 'distance function' or 'detection curve' for species x, and
describes its detectability for that particular habitat.
Bird Surveys 47
Figure 13. Bird detection curves from transect and point count methods.
Histograms (a) and (c) show the actual numbers of birds recorded in each
distance band, while (b) and (d) show the density of birds recorded per
distance band (i.e. number of birds divided by the area within the distance
band and expressed as individuals per km2). Note that histograms (c) and (d)
are different shapes because the area within distance bands increases
exponentially with increasing distance from the recorder.
a) Transects birds recorded

c) Point count birds recorded
b) Transects detection curve

d) Point count detection curve
Detection curves for point count data are similar, but there is the added
complication that the area within each distance band is different (Figure 13c
2
d). The area encompassed by, for example, the 010m band (314m ) is very
2
much smaller than the area within the 2030m band (1,550m ). Therefore, the
shape of the histogram for the number of birds recorded per distance band (c)
is different to that for the density of birds in each band (d). You must
combine data from many point counts or line transects to produce these
curves, but do not combine those from more than one habitat without careful
consideration. Different habitats tend to produce different curves because
birds are more or less conspicuous in them the main strength of distance
methods is that they account for these differences.
What the DISTANCE program does, in effect, is to draw out each

detection curve and then fit a mathematical model to it. The problem is that
there are a number of different models that could be applied and a variety of
ways in which the data could be manipulated to ensure better fit of a model
estimate. Although you do not actually have to produce a detection curve
yourself (the raw data are entered into DISTANCE), we strongly advise that
you do this for representative species before using the DISTANCE program.
The main reason is that a number of decisions about data manipulation and
model fit have to be made, and these depend upon an adequate knowledge of
Bird Surveys 49
your data set. Producing histograms of detection curves is the best way to do
this, and the optimal approach is to produce some initial curves during the
pilot study. In this way, you may be able to modify your data collection to
avoid some of the problems discussed below.
3.4.2 Shape criteria
Figure 14a shows a good detection curve it has narrow bands, the number of
contacts remains fairly constant over the first few distance bands (the curve
has a broad shoulder) and there is a smooth and rapidly declining tail. There
are a number of reasons why field data may not approach this ideal shape and
these are dealt with below.
Figure 14. Bird detection curves some problems.
a) Good detection curve with
b) Skulking bird often recorded
broad shoulder and steep tail
on paths
c) Birds move in response to

recorder presence
Bird Surveys 51
d) Outliers. There may also be a

problem with heaping (at 50 and
100m)
Lack of a broad shoulder in the detection curve
This can be caused by missing too many contacts close to the recorder, or by
the bird being attracted to the recorder. This is illustrated in Figure 14b and it
can have serious consequences for the reliability of the estimates produced. If
you have identified the problem during the pilot study you could consider
using point counts (if you were using line transects) or use longer point count
periods. If you stand still for longer you are likely to detect more of the birds
around you and this may produce a broader shoulder for the detection curve
for that species. If all the data have already been collected and a lack of a
broad shoulder is apparent then you could enter the data in distance bands but
manipulate the band distribution and widths to give the best curve.
Higher or lower than expected values at close distances
There can be a number of causes. Birds fleeing from the observer can
produce low values for the closer distances and Figure 14c illustrates such a
circumstance. Studies using point counts are particularly prone to large
fluctuations at these close distances, because the area sampled close to the
recorder is very small. This is one reason why you need more data for point
counts than for line transects. Altering the widths of the distance bands prior
to data entry can produce a better curve and more reliable estimates.
Outliers
Outliers are records of birds detected at large distances from the transect or
census station (see Figure 14d). These add little information about bird
density and make fitting a model more difficult. As a general rule, outliers
should be routinely removed or truncated from the analysis, to enable a
better model fit to the data. How much data have to be truncated will depend
on the actual detection curve, but 5% of all line transect data is an average
figure. A slightly higher percentage (around 10%) of data generated by point
counts is usually truncated because there are a higher proportion of detections
at larger distances, and these will distort the ideal shape of the detection curve
by flattening the tail.
Heaping
If there is a tendency to round off distance estimations to the nearest 10, 20 or
even 50 metres, there may be large heaps of records at particular distances
surrounded by very few records. This problem may have been recognised and
alleviated in the pilot study but if not, grouping the data into different
distance bands would help. The first distance band should be narrow and
should fall within the 'shoulder' whilst the width of the other bands should
increase with distance from the point or transect line.
Cluster bias
This is only a problem if the detection of species is a function of cluster size,
e.g. if observations at larger distances tend to be of larger flocks than those
close to the observers. Detection distance and cluster size should be
independent and drawing some scatter plots and calculation of correlation
coefficients will test this. If the observations at larger distances tend to be of
relatively large flocks it is a good idea to truncate the data prior to analysis to
remove some of the large groups.
Small sample sizes
It can be very difficult to ascertain the shape of a detection curve based on a
small sample size (the DISTANCE software will find it equally difficult). One
option would be to combine data on the same species from different sampling
units, in order to obtain a better detection curve (and eventually a more
reliable estimate). This is only valid if there is good reason to suppose that a
species will have the same detection curve (or level of conspicuousness) in
those different sampling units (for example, the habitats were similar and the
surveys were carried out at similar times of year).
3.5 Using the DISTANCE software

An overview of the use of the software and the different options available
within the program are given below. Some annotated sample inputs are given
in Section 9. For full details, see the program manual (Laake et al. 1994).
Bird Surveys 53
3.5.1 The basic model

In simple terms, the DISTANCE program draws out the detection curve for
each species in each sampling unit and then fits a mathematical model which
describes the data. The most important decision to be made is which model to
fit to the data. The three main models or key functions are called Uniform,
Half-normal and Hazard Rate, and the fit of each can be adjusted by using a
series expansion. By default (i.e. unless you tell the program otherwise), the
Uniform key function is used because it performs well in a variety of
situations.
3.5.2 Alternative models
The basic shapes of the uniform and alternative key functions are shown in
Figure 15. The half-normal key function is sometimes used when the level of
detection declines quickly over distance. In these circumstances, the data are
often not truncated and the half-normal model followed by a series expansion
called Hermite polynomial is applied. The Hazard Rate model is more
effective for data which show a flat shoulder and long flat tail. The Negative
Exponential model with a simple polynomial expansion is occasionally useful
for the analysis of poorly collected data. Whatever model is selected, you
should ensure that none of the four basic assumptions (given on pages 3536
have been broken.
Figure 15. The basic shapes of the uniform and alternative key functions.
3.5.3 Testing model fit
You can fit any of the models to your data but how do you Uniform
test which is the
best fit? Akaikes Information Criterion (AIC) provides a quantitative method
Detection
for model selection (see Buckland et al. 1993). The relative
of the
Hazardfit
Rate
Probability
alternative models may be evaluated using AIC, so that the model with the
best fit and least number of parameters will have the lowest AIC value.
Half-Normal
Therefore, rather than accepting the default Uniform model, you can ask the
program to examine the AIC values for each model fit and calculate the
Negative
density estimates by using the model with the lowest AIC value.
0
20
40
60
80 100
exponential
Just because a model is judged to be the best fit of those possible, it does
not necessarily mean that it Distance
is a close fit or one which will produce a precise
estimate. The DISTANCE program uses the statistic to assess the
goodness of fit of each model. For a number of reasons, it is not a
particularly sensitive test, but when you look at the output from the program a
significant test is a useful warning that the model might be a poor fit and/or
one of the four critical assumptions of distance sampling might be seriously
violated.
3.5.4 Inputting data
The following is a summary of how to input data into the DISTANCE
program. We strongly recommend that you treat this as a very basic
introduction and refer to Laake et al. (1994) and Buckland et al. (1993)
before producing your final estimates.
Basically, the data input can be divided into three sections as follows:
Options
Here you describe the parameters of the census method, i.e. whether the
estimate is based on line transects or point counts, what the units of area are
(e.g. per km or per hectare) and the units of distance estimation (usually
metres).
Data
You can actually enter the data here or you can refer the DISTANCE
program to another file in which the data are stored. The group sizes and
distance to each contact are arranged by sample effort, i.e. per line transect or
point count. You also state here how many times each sample was repeated
and whether the samples are arranged into different strata each stratum
could be a different habitat. Each sample can also be given a label.
Bird Surveys 55
Estimate
This is where you tell the program which model to fit to the data or how it
should decide which model to fit. You can also select whether to have a
density estimate for each transect, for each stratum or for the whole data set.
Some annotated examples of input files are shown in Section 9. When you
try and run the program, it may abort its run for a number of reasons. A
common problem is that a group size or a distance estimate is omitted
because of an error in data input, so check your data very carefully.
3.5.5 Understanding the output
When the program runs successfully, the results are put into a file which
(unless you tell it otherwise) is called dist.out. Much of the output is
concerned with fitting the models to the data. The final section contains the
population estimates with their standard errors and 95% confidence intervals.
Potentially useful statistics in this output are the effective detection distances
(effective detection radius (EDR) in point counts). The EDR is the distance
from the observer, beyond which as many bird contacts are missed as are
actually recorded within the EDR. By comparing the values between species
and habitats you can check to see, for instance, if a species is equally
conspicuous in different sampling units. If it is and it makes biological sense
to do so, you could combine the sampling units and get a larger sample size
and (hopefully) a more precise estimate.
Section 4
COUNTING SINGLE-SPECIES
Stuart J. Marsden
4.1 Is a single-species study appropriate?

The impetus to study a single-species or group will usually be a lack of
knowledge about, or particular concern for the plight of, that species or
group, and/or the funding available from a special interest group. So what is
the nature of a single-species study? On the one hand it is a tailor-made and
concentrated effort with specific aims and outputs. On the other it may be
impractical, unnecessary or a misguided waste of resources. In short, why
ignore 99 records of other bird species for the sake of a single record of one
species? Consider how the data you collect on the single-species fits into
biodiversity thinking you might find the best area for your species but what
about the rest? Single-species studies that can be incorporated into fuller
studies may be best in some situations but if you do choose to study just the
one species then you must choose a method to suit your bird precisely. The
key to doing this effectively is to gather information about your species, its
distribution and likely abundance, then fully understand the pros and cons of
different census methods.
4.2 Narrowing down the search: information is the key

A complete literature search on the species and its relatives, its habits,
habitats and the area to be visited is essential, as is contact with any workers
in the field. For globally threatened species, such information has been, or
will soon be, documented in Red Data Books covering Africa, the Americas,
and Asia. Local scientists may have limited access to such literature but they
can be in a better position to contact counterparts in the study area, local
community leaders and local hunters. Perhaps a brief visit to the study area to
identify possible research sites and methods will be valuable and not too
expensive.
It may be important first to look for a species where it was seen last. If the
species is present there, useful initial experience of its habits can be gained
before searching other areas. If the species is not there any more, then you
will still have some important data (without actually recording the bird). Why
is it absent (e.g. habitat change, hunting) and which other areas might still
support it based on this knowledge? It is important to note that in some
species which have suffered declines through over-harvest or through
Bird Surveys 57
predation by an introduced taxon, the relict distribution may not coincide with
its original distribution or reflect the habitat which it most favours.
Figure 16 shows aspects of narrowing down the search for a species.
Confirming absence or looking for a species in a likely but unknown area is a
sensible starting point. While it would perhaps be a waste of effort to visit an
island or region where the species is unknown (Figure 16a), there are
exceptions. Some species are known only from a handful of old specimens,
some of which can have localities mislabelled. Species can be misidentified in
the field, escaped individuals can be recorded in unnatural areas, and every
year new species and range extensions come to light.
The next scale down is local presence/absence. In the example (Figure
16b), the species prefers higher altitudes. Just as clear is that the positioning
of the sample effort greatly affects the number of birds recorded (A = species
extinct! whereas B = common highland species). Sample effort C is perhaps
the most successful (C = species common at higher altitude, uncommon at
mid-altitude and absent in the lowlands). Local differences in abundance can
also be due to rainfall patterns, longitude and latitude, hunting pressure and
many other factors, and remember that many tropical birds undertake
altitudinal and other local migrations.
The need to look in the right habitat is just as great. In Figure 16c, the
species is present within only a small proportion of the area. Again,
placement of sample effort is crucial, and in the example there is a need to
survey both riverine forest (the species favoured habitat) and the area of
remnant non-riverine forest. Knowledge of whether the species occurs in the
remnant forest may be extremely useful in describing its range and habitat
needs (is it a riverine forest specialist or is riverine forest just about the only
forest left in the area?).
As birds have general habitat requirements, so do they have microhabitat
requirements. You need to get right into the microhabitat where your species
lives. Again, information is the key: does the species nest in dead trees, does
it like the open understorey of primary forest or the closed understorey of
disturbed forest? Some information on your bird's microhabitat may be
gleaned from field guides or experienced birders. If specific information is
not available for your species, then details on related taxa may help.
Remember though, that available details of microhabitat may not represent
the species' true requirements, rather the habitats in which it is most easily
detected or, again, the only habitat left.
Figure 16. Narrowing down the search for a species.

a) Regional presence
b) Local presence
c) Restricted habitat
Bird Surveys 59
(x = species recorded)
Time of day is obviously an important consideration in survey design
(Figure 17). Some windows of study are fairly obvious, e.g. nocturnal owls
are best looked for during the night! Others are not so obvious. Nocturnal
birds may be sought (in their active phase) at night but searches for roost or
nest sites during the day may be just as important (and these require different
methods). When we take account of seasonal time factors, windows of study
can become quite complex (Figure 18). There may be a specific time of year
during which birds sing (some tropical birds may sing extremely
infrequently), or a certain time of day. The timing of your fieldwork and the
methods you choose should reflect these considerations.
Figure 17. Time windows for the study of a resident diurnal forest bird using
a distance method (the species sings between July and September). = good
time, = very good time.
Time/
month
Dawn

Morning

Afternoon
Dusk
Midday
Night
Bird Surveys 61
Time/
month
Dawn
Morning
Afternoon
Dusk
Night
Midday
Figure 18. Appropriate survey methods, time of day and month for a (mostly)
nocturnal owl which calls between March and May, and which has fledglings
in June/July.
1 = transect searches, 2 = playback of call, 3 = roost-site searches, 4 =
search for active adults/juveniles.
4.3. The nature of the beast: rarity

There are many forms of rarity. Absolute rarity means that numbers of a
particular bird are known to be low. With a wild population of only one,
Spix's Macaw is obviously one of the worlds rarest birds. The absence of
actual population figures for most bird species means that relative rarity is
often used. Thus, Species A is rarer than Species B, or common in one region
or habitat but rare in another, or rarer than it was 20 years ago. The above
terms for rarity are valid but some instances of perceived rarity are not. Some
species may be difficult to find for several reasons but this may be very
different from actual rarity. Some little-known species are seen as rare
because previous expeditions have looked for them in the wrong place or
used the wrong methods. There may be a tendency to look upon a species that
has not been seen in the wild for many years as rare: has anyone actually been
to look for it?
Rabinowitz (1981) described three components of actual rarity: small
global range, restricted habitat and low population density. Some species can
suffer from more than one of these, the worst case being of a species with a
highly localised range, within which it occurs in a very specific habitat and,
even in this habitat, it occurs at low density. These are natural ecological
patterns of rarity but they have serious implications for the bird surveyor. One
of the forms, small global range, has been introduced in Figure 16 so lets
presume that you are 1) in the right place, and 2) in the right habitat (the
habitat of a rare species may be poorly known, so you are looking in a range
of possible habitats in which it may occur).
Figure 19 shows some different patterns of distribution/rarity. For a
certain survey effort (which could be one person looking for one month),
the rarer the species, the fewer records the expedition will amass. Sample
efforts A and B are superimposed on the figures. In the case of a territorial
bird (19b), individuals tend to be fairly evenly spread out. Note that the
position of the sample effort does not make such a difference to the number of
birds actually recorded. Increasing the sample effort (say from one to two
months) may roughly double the number of contacts with individuals. In the
colonial or clumped system (possibly the result of restricted habitat), the
situation is very different: (B) records many birds while (A) records none. In
general, the more clumped a species distribution is, then the larger the area
that the sample effort must cover to get a true idea of the average abundance
of the bird.
One type of rarity which is important in bird surveys is that some
species, for various reasons, can be difficult to detect (they can be nocturnal,
cryptic, or they can be disturbed easily). For these species (19d shows a
cryptic, uniformly distributed one) the problem is being able to record the
individuals which are actually present. While it may be best to cover a lot of
ground to census colonial birds, for cryptic species it may be better to
concentrate on a smaller area and make sure that you search well enough to
find most of the birds present. It is very important to understand how your
species fits into these patterns of rarity.
In a multi-species survey, it will usually be unwise to jump from distance
sampling to another technique for the benefit of just one or two rare species.
Alternatives could be to stay longer, or to devise specific/focal studies for the
species as an aside to the distance sampling regime. In single-species studies,
the jump can be just as drastic, so the important thing is to decide fairly early
(during the pilot study) whether the standard method is going to be
appropriate. You should try a distance method and extrapolate the number of
records accumulated over the first days to the number you can expect during
the whole fieldwork period (see Section 2). Can you restrict the survey to the
types of habitats where the bird occurs?
For various reasons, you may not be recording the bird sufficiently often.
If it is a cryptic species, expanding the count period may work. Alternatively,
Bird Surveys 63
the bird's distribution may be so clumped that you are missing the
aggregations or colonies. In this case you need to locate the aggregation and
make a total count or estimate of each aggregation. You may be able to locate
the species more easily when it flies by looking over large areas of forest
from a vantage point. In cases of extreme rarity you may have to use all your
birding skills just to find it, or ask local people. Contact with local people
concerning your species and the project you are doing can be extremely
valuable (see Section 4.6.5). In other cases, you may have to search
specifically for a species and concentrate on its habitat associations (see
Section 6).
Figure 19. Effect of distribution and rarity on survey results.
a) Common species
b) Rare species (uniform)
c) Clumped
d) Cryptic (and uniform)

x bird recorded; o bird missed
4.4 Problem species and difficult habitats

Birds come in all shapes and sizes and their habits and habitats do not always
make counting them straightforward. This is not to say that unusual birds are
always more difficult to work with than ordinary ones. In fact, some
counting methods actually use the unusual characteristics shown by their
subjects to their advantage. Again, information about the natural history of
your subject species is crucial in tailoring methods to suit the situation. Below
are listed some special characteristics shown by birds, some examples, and
the implications of these for censusing. This is followed by more specific
methods to deal with problems that may arise.
4.4.1 Bird colonies/aggregations
The distribution of many bird species will be clumped to a certain extent. The
more clumped, the bigger the gaps will be between bird occurrences (Figure
19c). Clumping can be very pronounced in a number of situations:
Many birds breed in colonies although they may disperse in the nonbreeding season. Colonies may range from just a few pairs to many thousands
of birds. In the former case (e.g. a small colony of starlings or mynas in a
large dead tree), normal census methods may be suitable, but in the case of
single large colonies, specific searches followed by total counts would be
better. In these cases, information from local people can be crucial and not
too difficult to obtain if colonies are spectacular, or if birds, eggs or nests are
harvested. The focus for colonies may be obvious (e.g. caves for swiftlets or
oilbirds; cliffs for parrots and hirundines; hot springs or beaches for
megapodes; riverside trees for waterbirds) or less easy to predict (traditional
nesting grounds for megapodes, or individual trees for weavers).
Many birds will roost in groups for at least part of the year. As with
breeding colonies, both the size of roosts and the focus for the roost (large
trees, caves, cliffs, mangroves, or mudflats) varies considerably. Again, prior
knowledge of where the species may roost and specific information from
local people will be invaluable. In this case the fieldworker has some useful
options; birds can be studied away from the roost (using normal census
methods), as they fly to/from the roost, or actually at the roost. Birds will
often roost by night, but there are other cues, including tidally-driven roost
patterns and periods of post-feeding roosting in frugivorous birds.
Some species, such as manakins and birds of paradise, aggregate around
lekking grounds, to display to potential mates before breeding. Such
aggregations are usually quite small (fewer than 50 birds) and subject birds
may be counted as clusters with distance methods. Leks can vary in their
Bird Surveys 65
conspicuousness to the recorder and in their timing (e.g. early mornings at

certain months of the year). Local people may know of lekking sites (they
may collect birds or colourful feathers). Some lek sites can be confirmed out
of season because birds alter the local vegetation (clearing the ground) or
leave signs (feathers or faeces).
The above cases of aggregation can help the fieldworker to count them.
However, many birds (such as pigeons, finches, tanagers and jays) simply
travel and feed in large groups. Consequently, you may go a long time
without recording any and then stumble across too many to count (i.e.
encounter rate low and group size unknown). In all but the most extreme
cases, this problem is best tackled using standard methods but the following
must be considered. First, you should make sure that you have enough bird
records not the total number of birds seen but the number of groups
encountered. Second, you must be able to predict the number of birds in each
group every time you see a group (on census or not) try to estimate the
group size. You can then substitute the mean group size of these encounters
for those during the census where the group size could not be determined.
4.4.2 Cryptic and understorey birds
For a given population density and search effort, the detection rate of a
species will depend on its conspicuousness. Many of the world's most
beautiful and elusive birds are cryptic understorey or ground dwellers.
Remember, a birdwatcher usually aims to see a bird, whereas the bird
surveyor can record it by sight or sound. Knowing a bird's call with
confidence can increase its encounter rate tenfold or more. Get this
information from literature, tapes, previous visitors, local people, or firsthand during the pilot survey. Once this is known, as long as it is not too rare,
then the species can be censused using a variation of VCP method (section
3.3.6). Alternatively, if the species flushes easily (e.g. gamebirds and many
other ground-dwelling birds), then walking transects may be better. Your aim
could be to count the birds as you flush them (while trying to ensure that you
don't count them twice).
4.4.3 Canopy species
Many rainforest birds use the upper storeys (3070m), making their detection
difficult (especially as some move around quickly in mixed species flocks).
There is no easy way round this problem of detection, although a variant of
distance sampling may again be the best compromise (section 4.5). In a
nutshell, the answer is to find out the proportion of birds actually in the
canopy that you can record from the ground (in distance sampling, the
probability of detecting birds at zero metres <1). Perhaps the most feasible
way of doing this is for one team member to do a normal point count, while
several other recorders stand around him/her looking into the canopy through
gaps and trying to record every bird that is there. If this is done many times
then density estimates from the single observer can be corrected, by relating
them to numbers of birds seen by the other observers. For example, if the
single observer records five birds per point count and the other observers
record ten between them, then estimates should be doubled. Of course, this is
a very approximate method and in some cases it may be near impossible to
know the proportion of birds you are missing. Perhaps the most important
point to appreciate is that the proportion of canopy birds that you detect in a
forest with a low canpoy may be much higher than in tall forest.
4.4.4 Mixed-species flocks
Many species join mixed-species flocks for at least some part of the year. It is
important to distinguish between those, mostly insectivores, which move
around, from feeding aggregations such as several frugivorous species at a
fruiting tree. In some respects mixed-species flocks can be seen as a variant of
single-species flocks (section 4.4.1) but with the added problem that both
species-composition and group size are unknown. Flocks may be rarely
encountered, move very fast through the forest, and contain variable numbers
of individuals and species. For an estimate of abundance for one of its
constituent species, we need to know (1) the number of flocks in a given area,
and (2) the presence and number of target species within each flock.
4.4.5 Aerial birds
Although birds can spend between 0% (flightless) and almost 100% (swifts)
of time in flight, most spend less than 50% of their time flying. Also, in
forests, most birds do their important 'business' (feeding, breeding, roosting,
etc.) when they are not flying large distances. For the great majority of
species, a record of a perched bird is much more important than one in flight.
In most species, flying birds can simply be omitted from distance sampling
estimates. In a minority of very mobile birds (or ones which are cryptic at rest
and conspicuous in flight), however, specific techniques may be the only/best
way to count them. These include swifts, swallows, some raptors, etc.
4.4.6 Nocturnal and crepuscular birds
Birds which are difficult to count during the day are obviously a special case.
Distance methods will probably be unsuitable for fieldwork at night (not least
for reasons of safety). Spot or territory mapping of calls, encounter rates
along transects, or simple presence and absence in different habitats may be
the only way of collecting meaningful data. Marking trees with brightlycoloured tape may allow relocation of sites during daylight. Remember that
Bird Surveys 67
many nocturnal birds prefer edges or clearings so transects which proceed

along paths or roads (rather than randomly through forest) may overestimate
bird abundance. Transects could work if roosting birds or nests are looked for
during the day (several people could walk side by side, blanket searching for
likely cavities or roost sites).
4.4.7 Other problem birds (waterbirds, birds of prey, migrants
etc.)
Some tropical waterbirds can be counted at aggregations but others occur
along watercourses within forest (e.g. forktails, kingfishers, ibises, herons). In
some respects this makes them easier to count (their habitat is almost twodimensional). Line transects along streams may yield density figures of bird
pairs per kilometre of stream. In some cases (e.g. forktails) spot-mapping of
territories works as birds may flush to the edge of their territory and fly
behind the observer. Nests of stream birds may also be relatively easy to find.
Some birds of prey are very difficult to census using distance methods.
Here the answer may be to count the birds as they fly above the forest or to
spot map their nest-sites. Finally, some birds are either only known from
migration, or can only/best be counted as they migrate. Counting large birds
such as cranes or raptors at migration bottlenecks is similar in many ways to
counting birds flying to/from roosts.
4.4.8 Special habitats/niches
Any bird's niche or lifestyle will seem special if you look hard enough at it,
but in conservation studies, some will appear more specialised than others. A
bird which occurs mostly in, say, mangrove, riverine forest or bamboo, must
be looked for mainly (but not exclusively) in that habitat. In many cases, the
census method used can be exactly the same as for general habitats. In others,
such as mangrove, variations such as the use of a boat require only common
sense, as long as the effect of the boat on birds is considered. Other habitats
such as particularly steep terrain or montane areas (where bird density may be
very low) may need specific consideration.
Specialisation may be behavioural or resource-based. In the neotropics a
number of species are closely associated with swarming ants. A few
African/Asian species are associated with bee nests (honeyguides) or with
particular events (e.g. bee-eaters and raptors with bushfires). Some of the
above species can be counted with standard methods, but others are rare and
so specialised that studies must be focused specifically on the resources that
limit their distribution.
4.5 Tailoring distance sampling methods for individual

situations
This section describes how distance methods can be tailored to suit individual
species types. Of course they are only broad recommendations. It cannot be
stressed too much that for your census to be successful, you should take time
to understand the principles of distance sampling, and how the behaviour of
your species fits in with its assumptions (see Section 3). To recap:
your plots or transects should be positioned so as to form a representative

sample of the area under study;
the probability of detecting birds on the transect line or point must be
certain;
birds must be counted at their initial location, prior to any natural
movement or movement in response to the recorders presence;
distances to bird contacts should be known accurately.
Ordinary birds
Count period: Usually 510 minutes in multi-species surveys. For a
single-species survey, 5 minutes may be long enough.
Search effort: Normal (see Section 3).
Flying birds: Ignored/omitted.
Spacing of points: For VCP method distances of 200300m between
census points is usual. You must find the best compromise between
unnecessary walking and the risk of a bird being counted very close to one
station and then of it moving to a position very close to the next station.
Cryptic birds and cue counting
Count period: Should be longer than for the ordinary species. Perhaps
1012 minutes. This is to ensure that high proportions of the birds close to the
recorder are actually detected. A very different approach, which may be
useful when surveying gamebirds and other ground-dwelling species
(especially in grasslands), is to walk transects and to count birds as you flush
them. This is a variant of the standard transect method, where the aim is to
flush as many birds as possible from on, or near the centreline (without
double-counting them).
Search effort: Concentrate on visual and aural cues within 2030m of the
central point. Because the count period is long, be careful not to record birds
which you think may have entered the plot after the count period began. Some
cryptic birds can be disturbed easily so approach the station very carefully
Bird Surveys 69
and record any birds that flush due to your arrival as being present at the
station. At the end of the count period, it may be worth walking around the
plot or pishing to make sure you record stubborn birds.
Other: Stations may be positioned fairly close together (perhaps 100m
minimum). Dawn and dusk may be the best time to census many cryptic birds.
An alternative approach is to use a variant of distance sampling called cue
counting (see also Section 4.6.3). Perhaps the most important cue in the
census of a cryptic bird species is its call (the species could be a partridge,
pheasant or pitta, for example). The cue counting technique has two
components:
1. Estimating cue density: Sampling the cues using a standard VCP
method or a variant of a distance transect method. While in normal distance
sampling, the fieldworker records an individual bird only once per census
plot; in cue counting, he/she records every time the cue is given (i.e. every
time the bird(s) call(s)). For example, during a ten minute point count, there
might be three calls (cues) given by an unknown number of birds within 30m
of the recorder.
2. Estimating cue rate: This involves estimating the number of times that
an individual of the target species gives the cue during a given time period.
To do this, you will have to perform focal studies on several individual birds
for several hours to find out an average call rate for birds (see also Section
4.6.3). Take care to ensure that minimal bias creeps into your estimate of cue
rate. For example, you should estimate a species calling rate during the same
time of day as your census will take place (as birds will call more at some
times of day than others). Also, call rate may depend on the density at which
birds occur, so undertake focal studies of calling birds in several different
areas.
An option is available in DISTANCE to deal with cue count data. First,
TYPE=CUE must be chosen. Then, the cue rate must be entered using the
CUERATE option. If points are used, then you should express the cue rate as
the number of times that the cue is given per count period. Thus, if your count
period is ten minutes per census plot, then the cue rate will be the number of
times the call is given per ten minutes (e.g. if the bird calls six times per hour,
then cue rate equals 1, for count periods lasting ten minutes). Further
information is given in Buckland et al. (1993) and on pages 24 and 34 of
Laake et al. (1994). One complication is that since more than one cue counted
at a census plot may come from the same individual, then the distances
entered are not independent events. Because of this, estimate variances should
be calculated using the BOOTSTRAP command.
Aerial Species
Count period: Must be near-instantaneous as birds are extremely mobile.
Count periods lasting any more than a second or so will seriously
overestimate bird density.
Search effort: The recorder looks directly upwards and estimates the
horizontal distance to bird contacts at that moment. Remember that birds will
be much easier to detect in open areas than in closed canopy forest. An
alternative and possibly useful method is to look up through binoculars or
even a standard sized tube. This will restrict your view of birds to a standard
air volume. This method will not produce actual density estimates but may
still allow comparison of encounter rates between habitats or areas.
Other: Census points can be positioned very close together. Each
instantaneous count is then entered into DISTANCE as a replicate of the one
plot.
Parrots, hornbills, toucans, etc.
Count period: Ten minute counts may be necessary to increase likelihood
of recording birds which can be extremely cryptic at rest and obvious when
flying. Be careful not to record birds which fly into the plot during the count.
Search effort: Concentrate on perched birds within 50m of the recorder.
At the end of the count period check for stubborn birds (see Cryptic birds),
particularly ones which may be sitting silently in the canopy or in nest holes.
If this is to be done, then it is important to standardise this period of flushing
as much as possible, both between census stations and between different
recorders. Parrots may respond to your presence: they may give alarm calls,
stay silent, fly away, or even fly towards you. Remember that you must record
birds in their initial positions.
Other: Many such species are rare, so many stations are needed to
produce good population estimates. I prefer the VCP method to transect
methods for parrots. The main reason is that with point counts, there is a
better chance of recording all birds close to the observer. If transects are
chosen, then they must be walked very slowly and carefully (bird detection at
distance zero metres is paramount). Stations can be placed 200300m apart in
forest, but maybe 500600m in very open habitats. Flying birds must be
ignored in density calculations (except those which fly out of the census plot).
The best time of day for parrot and hornbill census (using VCP method) is
Bird Surveys 71
often between one hour after dawn and 1030h (and perhaps between 1500h
and an hour before dusk).
Large groups and mixed species flocks
Count period: Flocks of birds may be fairly conspicuous and move
around quickly. Short periods (25 minutes) are appropriate. This does not
mean that you cannot spend time after the count period has finished to
identify species and estimate group size (although birds must obviously be
recorded in their initial positions, not where they were when you identified
them).
Search effort: For bird groups, you can either estimate the distance to the
centre of the flock (and then enter the data into DISTANCE as clusters) or
estimate distances to individual birds (Section 3).
Difficult habitats
Count period: In areas of low bird density, the answer is not to increase
the count period but to increase the number of stations or kilometres of
transect walked.
Other: In some situations, transects may be the only way of gathering
enough bird records. They may be appropriate in some montane areas, where
trees are not as tall as in lowland forest (birds may be easier to detect).
However, surveying while walking may be dangerous, and you must make
sure that the habitat through which your walk route passes is fairly
representative of the forest as a whole. In some situations, your view at a
station may be obscured by cliffs. There is an option in DISTANCE to
account for this. For example if at one station you can only see half the plot
(i.e. 180), then you can enter this point as a half plot.
4.6 Specific techniques for special cases

4.6.1 Spot mapping
Spot mapping involves plotting the position of birds on a map. The technique
has been used primarily in temperate regions to count the territories of singing
males. The restraint is normally that it is very time-consuming. You need 6
10 visits to an area and these visits must be sufficiently spread in time as to be
independent. Although little-used in the tropics, the advent of GPS
technology may make spot mapping fairly useful in some situations (although
it will not usually be accurate enough for use in closed-canopy forest). Uses
of spot mapping include territory mapping of singing males and mapping
nests, colonies, roosts etc. They may be suitable when a species is too rare for
distance sampling, or when forest patches are so small that all the birds can be
counted and may work well for species such as thrushes and territorial
flycatchers. They may not be appropriate outside the breeding season, for
birds with asynchronous breeding seasons, for birds which do not hold strict
non-overlapping territories, or for birds which sing in more than one area in a
single season. Little or no data are available on the mating systems of many
tropical birds, so you may be left with unknowns which could jeopardise the
value of your population figures (e.g. there may be many non-breeders,
especially in populations of long-lived species). Likewise, counting singing
males alone may not provide the same index for all species in all
circumstances, when sex ratios are biased.
The method is time consuming. You will need to locate every individual
in the area, or use a calibration technique to predict the likely numbers that
you have missed. How much time or how many repeat visits it takes to find
the birds will depend on its ease of detection (e.g. strength and frequency of
song, conspicuousness of nest). There are several variations on the basic
method which may be appropriate in some cases. Playback of tape recorded
calls could be very useful, particularly in surveys of nocturnal owls. A useful
technique for some species is to flush birds to the edge of their territories
(making territory definition much easier). With playback and flushing
techniques, it is very important to minimise disruption to birds.
4.6.2 Counting birds at or near aggregations
Counting birds at roosts, colonies, leks etc. will usually (but not always)
involve direct counts of all the birds present. You may not be able to count all
birds, either because there are too many, or because you cannot see them all.
In cases of very large numbers, you may have to sample the population. In a
cave holding many birds, could you derive a measure of birds per m2 by some
cunning means? (e.g. counting the birds within the beam of a torch held a
known distance from the cave's wall). Another possibility might be to mark
out known areas on the cave or cliff side and then count birds within these
areas (marks could be used for several years). A different approach is to
select (many) birds at random and estimate the distance from each to their
nearest neighbour. Counting large bird roosts in trees is difficult. Maybe
select a few trees, count the birds and multiply up. Can you photograph all the
birds for counting later? A useful method may be to set up a telescope,
binoculars or even a tube, and count the birds visible in the field of view.
Then calculate the area of your sample field of view and relate to the total
area. In some cases, it may be better to count bird signs than the birds
themselves. In busy colonies, active nests or nest holes may be an appropriate
sampling unit.
Bird Surveys 73
Bird aggregations take many forms, as will the most appropriate methods
to count them. There is great scope for devising new and good methods for
individual cases, so think about it hard! Apart from some considerations in
presenting your results, there are a few specific considerations. First, can you
count the species in its aggregating and non-aggregating phases? Ideally, you
could count a species (e.g. a parrot) during the day using a standard method,
and then in the evening, at or near its roost. Second, are you sure that you
have located all the aggregations in the area, or the proportion of individuals
which join these groups? Third, have you adequately accounted for variability
in counts? You may need to take the average of several fieldworkers counts,
or count the aggregation on several different occasions (times and dates).
The other way to count roosts is to count birds as they move to or from
the roost. In many respects this is similar to counts of migrating birds such as
birds of prey or cranes at bottlenecks. Figure 20 shows a possible method. It
is helpful, but not absolutely necessary to have more than one fieldworker
counting at one time. The distance from the observers to the roost will also
vary: in general the bigger the roost, the further away the counters may have
to be, so as not to be overwhelmed by bird records. You can attempt to cover
the whole circumference of the circle (B), or just a sample of it (A). The latter
may work better as double-counting birds can be a problem with the former
regime, and will seriously overestimate roost size. Using (A), the proportion
of the whole circumference which is covered needs to be known. Remember
that some birds may enter/leave the roost in the dark. It is also unlikely that
departure or arrival directions will be uniformly spread.
A (sample)
B (all)
Proportion sampled = a+b
Figure 20. Counting birds flying to/from roosts.

C (2 people
per sector)
Birds will probably fly

in/out unevenly and in
large groups
360
4.6.3 'Look down' methods from vantage points

Whereas most census techniques are look up methods, there are useful
methods which involve looking down on forest from vantage points or
aircraft. Aerial surveys are used for waterfowl and many other birds (although
not often on expeditions in the tropics). Long watches from hillsides, cliffs
and large trees have been used in many studies (particularly by The Peregrine
Fund, USA), both to enumerate populations and to investigate raptor
behaviour and habitat associations. With look down methods, you usually
need to know the area of your survey. In discrete forest blocks or valleys this
can be gleaned from maps, but in other areas you may have to work out the
area using trigonometry (take compass bearings to points on the edges of your
site and draw your own map). Below are two examples of their use and
suitability:
Surveys of parrots in the Caribbean
Parrot populations on Caribbean islands have been monitored for many years
using long watches. The method could also work for some birds of prey,
pigeons, toucans and hornbills. Observers overlook forested valleys and plot
each birds flight path. In some cases, two or more observers are positioned in
different places around the valley/area, and these note the exact timing of
each flight: a picture of movement can then be developed for bigger areas.
This method has produced some very precise results, but may only work well
with conspicuous birds which fly above the canopy, in areas with suitable
vantage points and where forest occurs as fairly discrete blocks. Most
important, it may only work with rare species where there are so few birds
flying around that you can be sure of following individual birds. The method
also allows area usage and habitat association studies. For example, what
proportion of all birds flew to, or from, primary forest (rather than areas with
scattered trees)?
Cue counting from vantage points
In cases where birds are not extremely rare, it is often difficult to identify
individual birds with confidence. However, watches can still yield population
measures and these may be useful in very rugged terrain such as steep valleys
(where look up techniques are difficult). The method can have one or two
components:
a) Within a known area, the number of flights made is recorded and
expressed as mean number of flights per hour per unit area (a relative
abundance index). The assumption is that the more birds there are, the more
flights you will record. Remember, most birds tend to fly more early in the
morning and in the evening than in the middle of the day. A serious limitation
Bird Surveys 75
is that birds in one area may fly more than those in another. This could easily
be habitat-dependent, with birds in poor habitat perhaps having to fly further.
b) An extension of this method and a possible way round the above
problem, is to find out how much time an average bird spends in flight per
hour. From your vantage point, locate a bird in flight and follow it until it
perches. Then record the amount of time it spends sitting/feeding. When it
takes off again, time its flight. Obviously, you will need to do this many
times.
So now you know (1) what fraction of its time the average bird spends in
flight, and (2), how many flights are made (by an unknown number of birds)
per hour within a known area. To work out your population estimate, divide
(2) by (1). For example, an average bird spends 1 minute out of each hour
flying. You recorded ten minutes of flight by all birds in one hour. Your
population estimate in the area is ten birds. The essential parts of this method
are that you, (a) detect every flight made by all birds in the area (so several
observers may be necessary), and (b) your data on how much a single bird
flies is good enough. This form of censusing is cue counting and in this case,
the cue is bird flight. The cue can also be bird calls (see Section 4.5 on
cryptic birds for an example of cue count distance sampling).
4.6.4 Nest searching/counting limited resources
Many of the worlds threatened bird species have specific habitat
requirements which can make them sensitive to habitat change and make their
distributions very patchy. Some may be counted using standard techniques
but in others the census must focus on their most important or most
characteristic habitat association. Parrots, along with many other birds (such
as large woodpeckers) nest in cavities in very large trees. They can also be
very rare, and one reason for this may be that very few large trees remain in
certain habitats. Therefore, a focus for bird census could be nests, firstly
because nest availability may limit population size and secondly because
nests are characteristic and so are relatively easy to find. Similarly, many
birds nest in dead trees and again nests, rather than birds, can form the
sampling unit (see Figure 21).
Nests themselves can be characteristic, but so can bird behaviour around
that nest. Forest birds of prey can be rare, and they can range widely, making
census (from vantage points) difficult. If, however, the sampling unit is the
nest, rather than the birds themselves, then characteristic behaviours
(returning to nest, displaying and calling) can be used to calculate the number
of pairs in an area.
The limiting resource, or at least the focus, for birds associated with ant
swarms may be swarms themselves. It may make sense to count these
resources: for example spot mapping or counting swarms along transects. A
consideration here is that swarms may be much commoner in certain habitats
(e.g. primary forest) so stratify your sample. Another consideration is time of
day, with swarms being active in the middle of the day, rather than morning
when the majority of birds are most active. Once you know the frequency of
swarms, then you need to focus on the presence/abundance of your bird
species at those swarms, using direct observation. Perhaps the best method for
censusing antbirds is to use standard methods in the morning and then focus
on swarms in the middle of the day. Variants of the above method may be
appropriate in other cases; parrots at clay-licks, honeyguides at bee nests, or
birds which nest in holes in banks.
Figure 21. Counting nest trees (or other limited resources).
Bird Surveys 77
4.6.5 Interviews with local people

It cannot be stressed enough that information from local people can be an
essential part of bird census. In the initial stages, local people (who know the
forest) can lead you to likely study sites and extremely rare birds. Illustrations
of target species will help with communication but remember that local
people will recognise many species by their voice or habits, rather than
specifics of their plumage.
As the survey gets under way, information about relative abundances can
be gained from careful but informal interview. Is the species commoner in
area A than B, in one month more than another, or was it commoner ten years
ago? Local hunters can give useful information about a birds ecology,
particularly its nest sites, breeding season and favourite foods. Interviews are
a particularly good source of information on birds which are economically
important to local communities. Some birds will be pests, while others may
be traded, kept as pets or eaten. In the latter cases it is crucial that you keep
the discussions both amicable and neutral. Dont talk about global declines
and cruelty, or you may get either no information or, worse still, information
that is aimed at keeping you happy!
If interviews are well thought out, then you may be surprised by how
much accurate information on bird capture/trade can be gleaned, even if that
trade is illegal. It might help not to aim your questions at the interviewee (are
birds still caught in other areas?). Bird prices are easy to find out, and how
they change over the years might give useful clues as to population trends.
Dont assume that if you are told that many birds have been captured from a
certain area that there will be few left there: the opposite can be more likely.
Bird markets are a reasonable source of information. Again, keep it amicable
and remember that the further you are from the forest the less reliable your
source of information may be (particularly as to where birds have come
from).
Finally, information given by several different sources is much more
reliable than that given once. The most successful interviews are those which
ask the same simple and unambiguous questions to many different people.
Many more details about planning, implementing and interpreting the results
of interviews is given in Expedition Field Techniques: People Oriented
Research, published by the Expedition Advisory Centre (see References,
Section 8).
4.6.6 Mark-recapture/banding/home ranges
Catching birds (usually with mist-nets) has many uses in long-term ecological
studies but has limited use in short census studies. There may be some
advantages in being able to handle some birds (e.g. nightbirds, very cryptic or
difficult to identify species), to find out if birds are breeding, or recognise
individual birds (banding/marking) but bird-catching is usually too timeconsuming to be considered in the majority of surveys. It also requires
intensive training. In some cases it can be worthwhile, for example, capture
rates of birds per metre or m2 of net can be a useful technique for deriving
indices of abundance for understorey birds and can also be used for diversity
studies. If bird capture methods are used, a scouring of the vast amount of
available literature is an absolute necessity, both to get the method right and
to avoid harming the birds.
4.6.7 Scientific birding
I had quite a heated argument with someone once over the status of a littleknown bird: I was convinced that it was "fairly common in places" whereas
he was adamant that it was "rare to locally absent". Without quantification,
such assessments cannot be compared, and many studies for which the end
result is such a description deserve criticism. But some situations are
incredibly difficult: your species might be nocturnal, cryptic, have no
characteristic habits and be extremely rare (even local people may not be
familiar with it).
It should be apparent fairly early on in the single-species study whether
distance sampling or other techniques will not work (this may not be so
obvious in a multi-species survey). The species may be so rare that the
fieldwork turns from census to search, trying to find at least one record of the
bird (i.e. is the species extant in the area?). To get that record, the random
sampling approach so necessary in census should be replaced by a specific
search in the most likely areas/habitats. This change of approach (i.e. to
birding) should not mean the loss of all systematic recording (Section 5).
The calculation of encounter rates is better than just saying we recorded
the bird four times. Your data becomes an encounter rate (ER) of 0.04 per
hour of search (if you looked for 100 hours). Better still is stratifying the
sample by habitat: ER in primary forest = 0.06; ER in logged forest = 0.02.
This is the start of a repeatable method: describe your method even if it
doesnt seem much of a method! For example, did you start looking just after
dawn, or were you restricted to later in the day? Did you have to stick to
paths? Did you know its call or were you relying on actually seeing one?
Which habitats did you sample? Remember that the habitat in which you
didnt record the bird can be as important as the one in which you did record
it. Compare the above ERs with ER in primary forest = 0.08, ER in logged
forest = 0.00. Your lack of records in logged forest may have serious
Bird Surveys 79
implications for assessing the status of your species but only if you describe
your data properly.
4.7 Interpreting and presenting results of specific

studies
Whether you tailor existing methods to your situation, or you devise your own
method, you must describe your technique completely. Things that may seem
obvious to you (because you did them) may not be at all clear when someone
repeats your work in 50 years time! It may be important to show the locations
of vantage points, roosts, colonies or other features on maps (with
coordinates). Report how many fieldworkers collected the data, what time of
day it was collected, and anything else which will make the repeat census
more comparable with your own.
Likewise, your study may itself be a repeat survey. In this case, the
method used in the past may be the best one to use. If so, then there is no
problem and you must follow their methods as precisely as possible. In other
cases, however, you will be able to make improvements or devise totally new
methods. Do this, but remember that bird population monitoring depends on
comparability. Is there scope in your study to use both the previously used
methods and your new ones? This has an obvious advantage: you get a
population figure that, although possibly wrong, is comparable with the old
one, along with a new (and hopefully more accurate) one which can form the
baseline for the future.
As introduced in Section 2, there are two components to a population
estimate; its accuracy and its precision. In distance methods, the reliability of
results are, to a certain extent, computed statistically. If you devise or tailor
your own methods, then errors can be harder to quantify and yet are even
more important. For example, in a roost count, did different observers
estimates match? One person counted 100, while others reckoned 90, 80, 90,
110 and 130. So which estimate do you use? In this case, the mean might be
best, but maximum and minimum values are also important (they may be
correct anyway), as is standard deviation. These are important statistics which
convey to people (who werent there) how precise your figures are likely to
be. The way you convey the precision of your results will depend on your
method. In interviews with local people, how many people said one thing and
how many people contradicted this? During long watches, there may be doubt
as to which bird was which in this case what was the minimum and
maximum numbers that you are sure of?
Section 5
ASSESSMENT OF SITES: MEASUREMENT OF

SPECIES RICHNESS AND DIVERSITY
Peter A. Robertson & Durwyn Liley
5.1 Introduction
Many projects aim to assess the conservation importance of a site or the
relative values of different habitats and do this by determining the diversity of
species present. Such surveys provide baseline conservation data on the
distribution of key species, the richness of sites or habitats and allow
comparisons to be made between areas. For such data to be meaningful it is
necessary to know how accurate and how complete they are.
In tropical forests it is notoriously difficult to locate birds. The very
structure of the habitat, with high canopies and sometimes with dense
undergrowth, means that birds are difficult to see. Many species occur at very
low densities and the difficulty of identifying species compounds the
problem. With a high diversity of possible species, accurately describing the
avifauna of a tropical forest site presents considerable problems, particularly
if time is limited and the observers are unfamiliar with the species present.
For these reasons it may be excessively time consuming and unrealistic to
collect density information as described in the previous section. Measurement
and comparison of species richness may be a sufficiently challenging and
worthwhile aim.
This section is concerned with methods by which a team can collect data
which are as complete and meaningful as possible, given the constraints of the
habitat, observer experience and time. Methods are discussed by which it is
possible to record the bird species present at a site, to determine how
complete the list compiled is, to judge how much time is needed to sample a
site to some degree of accuracy and lastly, to compare between sites.
5.2 Compiling a species list

The most basic description of the avifauna of a site is a species list. A list
describes the diversity of a site, and shows the presence or absence of rare
species. Species that are globally threatened (Collar et al. 1994) are key
species for conservation and as such are key species to locate on any bird
survey. The number of rare species and the diversity of species present at a
site can be used as indicators of the importance of different sites or habitats
Bird Surveys 81
for bird conservation. It is important that a species list should be as

exhaustive as possible or that its incompleteness should be acknowledged and
understood.
Compiling a species list is principally a matter of spending time
birdwatching at a site. However, while the number of species recorded is
largely dependent on the time spent in the field, there are a number of
techniques that can be used to maximise the variety of species recorded in a
short time. They amount to the skilled use of a variety of observation
techniques supplemented with the use of supporting equipment. These
techniques are described below:
Habitats
The full range of habitats and altitudes at a site should be covered. Subtle
habitat variations can be important. Habitat breaks and changes in habitat
such as ridges and valley bottoms are good areas to focus on, as are streams
and marshes, particularly in dry regions or during the dry season. Forest
edges are well known to attract birds and can provide easier viewing. Many
species depend on restricted habitats such as bamboo clumps and it is
therefore important to locate and search such restricted habitats, particularly
as the species restricted to these habitats are often of conservation concern.
Canopy watching
Many species in tropical forests are more or less restricted to the forest
canopy and can be difficult to see because of the height of the canopy above
the ground. By making use of high ground, slopes, knolls, hill sides or by
climbing a tree it is possible to be level with the canopy and increase the
chance of seeing canopy species. If such an opportunity does not present itself
then an observer could choose a spot with an unrestricted view of the canopy
and lie on his/her back to allow a prolonged period to be spent concentrating
on the canopy whilst avoiding neck ache!
Sky watching
Vantage points providing a view over the canopy increase the likelihood of
seeing raptors, swifts, parrots and other frugivores. Some species (e.g. some
parrots and some pigeons) roost communally and can be counted flying to and
from the roost at dawn and dusk.
Speed of walking
A fast, quiet pace is better for detecting ground birds on a path and for
encountering flocks; slow walking is better for detecting species in the canopy
and away from a path. Skulking understorey species can be detected by
scanning ahead along paths and stretches of streams, particularly when first
rounding a bend in a path or stream. Frequent stops, listening for the
movements of understorey species, such as the rustling of leaves, can also
help to detect this difficult group of species. If an understorey species is
flushed without being identified, waiting silently in the spot or leaving and
cautiously returning shortly afterwards may allow the same individual to be
seen and identified.
Flushing/rope-dragging
Rails and other shy waterbirds in marshes and larks in grasslands can easily
escape detection by a single observer. Sometimes the only way to detect these
species is for a group of observers to walk in a line across an area of suitable
habitat in an attempt to flush any individuals present. A fast pace is required
to avoid birds fleeing without breaking cover. Similarly, a rope can be
dragged across the top of the vegetation in marshes and grasslands to flush
birds. The rope should be thick enough to disturb the vegetation without
being too heavy to drag. Nylon ropes are preferable in marshes as they do not
absorb water. Densities can be calculated from rope-dragging by calculating
the area of habitat disturbed and assuming that all birds present were flushed.
Sitting still
Certain points such as fruiting trees, streams, pools, breaks in undergrowth or
bamboo provide good vantage points for waiting for birds to appear. Patience
can be rewarding.
Timing
Activity patterns vary between species. In West African forests the frequency
of calling of many species decreases after 0930h. Mist-netting studies in
central America have shown that certain groups of species were active at
different times of day, for example more than half of all species trapped, and
most Tyrannidae, were trapped in the early afternoon. Between 0900 and
1200h is the time of peak activity for most soaring raptors. It is therefore
necessary to carry out searches at different times of the day and not
concentrate exclusively on the early morning period. Blake (1992) describes
the effect of timing on the results of point counts in a lowland wet forest in
Costa Rica.
Nocturnal species
As an extreme example of different activity peaks between species, certain
birds, particularly owls and nightjars, are active only at dawn, dusk and at
night. These species are often under-recorded and it is necessary to spend
time in the study area at night in order to stand a chance of recording them.
Bird Surveys 83
Many nocturnal species are very vocal and can usually be identified by call
alone, although a powerful torch is also useful. A tape recording of the calls
of possible species can be used to elicit a response.
Knowledge of calls
Knowledge of the calls of target species and of shy or skulking species will
greatly increase the chances of recording these species at a site. Tapes of bird
calls from many parts of the world are available (see appendix at the end of
this chapter) and can be used to learn calls before starting fieldwork, thus
saving time and energy during survey work. Unfamiliar calls heard while in
the field can be tape recorded (see below) or transcribed into a notebook and
identified later by reference to pre-recorded tapes.
Use of tape recorders
Small portable tape recorders and speakers are available relatively cheaply
and can be of great help in the field. Playing the call or song of a species will
often produce a response if there is an individual of that species within
earshot of the tape recorder, with the bird either coming out into the open or
calling in reply. The chance of encountering shy, skulking or quiet species
and nocturnal species can be greatly increased by tape playback. Walking
through suitable areas occasionally playing calls of potential species is a
possible method. In addition, the use of a microphone enables an unknown
call to be recorded and played back immediately to bring the bird in question
out into the open. When using these techniques the welfare of the bird should
always be carefully considered as the excessive use of tape playback can
cause disturbance to breeding birds. Marian et al. (1981) give a detailed
account of playback techniques and their possible side effects.
Attracting species
Some species can be attracted to a particular spot allowing observers to
record their presence. Bait can be used to lure species, for example fruit for
attracting both frugivores and insectivores feeding on insects attracted to the
fruit, honeycomb to attract some species of honeyguide and sugar solution to
attract hummingbirds and other nectar feeders. In an otherwise dry area,
drinking pools can be created which attract some species to drink, particularly
at the hottest part of the day. Certain noises will also attract birds to the
observer; pishing is a well known technique among birders, a squeaking
sound made with pursed lips and often the back of the hand which can draw
passerines in close. It is also possible, once learnt, to draw in flocks by
imitating owl species, for example, the call of a Ferruginous Pygmy-Owl in
South America or the Barred Owlet in East Africa. Alternatively, recorded
calls of these species could be played on a loop tape.
Special events
Certain events within a tropical forest tend to concentrate birds from a wide
area. Such events include trees fruiting or army ant swarms. Although these
events are unpredictable in terms of when and where they will occur, it is
worth putting some energy into locating them as they are often focal points
for the activity of many species within a forest. Mixed-species groups of birds
will often forage together as a flock. If such a group is encountered then it is
worth following until all the species present have been identified.
Mist-netting
Mist-nets can be used to add species to a site list, but training is necessary to
learn how to set the nets and more particularly how to remove birds from
them. The use of mist-nets by people with no training represents a serious
threat to the birds caught and should not be attempted. In some countries
there is a legal requirement for those who use mist-nets to possess a licence.
However, for the appropriately qualified, mist-nets are an effective means of
detecting skulking understorey species. They can also be used to catch
species in the canopy, although this is more difficult and capture rates tend to
be much lower than in the understorey making the effort necessary much less
worthwhile. Mist-nets are most frequently used to survey understorey birds,
and the number of birds caught will, to some extent, depend on where and
how the nets are set. Nets set by water, fruiting trees and low vegetation are
likely to be effective in catching birds. Some observation of areas of activity
can be useful in determining the best location to set nets. The use of mistnetting as a quantitative survey technique is dealt with later (Section 5.3.5).
Knowledge of the ecology of species and targeted searches
After the initial survey effort when the common species will have been found
it can be very helpful to review the list of possible species not yet recorded,
especially targets such as threatened species, and focus continuing searches
on these using a knowledge of the species ecology to select the methods from
the list above which are most likely to locate these species.
Local knowledge
Local people, particularly hunters and harvesters, often have a very good
knowledge of many of the species present at a site. Even unstructured,
informal interviews can be used to provide observers with an idea of species
that could be expected before visiting the area. The absence of key species,
particularly quarry or otherwise significant species, can also be concluded
from interviews with the local population. Simply showing someone a field
guide to an area and noting which species are recognised and which are
unfamiliar is a good technique but should be used with a sensible degree of
Bird Surveys 85
precaution, questioning to check reliability where possible. The number of

species apparently recognised that are highly unlikely to occur in the region
gives a good indication of accuracy. If such interviews are carried out in a
number of communities around a site, the consistency with which a certain
species is reported by different communities, groups or individuals can be
used as a means of measuring the reliability of the record, although some
errors may persist between interviews.
5.3 Standardising recording methods

Species lists can show considerable variation in how accurately they describe
the avifauna of a site. It would obviously be unwise to compare two species
lists from different sites if one was collected over a two week period and the
other was the result of many years of data collection with repeated visits to a
site by a variety of observers. Lists can vary according to factors such as the
length of time over which the data was collected (with more of the normally
resident species being recorded over a longer time period but also more
species of only chance occurrence), the quality of the observers and the
variety of habitats sampled. The value of a species list is greatly increased if
the level of effort is measured and if the techniques used in compiling the list
are standardised. Annotating a list with the relative abundance of each species
also makes it much more useful. There are a variety of ways in which effort
can be measured and relative abundances can be calculated.
5.3.1 Species discovery curves
The importance of the amount of time spent at a site has already been
stressed. The frequency of adding new species to a list declines with time; at
the start of fieldwork every species recorded will be new and as time spent in
the field increases so fewer and fewer new species will be recorded. Yet even
after spending months at some sites it is still possible to add new species to
the list. When collecting data for a species list the time period over which the
list was collected, the number of observers and the number of hours spent in
the field should be recorded. By also noting the time and date at which each
new species is recorded some simple analysis becomes possible.
The rate of species discovery can be recorded in the field by dividing the
overall survey effort at a site into standard units, and recording all species
noted during each unit. Survey effort is a function of the time spent surveying
and the number of observers. Each observer is only collecting independent
data if he is working at a different place from other observers. Thus, if
observers are working in pairs (e.g. for safety reasons) each pair is effectively
acting as a single observer. Units of effort should therefore be observer x unit
of time (e.g. observer hours or observer days). Simultaneous periods of

observation by different observers can be grouped together or treated as
consecutive periods of observation (e.g. four observers working in different
areas during the same one hour period can be treated as one unit of four
observer hours or as four consecutive units of one observer hour). The unit of
time used may vary from an hour to a day (or even longer). The advantage of
using a day as the time unit is that the activity pattern of species is
approximately the same for each recording unit, although this is only practical
if at least ten days, and ideally a rather longer period, is to be spent at the site.
If less than a day is used as the time unit, then changes in species activity
patterns will affect the species discovery curves, for example with fewer new
species likely to be discovered during a time period covering the middle of
the day. However, a shorter time unit will give more detail in the curve,
particularly in a short overall recording period and the problems of changes in
daily activity patterns can be reduced by not recording over the middle of the
day. The recording unit of time should be chosen such that a minimum of at
least ten recording units make up the total period of observation.
If the cumulative total of species recorded is plotted against survey effort
then a curve rising to a plateau will result (Figures 21 and 22), as fewer and
fewer new species are discovered with continuing effort. Such a curve can be
a useful indicator of the optimum length of time to spend at a site illustrating
when the majority of species have been recorded. The position of the plateau
of the curve can be used to compare species richness between sites. Figure 21
shows a graph from a site in Indonesia, the data for which was collected over
2
a four month period from a forest block of 3km . It is only after
approximately 50 days of fieldwork that the majority of species have been
recorded.
Figure 21. Species discovery curve. The graph shows the cumulative total of the
number of species seen during fieldwork between August and November in
West Java, Indonesia. The site included a range of primary forest, secondary
forest, forest edge and scrub habitats (from the authors own data).
Figure 22 160
gives results from surveys carried out in three different forest
types on the Freetown Peninsula, Sierra Leone, West Africa (Ausden and
Wood, 1991).120
It shows a slow but continuing rise in the closed canopy forest
compared with a relatively rapid rise towards a plateau in the cumulative total
Number of
80secondary regrowth. It also shows a shoulder in the curve for
of species in the
Species
degraded forest after 70 man hours, which results from a change of base camp
to a new part40of the forest. This illustrates the importance of thorough
coverage of the site if results are to be applied to the whole site.
0
20
40
Observer Days
60
80
Bird Surveys 87
Figure 22. Species/time curves for the three habitats surveyed (from Ausden
and Wood, 1991)
Total species recorded
5.3.2 Encounter rates

One means of incorporating the effort expended into the analysis of bird
140
survey results
is to record field hours for each observer and the number of
individuals120of each species observed. This allows an encounter rate to be
calculated for each species by dividing the number of birds recorded by the
number of100hours spent searching, giving a figure of birds per hour for each
species. Additional
information can be gained by determining separate
80
encounter rates for each broad habitat type (e.g. primary forest and logged
60
forest). At each site all the main habitat areas accessible should be visited.
40
Farmbush
The data provided by encounter rates Closed

do Canopy
not provide
an accurate
Forest
Degraded Forest
indication 20of abundance and are not a substitute
for density estimates.
However, if0 it is assumed that a species is as easy to locate at one site as
another then the
encounter
rates
comparable
species between
0
50
100 are crudely
150
200
250 for a300
of hours
in theinto
field crude ordinal categories of
sites. Encounter rate data Number
can be
split
abundance (e.g. abundant, common, frequent, uncommon and rare), making
these terms much more useful as they have some definition and allowing a
species list to be annotated in such a way that future surveys might detect
large scale changes in the abundance of individual species. An example of
abundance categories related to encounter rates, used on an expedition to
Paraguay, is given in Figure 23.
Figure 23. Using encounter rates to give a crude ordinal scale of abundance
(from Lowen et al. 1996).
Abundance category (Number of
individuals per 100 field hours)
<0.1
0.12.0
2.110.0
10.140.0
40.0+
Abundance
score
1
2
3
4
5
Ordinal scale
Rare
Uncommon
Frequent
Common
Abundant
One important bias in the use of encounter rates in the field is that, with
all the observers starting from the same base at the start of fieldwork, any
species roosting near the site will be recorded by all observers at the
beginning and end of each day. This bias can be reduced if observers move
rapidly to a starting point some distance from the base camp, and if these
starting points are varied between observers and between days. Vocal and
prominent species are also over recorded at the expense of more skulking or
quiet species, and this factor must be taken into account when describing the
results. Although this factor will most commonly prevent comparison of
different species, it might also prevent comparison of the same species at
different times if there is some seasonal change in the detectability of the
species, for example a difference in frequency of vocalisations between the
breeding and non-breeding seasons. These factors are commonly but
mistakenly ignored so that some species described as rare are actually just
difficult to detect and some species described as common may just have
obvious and far-carrying calls.
5.3.3 Mackinnon lists
Mackinnon Lists (Mackinnon and Phillips, 1993) provide another means of
calculating a species discovery curve and an index of relative abundance.
Mackinnon Lists differ from the other techniques in that the unit of effort is
the time taken for an observer to record a pre-determined number of species.
The advantage of this is that it makes the method relatively less susceptible to
differences in ability and concentration of the observer. If an inexperienced
observer takes a long time to identify each species detected this does not
greatly affect the results providing he/she does eventually identify all species
detected. Similarly recording during a period of low activity such as over
midday will not greatly affect the results it will just take longer to detect a
given number of species.
The observer makes a list by recording each new species until a predetermined number of species is reached. A species can only be recorded
once in each list but may be recorded in subsequent lists. The appropriate
length of list can vary between 8 and 20 species; the larger the likely total
number of species at the site the longer the length of list chosen. Comparisons
can only be made between surveys where the same length of list was chosen.
Surveys are repeated until a minimum of ten and preferably more than fifteen
lists have been produced for each site. When recording data the observer is
free to search for birds in as efficient a manner as possible, using whatever
search techniques from section 5.2 are appropriate for the site. However, the
observer should endeavour to cover different ground at least from one list to
the next to avoid recording the same individuals on repeated lists. A species
discovery curve, as described above, can then be drawn by replacing the unit
of survey effort with the number of lists and plotting this against the
cumulative total number of species. As above the position of the plateau of
the curve reflects species richness and the shape indicates how many more
species are still likely to be found in that locality (see Section 5.4.1 for
analysis of Species Discovery Curves). Figure 24 shows an example of such
Bird Surveys 89
curves from four localities in Indonesia, the shape of curve and steepness
differs between sites.
Figure 24. Species curves from four different sites in Indonesia. (From
Mackinnon & Phillips, 1993. Reproduced with permission of Oxford
University Press)
5.3.4 Timed species-counts (TSCs)
Timed Species-Counts
(TSCs) were developed by Pomeroy and Tengecho
160
(1986) for open woodland and bush habitats. They provide a simple method
Danum Valley
of comparing the
120avifaunas of extensive areas by sampling representative
habitats. Simple, rapid and effective, they give a reasonable measure of
Number
of
relative
abundance.
open Kulon
habitats
80 They are best suited to extensive areas ofUnjung
and Species
there are limitations to using the method in thick forest. Bennun and
Waiyaki (1993) give
40 an example of the use of TSCs in Kenya.
Baluran
Data for TSCs are recorded in six columns, corresponding to six 100
minute intervals during
an hour long survey. The observer walks
a slow
Bogor atBotanic
-1
pace (approximately 12
the 13
study15area for one
hour.
For
the
1 km/h
3 5 ) through
7 9 11
Gardens
first ten minutes, every species recorded is noted down in the first column,
Number
Lists of individuals. For the second
giving only the species name,
not theofnumber
10-minute period, any species not already recorded is noted in the second
column. The remainder of the hour is also divided into 10-minute periods and
any species recorded for the first time during any 10-minute period is noted
down in the relevant column, such that every species recorded during the hour
is written down only once, in the column relating to the 10-minute period
during which it was first seen. The analysis (described in Section 5.4.4) gives
an index of relative abundance based on the assumptions that the more
common species will be recorded earlier during each survey and in more
different surveys than rarer species.
A minimum of 15 surveys should be carried out at different parts of the
site. The choice for the length of time for each count is a trade off between
recording as many species as possible and not spending so long that few visits
can be fitted in. An hour of observation is recommended as standard so that
surveys by one team can be compared with similar surveys at different sites or
by different teams at the same site. Pomeroy and Tengecho (1986) also
2
recommend using an area of 1km for each count. In making the count the
observer intentionally visits as many parts of the area as possible and
concentrates on areas where bird activity is greatest.
Simple habitat and environmental variables can be included in the analysis
which will help to account for differences in bird communities. Pomeroy and
Tengecho recorded two variables: moisture index and percentage cover of

woody plants. Other variables could also be used (see section 6).
5.3.5 Mist-netting
Mist-netting is an effective means of recording quiet and skulking species of
the forest understorey which may not be recorded using other techniques.
Standardised mist-netting effort can be used to compare this element of the
avifauna between sites. However, mist-netting is a very labour intensive
technique and the purpose of using mist-nets should be carefully considered
before any netting is undertaken. They should only be used by people with
appropriate training to avoid any threat to the birds caught.
For comparison of sites a minimum of 100m of nets set in a standard form
(e.g. on a random grid or in a straight line) should be used with at least two
days of catching effort. Nets should be opened for 34 hours from dawn and
perhaps also for two hours before dusk each day, although capture rates in the
evening are often lower than those in the morning. The results from
standardised mist-netting studies are expressed in units of birds caught per
-1 -1
metre per hour (i.e. birds m h ). Marking individuals caught (with metal rings
or colour rings) allows re-captured birds to be excluded from capture totals.
This technique should only be used by those with appropriate training. Nets
should be checked at least every half hour or more frequently if it is very hot,
and should be closed immediately during rain. The effectiveness of mistnetting is greatly affected by the skill with which the nets are set and also by
the type and condition of the nets. More birds tend to be caught if the nets are
set along lines specially cut through moderately dense vegetation than if
existing tracks or open areas with little understorey vegetation are used. Nets
set parallel or perpendicular to streams or close to other water sources tend to
be particularly effective but such selective positioning of nets reduces the
comparability of results between different sites.
5.4 Analysis of data

There are a number of ways in which the data collected from the techniques
described above can be analysed. The purpose of such analysis is to enable
comparisons to be made between sites and allow predictions to be made of
total species richness of sites, to describe the relative abundance of different
species within a site and the relative abundance of the same species between
two or more sites. Rigorous recording of the methodology used and
particularly the effort expended is essential to allow such comparisons.
Bird Surveys 91
5.4.1 Predicting total number of species from species discovery

curves
Two types of species discovery curves have been described, using observer
time units as a measure of survey effort (Sections 5.3.2 and 5.3.4) and using
repeated species lists as a measure of survey effort (Section 5.3.3). In each
case, a curve can be fitted to the points plotted. If the curve is extrapolated
beyond the data obtained, a prediction of the level at which it reaches a
plateau can be made. The level of this plateau is equivalent to the total
number of species expected at the site. Plotting such a curve also enables an
estimate of the level of effort required to add a given number or percentage of
species, allowing approximate calculations to be made of the amount of time
to spend at a site in order to optimise the number of sites to be visited in a
given time. This analysis is best carried out on a computer with a suitable
statistical package, although the choice of model used to describe the curve is
not simple. Two alternative general curve shapes are a logarithmic curve and
an exponential curve. An exponential curve may be adequate when a well
known avifauna is sampled in a small, homogeneous area, but where a poorly
known avifauna is sampled in a large and heterogeneous area the logarithmic
curve may be preferable. Trying out different curves with a computer package
allows the curve with the lowest residual variance to be selected.
Alternatively the curve could be plotted by eye.
An alternative method of predicting the species richness of a site is to plot
the number of new species recorded for each unit of survey effort against the
log of the cumulative number of species recorded prior to that unit of effort: a
linear relationship is expected. The number of additional species will
decrease inversely with the cumulative total and a regression line can be
plotted using a computer statistics package. Where the line of best fit crosses
the x axis an estimate can be obtained of the total number of species present.
An example from Kenya is shown in Figure 25. Because a straight line
relationship is expected it is easier to fit the line by eye than it is to fit a curve
as described above.
15
Number
of
additiona
l species
0
Cumulative number of species (log scale)
100
Figure 25. An example of analysis used to predict the total number of species at
a site using cumulative count data (from Pomeroy and Tengecho, 1986). The
point at which the line crosses the x axis gives an estimate for the total number
of species present at the site and is 120. R2 = 0.880, P<0.001, for the regression
line sa = 34.87.3 (log10 sp) where sa is the number of species at successive
counts of 1.0 hours and sp is the cumulative number of species.
5.4.2 Analysis of encounter rate data
The analysis of encounter rate data is very straightforward. Figure 26 shows
the results from a survey carried out by three independent observers in a
forest in Madagascar. Observer 1 spent two hours in the forest while
observers 2 and 3 spent three hours each. The total observation period is thus
eight hours. The encounter rate for each species is equal to the total number
of individuals recorded by all three observers divided by the period of
observation and multiplied by ten to give a result in units of number of
individuals recorded per ten hours of survey.
Bird Surveys 93
Figure 26. Encounter rates from a forest in Madagascar (authors own data).
There were eight survey hours for each observation.
Species
Number of individuals
by each observer (3
observers)
1
2
3
Number of
individuals
/10 hours
Class or relative
abundance
1.25
Uncommon
2.50
Frequent
6.25
Frequent
11.25
Common
Madagascar Turtle Dove
3.75
Frequent
Madagascar Black Coucal
1.25
Uncommon
Francess Sparrowhawk
Madagascar Buzzard
Grey Cuckoo-shrike
Madagascar Blue Pigeon
Blue Coua
Giant Coua
4
1
Red-fronted Coua
Mascarene Martin
Cuckoo Roller
2
11
Red-tailed Vanga
2
4
Chaberts Vanga
1
1
Madagascar Fody
Forest Fody
Frequent
1.25
Uncommon
6.25
Frequent
2.50
Frequent
35.00
Common
White-headed Vanga
Velvet Asity
6.25
3.75
Frequent
2.50
Frequent
16.25
Common
2.50
Frequent
1.25
Uncommon
5.00
Frequent
1.25
Uncommon
Hook-billed Vanga
Uncommon
Blue Vanga
Madagascar Drongo
Frequent
1.25
Madagascar Sunbird
Souimanga Sunbird
10.00
1
1
2.50
Frequent
10.00
Frequent
13.75
Common
Lesser Vasa Parrot
12.50
Common
Madagascar Magpie-Robin
5.00
Frequent
Madagascar Bulbul
29
12
17
72.50
Abundant
Common Jeery
17.50
Common
Madagascar Scrub Warbler
22.50
Common
Common Newtonia
16.25
Common
Madagascar White-eye
27.50
Common
Grey-headed Lovebird
5.4.3. Analysis of Mackinnon list data

The data gathered by Mackinnon lists can be analysed in two ways. Species
discovery curves can be produced by plotting the cumulative total of species
against the number of lists and analysing the results as described in section
5.4.1. In addition, the results can be analysed to give an index of relative
abundance for each species. The relative abundance of each species at each
site is equivalent to the fraction of lists on which a species occurs, i.e. if a
species is recorded on 8 out of 10 lists at site A and on 3 out of 15 lists at site
B then its index of relative abundance is 0.8 at site A and 0.2 at site B. This
index can vary between 0 (species not recorded) and 1 (species recorded on
every list).
5.4.4 Analysis of TSC data
In analysing the results (Figure 28, in which survey 1 is that shown in Figure
27), each species is given a score depending on the 10-minute period in which
it was first recorded, such that species recorded in the first ten minutes are
given a score of six, species first recorded in the second ten minutes given a
score of five and so on, with species recorded in the final ten minutes being
given a score of one. If a species is not recorded from a survey then it has a
score of zero for that survey. An index of the relative abundance of species
recorded from repeated surveys is calculated as the mean of scores from each
survey, and therefore varies between a maximum value of six and a minimum
value of 1/n (where n is the number of repeated surveys).
Figure 27. Results from TSC in savannah forest, Sierra Leone, West Africa
(from authors own data).
010 mins.
1020 mins.
Common
Bulbul;
Great Blue
Turaco;
African
Golden
Oriole;
Grey-headed
Sparrow;
Black Kite;
Tawnyflanked Prinia
Turatis
Boubou;
Red-chested
Cuckoo;
Bush
Petronia;
Blue-spotted
Wood Dove
2030 mins.
Dybowskis
Twinspot;
Red-eyed
Dove;
Swallowtailed Beeeater
3040 mins.
Black
Cuckoo;
Yellowwinged
Pytilia
4050 mins.
Vieillots
Barbet;
Red-winged
Warbler;
5060 mins.
Grasshopper
Buzzard;
Yellowthroated
Leaflove
Bird Surveys 95
Figure 28. Analysis of TSC results from four surveys.

Species
1
Common Bulbul
Great Blue Turaco
African Golden Oriole
Grey-headed Sparrow
Black Kite
Tawny-flanked Prinia
Turati's Boubou
Red-chested Cuckoo
Bush Petronia
Blue-spotted Wood Dove
Dybowskis Twinspot
Red-eyed Dove
Swallow-tailed Bee-eater
Black Cuckoo
Yellow-winged Pytilia
Vieillots Barbet
Red-winged Warbler
Grasshopper Buzzard
Yellow-throated Leaflove
Survey scores
2
3
4
6
6
6
6
6
6
5
5
5
5
4
4
4
3
3
2
2
1
1
6
4
5
4
6
6
5
3
3
2
4
6
4
5
5
2
5
4
2
2
3
4
4
3
5
3
1
3
1
6
2
3
6
3
5
6
4
2
3
2
3
3
2
3
2
Total
score
24
16
19
21
17
22
20
14
12
11
8
15
7
9
9
12
6
2
1
Mean
score
6
4
4.75
5.25
4.25
5.5
5
3.5
3
2.75
2
3.75
1.75
2.25
2.25
3
1.5
0.5
0.25
Species
rank
1
7
5
3
6
2
4
9
10=
12
15
8
16
13=
13=
10=
17
18
19
5.4.5 Analysis of mist-net data

A survey comparing the communities of different forest types around Gola
Forest, Sierra Leone, West Africa (Allport et al. 1988) used mist-netting as
one of a variety of methods. The results were analysed using diversity,
equability and similarity indices. These are a convenient, although not
foolproof, means of combining the species richness (total number of species)
and the evenness (extent to which all species are equally common) of a bird
community and as such may be used whenever there are data available on
number of species and their relative abundance. The most widely used indices
of diversity and equability are the ShannonWiener indices;
Diversity Index H' = - pI . ln (pi)
where pi is the proportion of species i expressed as a proportion of the total
number of individuals of all species, ln is the natural logarithm, and
represents the total pI . ln (pi) for all species.
Equability Index J = H'/Hmax = pi . ln (pi)/ln (s)
where s = number of species.
The results from Gola Forest are shown in Figure 29.
Forest Type
Secondary
Regrowth
Plantation
Logged
Primary
Total Catch
Diversity
Equability
43 bird sub-sample
Diversity
Equability
2.72
0.88
2.58
2.31
2.75
2.59
0.81
0.92
0.91
2.13
2.58
2.59
0.88
Number
of birds
56
Number of
species
22
0.81
0.91
0.91
50
65
43
17
20
17
Figure 29. Analysis of mist-net catches in different habitats around Gola

Forest, Sierra Leone, W. Africa (authors own data).
The results were analysed for the total catch in each forest type and for a
sub-sample of 43 birds, the minimum number caught in any one forest type.
The higher the diversity index the greater the number of species and evenness
of their populations. This can result in communities with higher species
richness and lower evenness having the same diversity index as communities
with a lower species richness and a higher evenness. Equability Indices vary
from 0 to 1, with communities where all species are equally abundant having
an index of 1. In the Gola example, the plantation forest understorey
community has a lower diversity and equability index than the other forest
types, which reflects the fact that the understorey vegetation in plantations is
very uniform and 46% of the birds caught in it were of just two species. The
relatively high diversity of the secondary regrowth may have been due to the
fact that the mist-nets covered almost the entire height range of this habitat, so
that species which in other forest types would not have been caught in the
nets were caught in this forest type.
Similarity Indices measure the degree to which the species and their
relative abundances are shared between different bird communities.
Completely similar communities have an index of 1 while completely
dissimilar communities have an index of 0. The Czekanowski similarity index
is widely used;
Similarity Index Sc = [2 min (Xi, Yi)]/[ Xi + Yi]
where Xi and Yi are the abundance of species i in habitat X and Habitat Y and
min (Xi, Yi) is the sum of the lowest abundance where species i occurs in
both habitat X and habitat Y.
A similarity matrix of the four forest types sampled using mist-netting in Gola
Forest is given in Figure 30.
Bird Surveys 97
Secondary Regrowth
Plantation
Logged
Plantation
0.991
Logged
0.516
0.323
Primary
0.475
0.315
0.989
Figure 30. Similarity indices of bird communities from different habitats

around Gola Forest, Sierra Leone, W. Africa (from mist-net data) (authors
own data).
This analysis shows the similarity between the secondary regrowth and the
plantation and between the logged and primary understorey forest
communities, and the lack of similarity between these two groups.
5.5 Discussion
The methods described in section 5.2 are aimed at increasing the efficiency of
compilation of a species list, i.e. ensuring that all species present are found in
the shortest possible time. The methods and analyses described in sections 5.3
and 5.4 are aimed at standardising the application of the methods described in
5.2 to allow comparisons between sites and between surveys. The usefulness
of a species list is significantly increased if the methods by which it is
compiled are carefully designed. It is necessary to include or describe
observer experience, effort, habitat, time of day and season. Species
Discovery Curves are a means of estimating how much of the total bird
community has been recorded during a survey and the total number of species
that is likely to be recorded. Encounter rates, Mackinnon lists and Timed
Species-Counts are simple methods of making an estimate of the relative
abundance of species and may be used to compare different species within a
site or the same species between different sites. However, such comparisons
must be made with care, taking into account the possible effects of various
biases on the results. Perhaps the most significant bias in the use of methods
measuring the relative abundance of species is the difference in detectability
between species. Thus, rare but large and vocal species may be recorded as
more common than common but small and secretive species. The effect of
this bias can be reduced for all the methods by setting a distance limit of say
20m on all records, such that any individual recorded at a distance of greater
than 20m is discounted. This will, however, greatly reduce the amount of data
gathered. Alternatively data could be gathered from two width bands of less
than and greater than 20m and the results analysed separately or combined. A
more effective way to rule out this bias is to use methods based on distance
estimation (Section 3). However, the rigorous methodology of these methods
requires preparation before the survey can start, e.g. cutting transects or
marking random points; limits the observers freedom to apply the search
methods described in section 5.2; and require the gathering of a relatively
larger amount of data before a valid analysis, which is in itself complex, can
be carried out.
The quality of data gathered by Mackinnon Lists is relatively unaffected
by the skill of different observers and by the change in expertise of a team as
it gets to know a bird community. If an inexperienced observer takes a long
time to identify each species seen he/she may still compile the same list as an
experienced observer although it would take him/her longer to do so. This
would not effect the results as this method is not based on units of time as is
the case for encounter rates and Timed Species-Counts. Similarly, lapses of
concentration, which are likely to occur during long periods of observation,
would have relatively little effect on the results.
Mackinnon Lists and Timed Species-Counts tend to underestimate the
abundance of species which are found in flocks, as they take no account of
the number of individuals encountered. Thus, a flock of thousands would be
recorded the same as a single individual using Mackinnon lists and Timed
Species-Counts. The difference between a flock and an individual is only
taken into account when using encounter rates.
Mist-netting can be used to record species in the understorey which might
otherwise go unrecorded and can also be used to estimate the relative
abundance of understorey species.
Figure 31. Summary of uses, advantages and disadvantages of the methods

described in this section
Method
Simple
Species List
Uses
Determining which
species are present.
Advantages
No analysis
involved. Simple.
Species
Discovery
Curve
Analysing the
completeness of a
species list,
estimating the total
number of species
present and
Allows comparisons
between species lists
for the same site
over time or for
different sites.
Disadvantages
No account taken of
effort, thus difficult
to make comparisons
between surveys and
between sites.
Accurate plotting of
curves requires
computer analysis.
Bird Surveys 99
Encounter
Rate
comparing lists from

different sites.
Annotating a species
lists with an index of
relative abundance
based on the number
of encounters with
individuals per block
of time.
Mackinnon
Lists
list with an index of
relative abundance
based on the number
of encounters with
species per block of
effort. Plotting a
species discovery
curve.
Timed
SpeciesCounts
list with an index of
relative abundance
based on the number
of encounters with
species per weighted
block of time.
Plotting a species
discovery curve.
Mist-netting
Discovery of
secretive understorey
species. Index of
relative abundance
based on the number
of encounters with
individuals per unit
of effort.
Allows crude
comparisons of
abundance between
species within a site
and within species
between sites.
Allows crude
comparisons of
abundance between
and within species
between sites. Data
collection is simple,
allowing the
observer freedom to
roam. Relatively
unaffected by
observer skill and
concentration.
Allows crude
comparisons of
abundance between
and within species
between sites. Data
collection is fairly
simple, allowing the
observer freedom to
roam.
Rigorous.
Prone to bias
resulting in
differences in
species
detectability. The
requirement to count
all individuals of all
species presents
practical problems.
Prone to bias
resulting in
differences in
species
detectability. Underestimation of
flocking species.
Underestimation of
flocking species.
Time consuming,
requires specialised
equipment and
highly trained
personnel.
Species lists and indices of relative abundance can be used for priority
setting for conservation. It is not the number of species present which is of
prime importance but which particular species are present, with judgements
usually being made on the basis of the conservation status of those species
present. BirdLife Internationals Important Bird Areas programme uses four
categories for selecting priority sites:
the presence of one or more threatened species;

the presence of a group of species with a restricted range belonging to an
Endemic Bird Area;
the presence of a group of species restricted to a biome;
the presence of a large number of individuals.
The methods described in this Section can be used to select priority sites
for conservation by making comparisons between different candidate sites or
judging an individual site against a standardised set of criteria such as those
used in the Important Bird Areas programme. Choices between sites with
similar species lists can be made based on the relative abundance of key
species at the different sites. The data gathered by these methods can also be
used as the basis of a monitoring programme, allowing comparisons within
sites over time.
5.6 Sources of information for the recording of bird

sounds
The Wildlife Section of the British Library National Sound Archive (NSA)
has over 100,000 recordings of animal sounds. It covers over 7,500 species of
birds, 770 mammals, 700 amphibians, 700 invertebrates and a few reptiles
and fish. Any recording may be listened to free of charge, by appointment, at
the British Library building in London. Several visitors listening to different
recordings may be accommodated at the same time. Copies of recordings and
spectrograms can be made to order. Advice on recording equipment and
techniques can be given, including equipment loans to approved expeditions,
and the NSA runs an annual workshop in conjunction with the Royal
Geographical Society/Birdlife International.
Contact:
Richard Ranft
Curator
NSA Wildlife Section
The British Library
National Sound Archive
96 Euston Road
London NWI 2DB
Tel +44 (0)171 412 7402/3

Fax +44 (0)171 412 7441
Email: [email protected]
http://www.bl.uk/collections/sound-archive/
Bird Surveys 101
UK
The Library of Natural Sounds (LNS) at the Cornell Laboratory of

Ornithology contains approximately 130,000 recordings covering over 6,000
species of birds. Other holdings include insects, amphibians and mammals.
LNS offers collaborators state-of-the-art archival facilities, database software,
technical advice and field recording equipment loans:
Contact:
Gregory F. Budney
Curator
Library of Natural Sounds
Cornell Laboratory of Ornithology
159 Sapsucker Woods Road
Ithaca, New York 14850
UNITED STATES
Tel +1 (607) 254 2404

Fax +1 (607) 254 2439 or 2415
http://www.birds.cornell.edu
Wildsounds is a supplier of recording equipment and commercially available

bird recordings. Their web-site has a facility to search all available recordings
by species.
Wildsounds
Cross Street
Salthouse
Norfolk
NR25 7XH
UK
Tel/Fax +44 (0)1263 741100

http://www.wildsounds.com
Bird Surveys 103
Section 6
BIRD-HABITAT STUDIES
Colin Bibby, Stuart Marsden and Alan Fielding
6.1 Why study habitats?

Understanding the habitat associations or usage by a species is fundamental to
getting to know about its conservation status. Such information can be
collected by direct or indirect field methods. Beyond the level of descriptive
natural history, there will be a need for some mathematical treatment of data
to identify important habitat associations. As emphasised throughout this
book, analysis of data depends very much on having good data in the first
place and this depends on having methods well suited to the purpose. Habitat
data will usually be collected at the same time as the bird data.
Ecology is the study of relationships between organisms and their
environment, where the environment consists of a set of gradients or
categories. Simple gradients such as altitude are relatively easy to appreciate
and study. More complex gradients, for example forest disturbance by
humans, are more difficult to characterise and often may only be described
mathematically. Any one place, or the requirements of a bird, may lie on
several independent habitat gradients. Explaining the patterns of relationship
between birds and habitats is the challenge for the researcher. The value of
such an approach may come from:
predicting distribution and numbers in unsurveyed areas;
providing an understanding of the nature of the relationship between a
bird and its habitat;
predicting possible consequences of future changes of land use.
It may be necessary to adopt a variety of methods to meet all three of
these aims. Different kinds of investigation take place at different scales, and
this will influence the kinds of field measuring that are possible.
We have separated the large and fine scales. Large scale approaches lend
themselves to questions about distribution and numbers and extrapolation to
areas not actually surveyed. Their planning and measurement relate to the
design stage of studies. Fine scale studies focus on individual birds and
address questions about habitat relationships in more detail. Predicting the
consequences of land use changes may require data at both scales.
6.2 Broad scale habitat studies

The great advantage of a broad scale study is that you can be more confident
about the area to which your findings refer. The downside is the large amount
of preparatory work required and the demand for access to a bigger area. The
approach requires mapping the habitats of the whole target area, which might
measure up to thousands of square kilometres.
6.2.1 What features to map?
Deciding what to map is a trade-off between the kinds of features that are
possible to map and those that are known or suspected to be significant for
predicting the occurrence of birds. What is possible is influenced by the size
of area to be mapped and by the availability of data.
Altitude
If you were only allowed one factor to map, then for many birds, altitude
would have to be the choice. Wherever there is a strong topography, altitude
is a major predictor of the occurrence of individual species. Altitude data are
relatively easy to find from various sources as maps or digital data.
Forest structure
Numerous studies have shown the over riding impact of structure of
vegetation on the distribution of bird species. For large scale mapping, the
difficulty is identifying parameters of the kind which influence birds but
which can also be mapped. Satellite images might be able to highlight
differences between seasonal and evergreen forest. Aerial photographs might
be used to separate different degrees of canopy closure. Either approach or
the use of forestry maps might be able to detect the effects of logging, but this
is notoriously variable in its intensity and often difficult to detect following
regrowth.
Other vegetation
Structural features are also important for birds in more open habitats. It might
be possible to map grasslands, scrub or savannah in terms of classes of
vegetation cover. Water features are important for many birds and can readily
be mapped, though there can be great seasonal variation. Satellite images are
particularly good at detecting water cover. Associations with particular
habitat features, such as one species of fruiting tree, or large hole-bearing
trees in general, may only become obvious from more detailed studies.
Generally, such features are not easy to map on a large scale.
Bird Surveys 105
Soils
Soil maps are readily available. Given the importance of soil types in
influencing vegetation, an impact on birds might be expected. Contrary to
expectation, soils do not seem to be an important factor for bird habitats. In
spite of the ease of obtaining the data, this is probably not worthwhile unless
you have good reason to believe otherwise in a particular study.
Human factors
The ubiquity of human influence is the constant refrain of conservation.
Humans can influence birds and their habitats directly by modifying
vegetation or by hunting. They can also have indirect impacts since habitat
change can alter the impact of predators or allows the spread of invasive
species. Indirect measures such as distance from road or village might be used
as surrogate measures of human impact, so it might be worth mapping roads
and villages from aerial photos, or other maps. Local people might be able to
map the areas they regularly visit.
6.2.2 Sources of data for mapping
There are three common sources for producing habitat maps. They differ
considerably in their costs and effectiveness.
Satellite images
Satellite images are the 'top of the range' approach but they are still expensive
to purchase and process. There is a huge literature on interpretation, which is
a specialist field in its own right. Provided there are no problems with shade
or clouds, satellite images are good for drawing broad brush habitat maps that
show major differences, such as between forest and grassland. If the images
are classified without field data (the normal situation), they may map all sorts
of variation, (e.g. within forest types) but these will not necessarily be clearly
explicable or biologically meaningful. Maps derived from satellite images
need especially careful field verification. If the field effort is sufficient, it may
be possible to use the field data to produce alternative classifications of the
imagery which bear more relation to the birds and their habitats.
Because they are handled on Geographical Information Systems (GIS),
satellite images lend themselves to studies requiring this kind of analysis. For
instance, it is very easy to measure indices of fragmentation, patch size and
isolation from a satellite image.
It is possible and much cheaper to obtain satellite data as false colour
printed photographs. Obviously, these cannot be used for any of the
sophisticated GIS applications. They can however be used to draw
generalised habitat maps to show major boundaries, such as between

evergreen and deciduous forest (if taken at the appropriate time of year).
Aerial photographs
Aerial photographs are generally easier to interpret at a broad level especially
if viewed as stereo pairs. They present much smaller areas at higher
resolution. For this reason they are quite good for mapping plots at the lower
range of sizes no more than hundreds of square kilometres. Aerial
photographs are easier to find and access than satellite images. Obviously,
they do not allow such sophisticated analysis. Typically you might expect to
generate a map which traces up to half a dozen habitat types.
Pre-existing maps
You might be able to find previously published maps from all manner of
possible sources. They may also have political significance, if for instance
they were generated by a forestry service that has management authority.
They have the disadvantage that they may or may not be suitable for your
needs in terms of accurate discrimination of relevant habitat boundaries. They
may also be substantially out of date.
Altitude as already noted is a factor that is important as a determinant of
bird distribution and is reliably available from maps. Contours can also be
extracted from computerised digital elevation models if you are going to
handle any other data in a GIS format.
6.2.3 Verification of a map and sampling
Issues concerned with design and sampling are covered further in Section 2.
Having generated a habitat map, you should use it in the design process. It
should be clear from the above that some of the ways of generating the maps
can be ambiguous. So you will want to verify the map at the same time as
pursuing more detailed habitat assessments (described below). Essentially,
you want to be able to show that the mapped units are internally consistent
and distinct one from another across the range of the map. Provided you have
a range of plots or samples where you have measured bird and habitat data,
then you can make such an assessment in retrospect.
6.3. Fine scale bird-habitat studies

6.3.1 Different approaches/survey designs
The way you link your bird data to your habitat data will depend on the bird
census method you have chosen. After all, it will be easier to collect habitat
data at the same time as your bird data, rather than having to organise a
Bird Surveys 107
separate habitat survey. It is easier to relate habitat to birds with some bird
census methods, such as point counts or line transects, than with others.
Figure 32 shows some basic approaches to bird-habitat studies based mainly
on the census methods described in Section 4.
Figure 32. Relating birds to habitat data using different census methods.
Census Method
Point counts
Line transects
Bird based
studies
Aggregation
counts
Look down
methods
Approach
Collect habitat data at, and around the points themselves
bird presence/absence at points is related directly to the
habitat there
Transect is divided into short sections (50100m in
length) bird presence along each section can be related
to habitat directly.
If focal studies are made of, say the behaviour or home
range of individual birds, then the birds use of the
habitat around it is of primary importance. For example,
within its home range does it stick to forest gaps, edges,
or areas with dead trees? What proportion of time does it
spend in each habitat type? How does the habitat it uses
differ from that which it does not?
Habitat recordings are made at each aggregation. The
same readings might also be made at random points or
places which look as if they might be suitable as
aggregation sites. What features are shown by the
aggregation sites and not by the areas that do not hold
aggregations?
More difficult. Very broad habitat features (e.g. density
of emergent or dead trees) can be estimated from the
vantage point. Habitat readings could be taken along
random transects through the survey area. Bird usage,
either in different parts of a single area or between
different sites could be related to the habitat features
using regression analyses.
6.3.2 What habitat features to record and how?

A species relationship to its habitat is bound to be complex. It is unlikely
during a relatively short study that you will discover exactly what makes the
species tick what it needs for feeding throughout the year or for successful
raising of its young. What you can do, however, is to characterise the sorts of
habitats in which it seems to thrive and those from which it is absent.
The choice of habitat features to record is difficult, and will depend on the
amount of time you have, the type of bird to be studied, its broad habitat and
the threats that it might face. You cannot record everything and you won't
want to because habitat recording is time consuming.
Altitude, gradient and other physical features
Of all possible bird-habitat associations, altitude or elevation is the most
commonly cited. Details of a species altitudinal range (maximum and
minimum), the altitude at which it is most abundant, and interactions between
altitude and the patterns of human habitat alteration can be crucial in defining
species ranges, explaining differences in abundance between areas and hence
positioning protected areas correctly. Altitude is easily measured using an
altimeter or can be taken from maps if these are accurate enough.
Gradient is used less frequently but it can often be important in defining
species distributions. Just as altitude has an effect on vegetation, so does
gradient, with, for example, trees on steep slopes being in general smaller and
those on ridges more likely to be deciduous. Birds may react to these
differences, and there may be a tendency for species diversity and abundance
to be lower on slopes. Gradients can be expressed as degrees or percentages
and be measured or estimated at a single point or expressed as a mean
gradient over several points (you could take gradient measures at four random
points around each census site).
Many other physical features can be recorded, although which ones are
important will depend on the type of bird/habitat you are working with.
Distance to the nearest river or stream can be important in many species,
while in others, the presence or absence of standing water (e.g. ponds,
swamps) can be important. The presence of bare ground, rocky outcrops or
boulders may be significant for some bird species. Although not strictly
physical features, the distance from the census site to the forest edge, to a
logging road, or to the nearest village may be potentially important. If the
census site is outside a forest block, how far is it from the forest edge?
Tree sizes, tree densities and biomass indices
Tree heights and diameters at breast height (DBH) are important habitat
features, which can be estimated, or measured with a tape measure at census
plots, along transects, or at random points within areas. With points, a useful
method is to estimate the heights (or measure using an inclinometer) and
measure the girths of ten trees larger than a given DBH (e.g. >0.2m DBH)
which are closest to the centre of the plot. These values can be used to
calculate the mean tree height and DBH at each census plot. A measure of the
Bird Surveys 109
distance to the furthest of these ten trees will allow you to estimate the density
of trees around the point using;
D = 100,000 . . (dmax )
2
[NB. = pi]
where
D = tree density per hectare
dmax = distance to furthest of the 10 trees (m)
Another useful statistic is basal area of trees per hectare. This can be
calculated again using distance to the furthest tree (dmax) along with the
summed basal areas of each of the ten nearest trees. This measure can be
extended to produce a woody biomass index, which approximates to the basal
area of each tree multiplied by their heights, and can be expressed as volume
of wood per hectare.
Measuring trees is time consuming and hard work, but tree sizes and
densities can be very useful in describing the distributions of all sorts of birds.
As well as considering trees greater than a certain DBH, it may also be worth
counting saplings within a given area, the number of fallen trees, or the
density of stumps left after recent timber removal. If you really do not have
time to measure several trees around your census site, then measuring or
estimating the girth of the largest tree, or largest two trees within your census
area (e.g. within 20m of the centre of your plot) is a useful surrogate.
Tree architectures
As well as their sizes, and whether they are evergreen or deciduous, the
shapes of trees can often be important predictors of bird distribution. One
regime to relate tree architectures to bird distributions was developed from
work by Torquebiau (1986) and by Jones et al. (1995). Each of the ten trees
nearest the census plots central point was allocated to one of the following
groups (Figure 33);
Branching above half its height (A): trees which have grown up under the
closed canopy of primary forest tend to have their first major branch well
above half their height;
Branching below half its height (B): trees which have grown up in open
canopied areas usually branch below half their height;
Branching above but with scars (C): trees which have major scars from
dropped branches tend to be characteristic of regenerating forest i.e. they
have grown up under a canopy that is closing up following a tree fall or
logging extraction;
Vertical branching below half its height (D): an alternative to (C) where
the lower branches are maintained but grow vertically into the closing
canopy.
Quantification of the above architectures, expressed as the proportions of
trees at a site which show (A), (B), or (C+D) are useful indicators of the
recent history of forest at census sites. Another useful measurement is the
number of dead trees and stumps at sites. Relating bird presence or
abundance to the proportions of different tree architectures at sites can yield
information on whether the species prefers primary forest (A), recently
disturbed forest (B), or older regenerating forest (C+D).
Figure 33. Tree architectures. A = branching above, B = branching below,

C = branching above with scars below, D = branching below with vertical
growth.
Foliage profiles/canopy covers

The percentage of vegetation cover at various strata of different habitats can,
if estimated reasonably accurately, be useful in describing bird distributions.
A useful regime is to estimate vegetation covers at ground level (01m), lowlevel (15m), mid-level (520?m), sub-canopy, and canopy, within a circle
around the recorder of radius about ten metres. Quantification of vegetation
covers is very difficult, so estimation is necessary. This should be well
practised and standardised between different recorders. Disturbed forests and
non-forest areas will have, in general, sparser canopy cover than primary
forest, and there may be a more extensive ground and shrub or understorey
layer. There can, in many habitats, be a negative correlation between the
amount of vegetation in the upper and lower strata e.g. in primary forests
most foliage is in the canopy and sub-canopy, while the understorey is sparse.
Bird Surveys 111
Plant/tree species recording

While it is desirable to relate bird distribution directly to the tree and plant
species which occur in an area, this is often difficult. Tropical tree species can
be very difficult to identify and most occur at extremely low densities, which
makes linking a bird's occurrence directly to the presence of a particular tree
species problematic. This situation may be helped if trees are lumped into
genera or groups, or if a technique such as Correspondence Analysis is used
to identify gradients along which communities of tree species may occur.
There are some situations where recording tree species may be very useful
in fairly short studies. The abundance of fruiting trees, such as figs (Ficus)
may be related to the abundance of frugivores. Some species, such as parrots
or hornbills, may nest in just a few characteristic tree species and the density
of these could be recorded, and related to bird abundance. Some pioneer trees
such as Macaranga or Lecropia can be characteristic of forest disturbance
and the abundance of these can be used both to classify forest types and to
relate to bird species occurrence or diversity.
Indices of human impact
In conservation studies, it is often important to quantify human impact on
habitats. To a large extent, human impact through timber extraction will be
characterised by changes in forest biomass and tree size parameters, different
tree architectures and tree species compositions. However, in some situations,
more direct measures of human impact may be appropriate. The distance from
a forest site to the nearest settlement or road may be correlated with the
degree to which it has been disturbed or is used by people. Similarly, the
number and width of paths in an area may be useful in characterising human
disturbance, as may counting people directly. Which measures of human
impact you choose will depend on the sort of impact thought to be important
in the species studied.
6.3.3 Preparation of data for analysis
Ensuring accuracy
Estimation of habitat parameters saves time and energy, but measurements
will always be more accurate than estimations. An alternative to estimating
many parameters is to measure a few, though in this case you must try to
ensure you will record the right habitat features. Many of these will be
correlated with each other (see later), so for maximum effect for minimum
effort, you might try to measure or carefully estimate around ten variables
which are thought not to be closely autocorrelated.
As with estimating distances to bird contacts (Section 3), estimates of tree

girths, distances to trees, percentage vegetation covers and other parameters
are open to between-recorder errors. There may also be a tendency for
recorders estimates to drift during the fieldwork. Periods of habitat recording
training prior to fieldwork and re-training during the fieldwork are essential if
your habitat data are to be meaningful. With some parameters, such as
vegetation cover, it may not be the actual value that is important, but how that
value differs from the values of other sites.
If you think that there may be a problem with the accuracy of your habitat
data, then rounding up during the analysis phase might help. For example,
ground vegetation covers, initially estimated to the nearest 5% could be
lumped into 10% or even 20% bands. At its extreme, continuous data could
be transformed to a simple one-zero format and perhaps some particularly
inaccurate data points excluded from the analysis. Finally, if different
recorders estimates of, say, canopy cover are clearly different, then all values
for a particular recorder could be scaled to the same mean as those of a
different recorder.
Identifying autocorrelation
Many habitat features will be correlated with one another. For example, trees
with large girths will tend to be tall, while forest with a full canopy may have
a sparse ground cover (Figure 34). If habitat features are correlated with one
another, then it may not be necessary for you to record all of the features.
Identifying autocorrelation between your habitat variables can save you time
in the field and will also simplify your habitat data set. During the data
analysis phase, relating your bird data to relatively few habitat variables is
usually better than relating them to many (as long as the few habitat features
you have considered are appropriate). This is because the more different
habitat features you consider, then the greater is the chance that you will
come up with spurious correlations between your bird distribution and your
habitat features. An alternative to collecting fewer habitat variables is to
condense your many, inter-related habitat features onto just a few axes of
habitat variability (section 6.4.5).
Figure 34. Correlation matrix of selected habitat variables (data from forests
on Sumba, Indonesia). Figures shown are Pearsons Correlation coefficients.
* p<0.01, ** p<0.001, NS: not significant.
Canopy
Altitude
+0.21**
Low
Ground
+0.21**
+0.09 NS
Average
Girth
-0.14*
Above
+0.27**
Bird Surveys 113
Canopy
Low
Ground
Average
Girth
-0.28**
-0.41**
+0.49**
+0.26**
-0.11 NS
-0.20*
+0.44**
-0.24**
-0.33**
+0.09 NS
Data transformations
Habitat data will often need transforming prior to analysis. Values of some
variables will have a positive skew (too many very large values). In these
cases, values (x) should be log transformed (log(x)), or if some values are
zero in the original data, then log(x+1) should be used. If values of a habitat
variable conform to a Poisson distribution, then square root transformation
will make the distribution of values more symmetrical (or square root (x+0.5)
if there are zeros in the data).
Many habitat variables will be collected as percentages (e.g. foliage
profiles) or proportions (e.g. tree architectures). In these cases, arcsine
transformation is usually performed. Computer programs will require you to
transform the original data to proportions (between 0 and 1) before arcsine
transformation.
Preparing your bird data
The sort of bird data you have will depend on the census method you used.
Most commonly, you will be relating habitat features to the presence or
absence of birds in a particular area. This might be variable circular plots at
which you did and did not record a bird species, or sections of a transect, or
small patches of forest in which a bird species sang and those in which it
didn't. Each of your census points then becomes either a positive (species
recorded) or a negative (species not recorded), and you look for differences,
in terms of habitat, between the positives and negatives. An important
consideration is the degree of confidence you have in your bird data - how
sure are you that some of your negatives werent in fact positives but you just
failed to record the bird there? Clearly, the more times you have looked for a
bird in an area and failed to record it, then the more certain you are that it is
not there. You may want to only classify as positive those plots at which you
recorded the bird perched (i.e. not just flying over the area) and those at
which the species was recorded at less than a certain distance away from the
plot (e.g. 30100m). Also, if you are not certain if the plot is positive or
negative, then you may want to omit it from the analysis.
Because it is easier to prove presence than confirm absence, a better

alternative is to use the abundance of the bird species at each of your sites. If
for example you repeat-surveyed each of your census sites ten times, then you
could express the abundance of a bird as the proportion of times out of ten
that you recorded it at each site. It may be appropriate to combine bird data
from several census plots together for analysis. All stations along one transect
could be combined, as could all stations which share similar altitudes. Even
better would be an estimate of bird density at each site, so long as estimates
were precise enough.
6.4 Analytical approaches

6.4.1 Summary statistics
Although there is no doubt that all birds live in a multivariate world, it is still
possible to obtain important information by studying the relationships
between a species and a single habitat feature. Univariate and bivariate
statistical tests should always precede more complex methods.
A simple table of means or median values for different habitat measures
can contain a lot of information about habitat associations. Such tables can
help in the interpretation of multivariate analyses. The inclusion of measures
of dispersion, such as the standard deviation, is also helpful. Statistical tests
such as Students t or the MannWhitney U can be used to compare single
habitat variables between sampling units with and without a species. If data
are available from more than two areas, the habitat uses can be compared with
an analysis of variance or an equivalent non-parametric method.
6.4.2 Indices
The index approach investigates if resources are used in proportion to their
availability by calculating an index (Bookhout, 1994). Neus method is
commonly used (Figure 35).
Figure 35. Neus selection index using simulated data.
Habitat
Primary forest
Secondary forest
Logged forest
Agriculture
Total
Calculations:
Selection index
Availability
Usage
Index
Proportion a
Records
Proportion r
Selection w
Standardised
0.503
0.220
0.145
0.133
1.000
47
21
3
0
71
0.662
0.296
0.042
0.000
1.000
1.316
1.344
0.291
0.000
2.951
0.446
0.455
0.099
0.000
1.000
w = r/a
e.g. 0.662/0.503 = 1.316
Bird Surveys 115
Standardised index
B= w/w
e.g. 1.316/2.951 = 0.446
If the selection index >1 the habitat is preferred usage is greater than
availability. Standardising the indices allows comparison between studies
because they always sum to one.
6.4.3 Graphical and linear regression approaches
If data have been collected from VCPs or transects, there will be quite a few
data points. These can be ranked along a habitat gradient and grouped. Either
the percentage of points occupied or the density estimate derived from
DISTANCE can be plotted against a habitat gradient. Figure 36 shows the
relationship between population densities of a parrot species and altitude in
eight forest patches on an Indonesian island. In this case there is a negative,
non-linear relationship between bird abundance and altitude. Such data are
often tested by a regression analysis which models the relationship between
bird abundance and habitat.
Figure 36. Relationship between cockatoo density and altitude.
A caution is appropriate here. There is a theoretical notion that species
live among gradients
of habitat. These will undoubtedly be of many
1.6
dimensions, but imagine the simplest one dimensional example, say between
1.2
abundance and altitude
for the study bird (Figure 37). There is a preferred
Birds
per is where the highest densities occur. Above and below this
altitude,
which
0.8 factors come in perhaps competition from a congener
altitude,
constraining
sq.
km
or maybe weather
0.4factors. At these altitudes, the bird is less common and
eventually, some way from its optimum, it never occurs. If this situation was
studied, the mean (or
0 median) altitude and its dispersion would well describe
the findings. A correlation
of abundance
with
show no
0
200
400
600 altitude
800 would
1000
relationship. In another place, the forest available for study may be limited in
Altitude
(metres
above
altitude because the particular
mountain
is less
high sea
or thelevel)
lowland forest has
been felled. Here, the mean altitude is a biased estimate of the preferred
altitude. On the other hand, there will be significant relationships between
abundance and altitude (one positive and one negative) depending on the
circumstances studied. Clearly these results depend on the circumstances in
which they were measured and cannot be generalised.
Figure 37. The effect of survey area on bird-habitat correlations. The
example shows a bird species which occurs at maximum abundance at
around 600 m a.s.l. If study areas included only altitudes less than 600m,
then there would be a positive correlation between bird frequency and
altitude (A). The converse would be found if only altitudes above 600m were
surveyed (B). Study design C gives no linear relationship.
6.4.4 Logistic regression

If the data take the form of presence or absence at a series of plots (VCPs or
transect stretches), a linear regression model is inappropriate, since each
record can only take the value of 0 or 1, with nothing in between. Logistic
regression is a powerful approach which can handle this problem, either for a
single variable or for many (Jongman et al. 1995). There are ways to pare
down the predictor variables to a small but effective (parsimonious) set. An
attractive feature of the approach is that the model equation predicts the
probability of the bird occurring at a point with any one set of habitat
variables. So, within the normal constraints of regression models, you have an
equation with the capacity to model effects of habitat changes. Logistic
models might be developed for a number of species.
6.4.5 Reducing the dimensions
It is typical to measure many habitat variables which turn out to be correlated
one with another. For instance, shrub cover might be inversely related to
canopy cover (because scrub grows best where the canopy lets in more light).
A canopy dwelling nectivore would be positively correlated with scrub cover
but the relationship is deceptive. One way round this is to reduce the
dimension of the habitat variables. Principal Components Analysis (PCA) is a
common approach. The output is a small set of variables which are weighted
sums of the original variables, explain a high proportion of the original
variation and are independent of each other. Provided they can be interpreted
in plain language, this can be quite a helpful approach. The difficulty is that it
is possible to explain almost anything with enough imagination. PCA, logistic
regression and other multivariate techniques are available in SPSS (Noruis
1993).
Correspondence Analysis or Detrended Correspondence Analysis are
related to PCA. The latter is particularly popular in ecology and is
implemented by a computer program DECORANA. DECORANA avoids
some of the statistical assumptions of PCA which ecological data often
violate. The habitat data set might be simplified by DECORANA before
being entered into regression models.
It is possible to go one stage further and ask how bird communities might
relate to habitat structure. In this case, a set of models for individual bird
species is not very helpful. Canonical Correspondence Analysis (CCA) is a
technique that identifies correlations between simplified bird and habitat axes.
A program called CANOCO can implement CCA (ter Braak, 1987).
Bird Surveys 117
6.4.6 Interpreting and testing the results

The correct way to view these techniques is that they help with the
exploration of data and the generation of hypotheses to explain what is
happening. So, you may be able to say that a bird species is associated with
low altitude, or that it occurs at higher population densities in forest areas
with large trees than in logged areas. Such hypotheses should then be
subjected to independent testing on new data sets. These data sets could be
from a different area within the region, or in a different habitat. Several
methods are available to test the validity of the habitat models you have
formulated:
Resubstitution
The same data are used both for training

(formulating the model) and for testing
(in the new situation). Tends to
overestimate predictive power.
Prospective sampling
The habitat model is developed from the

original data, the model is tested on a new
sample of cases (e.g. from a different
area).
Data partitioning
The original data set is split up into a

training and a test subset. The validity of
the training model is tested on the test
subset. Techniques include Bootstrapping
and Jack-knife sampling.
Section 7
MAXIMISING THE IMPACT OF THE WORK

Colin Bibby
7.1 Basic communications

The political and economic issues surrounding biodiversity loss make the
subject of resource conservation a sensitive one. Local people may be
suspicious of outsiders. Government and people may have conflicting views
as to who should benefit from natural resources such as timber and who
should pay for the consequences of damage to the environment. Very
commonly the benefits are taken by external groups and in a short term while
the price is paid locally or by future generations. Protected areas may be
perceived differently in a government office and on the ground. For these
reasons, biologists from the capital or from abroad need to be sensitive to the
fact that their aspirations for conservation may not be understood or shared by
local people.
Communication is about dialogue. Maximising the impact of a study
involves working out who the target audiences need to be and what messages
are appropriate. Methods of delivery can then be considered. A simple
scheme is illustrated in Figure 38. Note that some of the messages are written,
some spoken and some communicated non-verbally, by behaviour and
attitude. Scientists tend to think about the written word and its quality. The
majority of the people directly involved in the environment do not. A
technically excellent study might have limited or even negative impact on the
ground if the scientists were perceived as rude, discourteous, or insensitive to
local issues and culture.
Figure 38. A simple communications matrix for maximising the impact of a
bird survey.
Audiences
Local people
Messages
We are interested in
this area because..
Delivery
Clear verbal
Our interests are not a

threat to you
Tactful and respectful

attitude
Bird Surveys 119
Regional and national

technical (e.g. forestry
department, protected
area manager)
Here is some
information that you
may find useful
Scientific NGO or
government; national
or international
Here is a sound
description of the
status of a bird, a site
or a habitat
Openness
Good diplomacy
Clear and simple
written reporting which
fits with their needs
Scientific publication
Unpublished report
Archive data
The underlying causes of threat to the worlds birds are some very large
issues (Figure 39). Amongst these, lack of knowledge and poor use of the
information that is available is only a relatively small part. Knowledge about
birds is again only a small part of the total gap in data and information
required on ecology, politics and economics. In spite of this cautionary note,
ornithologists have been major leaders in the conservation world for the
simple reason that they have been good at exploiting the value of birds as
indicators and communicating key information in an effective way.
Figure 39. Fundamental causes of biodiversity loss.
Unsustainably high rates of human population growth and natural

resource consumption.
Steadily narrowing spectrum of traded products from agriculture and
forestry, and introductions of exotic species associated with agriculture,
forestry and fisheries.
Economic systems and policies that fail to value the environment and
its resources.
Inequity in ownership and access to natural resources, including the
benefits from use and conservation of biodiversity.
Inadequate knowledge and inefficient use of information.
Legal and institutional systems that promote unsustainable exploitation.
From the Global Biodiversity Strategy (WRI/IUCN/UNEP 1992)

Figure 38 could be considerably elaborated for a particular study. For the
purpose of this section, its three major divisions of audience will suffice.
7.2 Culture, politics and diplomacy

Diplomacy is about being sensitive to circumstances which influence the
prospects for effective dialogue.
Legal requirements
Complying with legal requirements can be quite hard work and may appear to
be slow and bureaucratic. Ignoring legal requirements puts your own work in
jeopardy but is also unhelpful to future relations between scientists and
whoevers regulation it is that you have ignored. Things that might require
permits over and above entry visas include travel, access to protected areas,
research, mist-netting, collecting specimens (birds or anything else), or
exporting specimens. To be effective, you need to find out what permits are
required and how to get them you may need letters of invitation or
reference, photographs, or a preceding permit such as the right kind of visa.
You will certainly need time.
Local involvement
If you are working in a distant place, in your own country or abroad, one
essential step to good diplomacy is to work with local people. Students from
the region will often appreciate the chance to join a survey and learn new
things. You might employ guides or other local helpers. They will pay for
themselves abundantly in their ability to speak the right language, interpret
local nuances of behaviour or meaning, know how to go about things in the
area, and maybe even know some of the birds. Few sponsors will want to
support expeditions without local involvement for the simple reason that they
are less likely to be successful. If you ask for help from local individuals or
organisations remember that you might be involving them in costs.
Internationally funded studies should certainly be sensitive to this and willing
to defray such costs.
Cultural sensitivity
The risk of causing offence by oversight of cultural differences is ever
present. The best way to deal with it is to have local participation who will be
sensitive to the importance of everything from how you sit, how you dress,
how you eat, where you wash and which sacred forest is a no-go area even if
it is the oldest and best forest in the area. The only other rule is to behave at
least as well as would be expected at home even if it is tempting to do
otherwise while living in the forest. You will be ambassadors for conservation
and your impact, for good or bad, will last longer than your visit.
Respect
You will never communicate anything to someone who feels that you have
slighted them with disrespect. This might include failing to visit them and talk
and keep them in touch with what you are doing. It takes time which could
otherwise be spent in the field but if you want the results to have impact then
you need your audience on your side.
Bird Surveys 121
Talking
Essentially good diplomacy comes down to forming good relations with
people being interested enough to talk with them and listen to them. Avoid
being condescending but tell people what you are doing and why in plain
language. If you can speak even a few words of the local language, you will
come over as much more genuinely interested and involved. Local people
may very well not know that the bird which occurs in their forest is very
special and does not occur all over the world. You can give people something
to be proud of by helping them to understand better the birds in their own
area. At the same time, listen to what they have to say. Rural people may be
very knowledgeable, but in a different way from that understood from a
scientific training.
Figure 40. Five areas where good diplomacy will help your work to be
influential. Poor diplomacy may override otherwise good work and thwart its
influence.
Legal requirements
Local involvement
Cultural sensitivity
Respect
Talking
7.3 Summary reporting

Field expeditions often produce two kinds of report. Commonly they produce
their own report for limited circulation to helpers, sponsors or friends. They
might later go on to publish something in the scientific literature. I will
discuss these two separately though it might be a good idea to use the
scientific paper as your summary report if you can write it quickly enough.
A few key considerations can increase the chances that your summary
report will have the best chances of being effective.
Figure 41. Key points for the impact of your technical report
Get it to the right people

Produce it reasonably quickly
Give people relevant information
Summarise the key results
Why are these results important?
Keep the report brief
Leave out irrelevant material
Avoid naive political or economic recommendations
Get it to the right people

Good contacts and discussions should have helped in the process of
identifying the key people from local communities, regional or national
government departments and NGOs who might welcome your report. Some
of them will already have been helpful in organising permits or other ways.
These people will be looking forward to hearing about your findings,
especially if you have communicated well with them during the trip.
Remember to acknowledge and thank all your helpers and supporters.
Consider translating it into the appropriate local language if this is different
from yours they will feel even better about the project when they see this.
Get it written by the right people
The further you are from home, the more important is the authorship. Local
co-authors are the more likely to feel a share in ownership of the work if they
have been given a chance to make a real contribution as co-authors. Local
ownership is likely to make a big difference to the long term impact of a
project.
Produce it reasonably quickly
If you take ages to produce a report you give the impression that it is not very
important to you. How, in these circumstances can you expect your readers to
think that your work is important. Some permits or grant-giving bodies may
request a report within a given time period. Foreign visitors with research
permits in some countries are expected to report before leaving. This need not
be very difficult and is certainly a good discipline. Remember that all sorts of
other things in life will take over once you have left and it might seem very
difficult to write anything. It also gets harder to write once the fieldwork
seems a distant memory. One way to increase the chances of writing a quick
report is to be thinking about it while in the field. Keep track of the data you
are collecting as they come in. Add up simple statistics as you go along such
as the numbers of point counts completed in different habitats and the
numbers of species recorded. Keeping all the data in one place rather than
entrusting them to various peoples field notebooks is another way to save
time when it comes to sorting results out.
Give people relevant information
Remember who your audience is and provide them with information that they
might find interesting and useful. This calls for some thought and
understanding of the context of the recipients. What powers might they have
to do something with the results? Which results do they most need to
stimulate some action? Remember that many of your important readers will
Bird Surveys 123
not share your passion for conservation but will see it in relation to their lives
and jobs.
Summarise the key results
Results are much easier to read if the author knows what the key points are
and has made them clearly. A good discipline to help do this is to write the
summary first as a series of single sentence factual points. There should be no
more than about 1015. Get these into a logical sequence and you then have
not only a summary but also a synopsis of the report which will then almost
write itself. Any material which is not logically needed to support the
summary is of questionable value. The summary should come first many
readers will only have time for the first few pages.
Why are these results important?
Remember to think why your results are important and set sufficient
background context to draw out their value. Point out that your data extend
the known range of a globally threatened species. Indicate that you have
found a very rich concentration of species in an Endemic Bird Area. Say that
the data you have collected provide a baseline against which future changes
of management in a protected area might be monitored. Be proud that you
have trained a local student to be able to conduct further bird surveys in the
region.
Keep the report brief
Short reports are harder to write than long ones but much easier to read.
Many of your readers will be busy. Your report may not be in the language
that they find easiest to read. It is a courtesy to readers to make the effort to
express yourself briefly and clearly and it greatly enhances the chances of
them reading and taking in what you have said. It is a mistake to believe that
long and heavy reports are a sign of a weighty piece of work. I can think of no
reason why anyone might prefer a long report to a brief and well written one
try to aim for a maximum length of ten pages. These comments apply to all
possible readers including friends, funders or potential future employers
whom you will want to impress. It is a courtesy and a help to local readers to
provide at least the summary in the right language. Your local collaborators
will presumably be able to translate if for you.
It may be that you have extensive scientific analysis to make and these
will not be ready in time for a report to be read by your first target audience.
Dont worry. The scientific report can come later. It will undoubtedly include
some further new information. It may well be that you can already make many
of the key points without the analysis you can certainly report on the areas
that the analysis will cover. You could say for instance that 1695 bird
records were collected at 120 point counts in six different habitats. Locally
special (endemic) species appeared to be more frequent at higher altitudes
but richness of species was greater in the lowlands. A further analysis will
correlate species occurrence with habitat features. This will not prejudice
your scientific report but will communicate much of what you have done and
found to a key audience. There is a risk that your summary report will be
quoted by other authors as if it was a scientific paper. Be scrupulously careful
about what you say to avoid being misunderstood. Numbers and standard
errors are a particular problem. If you have estimated a population as between
2000 and 10000 say so. 6000 4000 might be misquoted as 6000! And what
about the units? Did you count pairs, singing males, or the whole population?
The literature is full of erroneous quoting of pairs for individuals and vice
versa.
Leave out irrelevant material
It should be obvious from the above that irrelevant material has no place in
such a report. This is a difficult lesson. It is much easier to write pages of
irrelevance than to produce a simple and sharp summary report. Think
carefully about your audience before describing how you travelled, what you
ate, what medicines you took, who lost their binoculars and sundry items that
I have read in many a report from a field trip to an exciting place. Who wants
to know about this sort of thing? Is it just yourselves? If so, the report is no
place for it. Keep these topics for newspaper pieces or talks where they will
go down well. If you must, then why not have a 200 page appendix to a brief
report?
Avoid nave political or economic recommendations
Unless you are very knowledgeable on the political and economic
circumstances of the area you are studying, it may be ill advised to make any
comments on these obviously sensitive subjects. Doing so risks annoying or
offending the reader and may encourage them to suspect that if your politics
are so nave then perhaps your ornithology is as well. It is legitimate to say
that the forests on such and such a mountain are being extensively cut and this
is having an adverse effect on this threatened species whose population has
become so small as to be at risk of extinction. To say that this practice must
be stopped risks sounding nave if you do not know about its legality or its
economics (it may be the greatest source of revenue to the area). It may be
legitimate to say that such and such a species is hunted by people in such and
such a village. Again it may be nave to say that this must be stopped the
relationship between the village and the management of the national park may
Bird Surveys 125
be a matter of considerable political sensitivity. There are places where these

issues can be addressed but a summary report of a field expedition is not one
of them. Campaigning NGOs find it easier to base their arguments on
scientific data if they can claim that the data are collected by sound scientists
simply reporting facts. The local NGO can add the political spin it judges
appropriate in advocating a particular action. If a scientific reports appears to
contain questionable political or economic sentiments it is easy for officials to
dismiss.
Make it look good
It may seem a shame but the appearance of a report is almost as important as
its content when it comes to making an impression. With modern word
processing it is not difficult to make a report look smartly designed and laid
out. The time put into this will certainly give a better impression. Colour
photographs can be reproduced quite cheaply but do not use them unless they
say something worthwhile.
7.4 Scientific reporting

I do not know who first said that work unpublished is as good as work not
done at all. Excepting the consideration that a summary report might be
enough for some target audiences, this aphorism is largely true. The reason
for publishing a scientific paper is that your information will potentially be
available to ornithologists and conservationists the world over and for all
time. A published paper means that the data and analysis have been refereed
and also that they can be checked by someone doing subsequent work.
Published papers are valuable, of course, for career prospects and personal
satisfaction, if this is important to you.
A sensible precaution before writing a scientific paper is to choose the
journal in which you would like to get it published. Read a few issues to see if
they contain the sort of paper you expect to write. Read the guidelines for
authors to check that your paper will be written within the scope of the
journal and appropriate in length and style. Before you finish, pay scrupulous
attention to the guidelines on such matters as citation of references, layout of
legends, headings and other conventions. There is no greater way of irritating
editors than to ignore their rules and guidelines. To do so suggests that your
respect for the journal is low and irritated editors may be less sympathetic if
they have any other reasons to hesitate over accepting your work.
There are many texts on writing scientific papers but not much sign from
the literature that all authors have ever read any of them! Many of the points
in the previous section are just as relevant to a scientific paper as to a
summary report. Clarity of a paper depends on brevity, simplicity and good

organisation. Writing a draft of the summary first is a good idea because it
give focus to the important content of the paper which the rest needs to
document and support. If the summary consists of single sentences each of
which summarises something of the aim, method, finding or conclusion, so
much the better. I find it helpful next to plan the tables and figures needed to
document the evidence and present the analytical arguments and conclusions.
Few journals will accept more than about ten tables and figures in total, but if
they are the right ones, you can say a lot with ten. Major headings and a
synopsis come next. After that it is easy! It should be pretty obvious what is
needed to hang it all together.
Essentially your paper needs to say why your study was important, how
you did it and what you found. You will have dealt with the purpose of the
study in the planning stage (and having read Section 1). The methods section
needs to deal with location, field methods, steps taken to minimise bias, and
how the effort was designed and distributed. A key test, within the confines of
brevity, is that the reader should have a good chance of going to the same
area and conducting a comparable study 20 years from now. The results
section needs a general overview of the data collected and portrayal of the
analysis from which conclusions were drawn. Show your draft to other people
for comment. Try it on scientists but also on the person nearest and dearest to
you. He or she may not follow the scientific detail but they will tell you what
needs improving with greater honesty than most other people might.
Figure 42. Key points for a scientific report
Summary
Write this first
1015 single sentence factual points
Informative and interesting in its own right
It is all that some people will read
Methods
Write this last
Sufficient detail to be repeatable
Emphasise sampling design and control of bias
Keep it brief
Results
Up to 10 tables or figures to tell the story clearly
Sound statistical analysis to support the conclusions
Enough of the detailed findings summarised for future to communicate the
shape of the basic information but editors will not allow masses of data
Discussion
Say why the study is important
What are the most significant results?
Bird Surveys 127

How do the findings fit with previous knowledge on the species or place or
whatever?
What further scientific work is required?
7.5 Archiving the data

You may well have far more detailed information than you can readily
publish in a scientific paper but which is potentially very valuable for the
future. This can be written as an archive which might contain detailed maps
of study plots and census routs, original census data referring to mapped and
dated transects or point counts. Most likely, these data will have been
computerised and can be archived as a database or spreadsheet file but most
safely will have a paper printout as well (or alone). You may also have a
systematic list of miscellaneous bird observations which are potentially
valuable but generally too bulky to be acceptable to a journal. The archive
report will be weighty but it has a limited audience. Most people would prefer
to see either a short report or a scientific paper and only rather few will want
the full details.
Your archive data would be appreciated by the nearest relevant
institutions, be they a university, protected area headquarters, local
government office or NGO. There may be national conservation
organisations, such as a BirdLife Partner who would value and could use such
information. Many countries are currently working on establishing national
biodiversity data centres often as partnership organisations hosted by
government but supported by NGOs. If there is a national ornithological or
bird conservation NGO this would be another place to leave an archive copy.
Finally, BirdLife International maintains an extensive database and reference
library of published and unpublished information on globally threatened
species and important sites. This material is available to and much used by
people planning further studies or seeking particular information on
threatened species. Unpublished material held in BirdLifes library is cited in
Birds to Watch, the global inventory of threatened birds. I hope the results of
your study will contribute to the next edition. You can ensure this by
publishing or by lodging unpublished archive material for the use of others.
Section 8
REFERENCES AND FURTHER READING

Allport, G., Ausden, M., Hayman, P.V., Robertson, P.A. and Wood, P.N.
(1988) The Birds of Gola Forest, Sierra Leone. ICBP Study Report 38,
Cambridge.
Ausden, M. and Wood, P.N. (1991) The Wildlife of the Western Area Forest,
Sierra Leone. Special Report to the Forestry Department, Sierra Leone.
ICBP/RSPB, Sandy, Bedfordshire, UK.
Bennun, L. and Waiyaki, E.M. (1993) Using timed species-counts to compare
avifaunas in the Mau Forests, South West Kenya. Proceedings of the VIII
Pan-African Ornithological Congress 366.
Bibby, C.J., Collar, N.J., Crosby, M.J., Heath, M.F., Imboden, C., Johnson,
T.H., Long, A.J., Stattersfield, A.J. and Thirgood, S.J. (1992) Putting
Biodiversity on the Map: Priority Areas for Global Conservation.
International Council for Bird Preservation, Cambridge, UK.
Bibby, C.J., Burgess, N.D. and Hill, D.A. (1992) Bird Census Techniques.
Academic Press, London.
Blake, J.G. (1992) Temporal variation in point counts of birds in a lowland
wet forest in Costa Rica. Condor 94: 265275.
Bookhout, T. A. (ed) (1994) Research and Management Techniques for
Wildlife and Habitats. 5th Edition. Wildlife Society, USA.
Buckland, S.T., Anderson, D.R., Burnham, K.P. and Laake, J.L. (1993)
Distance Sampling: Estimating abundance of biological populations.
Chapman & Hall, London.
Collar, N.J., Crosby, M.J. and Stattersfield, A.J. (1994) Birds to Watch 2, The
World List of Threatened Birds. BirdLife International, Cambridge, UK.
Collar, N.J., Gonzaga, L.P., Krabbe, N., Madroo Nieto, A., Naranjo, L.G.,
Parker, T.A. and Wege, D.C. (1992) Threatened Birds of the Americas:
The ICBP/IUCN Red Data Book. International Council for Bird
Preservation, Cambridge.
Collar, N.J. and Stuart, S.M. (1985) Threatened Birds of Africa and Related
Islands: The ICBP/IUCN Red Data Book. International Council for Bird
Preservation, Cambridge.
Evans, M.I. (1994) Important Bird Areas in the Middle East. BirdLife
International, Cambridge, UK.
Grimmett, R.F.A. and Jones, T.A. (1989) Important Bird Areas in Europe.
Technical Publication No.9. International Council for Bird Preservation,
Cambridge, UK.
Bird Surveys 129
IUCN Species Survival Commission (1994) IUCN Red List Categories.

Prepared by the 40th Meeting of the IUCN Council, Gland, Switzerland.
Jones, M.J., Linsley, M.D. and Marsden, S.J. (1995) Population sizes, status
and habitat associations of the restricted-range bird species of Sumba,
Indonesia. Bird Conservation International 5: 21-52.
Jongman, R.H.G., ter Braak, C.J.F. and Van Tongeren, O.F.R. (eds) (1995)
Data Analysis in Community and Landscape Ecology. Cambridge
University Press, Cambridge, UK.
Kapila, S. and Lyon, F. (1994) Expedition Field Techniques: People
Oriented Research. Expedition Advisory Centre, London.
Laake, J.L., Buckland, S.T., Anderson, D.R. and Burnham, K.P. (1994)
DISTANCE User's guide Version 2.1. Colorado Cooperative Fish and
Wildlife Research Unit, Colorado State University, Fort Collins.
Mackinnon, J. and Phillips, K. (1993). A Field Guide to the Birds of Sumatra,
Java and Bali. Oxford University Press, Oxford.
Marian, W.R., O'Meare, T.E. and Maehr, D.S. (1981). Use of playback in
sampling elusive or secretive birds. Studies in Avian Biology 6: 8185.
Noruis, M.J. (1993) SPSS for Windows Professional Statistics 6.1. SPSS
Inc., Chicago.
Pomeroy, D. and Tengecho, B. (1986). Studies of birds in a semi-arid area of
Kenya. III the use of timed species-counts for studying regional
avifaunas. Journal of Tropical Ecology 2: 231247.
Rabinowitz, D. (1981) Seven forms of rarity, in Synge, H. (ed) Biological
Aspects of Rare Plant Conservation. Wiley, Chichester: 205217.
Robertson, P.A. and Raminoarisoa, V. (1997) Manuel de Formation Pratique
en Ornithologie. Projet ZICOMA, BirdLife International, Antananarivo.
Stattersfield, A.J., Crosby, M.J., Long, A.J. and Wege, D.C., (1998) Endemic
Bird Areas of the World. BirdLife International, Cambridge, UK.
ter Braak, C.J.F. (1987) CANOCO a FORTRAN program for canonical
community ordination by correspondence analysis, principal components
analysis and redundancy analysis (version 2.1). Agricultural Mathematics
Group, Wageningen, 95pp.
Torquebiau, E.F. (1986) Mosaic patterns in Dipterocarp rain forest in
Indonesia and their implications for practical forestry. Journal of Tropical
Ecology 2: 301-325.
Wege, D.C. and Long, A.J. (1995). Key Areas for Threatened Birds in the
Neotropics. BirdLife International, Cambridge.
WRI, IUCN and UNEP (1992) Global Biodiversity Strategy: Guidelines for
action to save, study and use earths biotic wealth, sustainably and
equitably. UNESCO, Paris.
Section 9
SAMPLE INPUT AND OUTPUT FILES FOR

DISTANCE PROGRAM
Data are entered into DISTANCE in the form of a syntax file. In this file, as
well as entering your data set, you will need to enter several commands
specifying the format that your data are in, i.e. whether you used line transects
or VCP methodology, whether measurements are in metric or imperial units,
etc. In addition, DISTANCE provides a number of optional commands that
can be used to override the default program and tailor it to your particular
data set/specifications. The commands you choose will vary depending on
study design, methodology used to collect the data and what you need in the
output.
To illustrate these points, we have included two basic examples of real
syntax files that can be used as guidelines. You will of course want to tailor
these files to your own particular needs and should consult the DISTANCE
users guide manual (Laake et al. 1994) for further information.
Sample input
The left hand column illustrates an actual syntax command file that can be
adapted to meet your specific requirements. The right hand column is a brief
explanation of what the commands mean and why they were chosen. Unless
otherwise stated, the commands used in the two examples are not specific to
either line transects or point counts but can be used interchangably. It is
important that the correct punctuation is followed throughout the input file, as
the DISTANCE program will not recognise your commands otherwise.
Example one is taken from a line transect study of hornbills in Zambia
(the data set is not complete). In this example, data were collected in two
types of woodland (chipya and miombo) and several transects were conducted
in each. The habitat types have been allocated as Stratum and the transects
as Samples within these. This enables the calculation of a density estimate
for each transect and habitat type individually, as well as a pooled estimate to
be made across habitat types, if appropriate.
Bird Surveys 131
Options:
Title=HORNBILLS IN
ZAMBIAN WOODLAND;
* A title may be displayed.
Distance=Perp/Exact;
* Specifies that the distance data

entered is of the perpendicular and
exact format.
Select=All;
* All distance data is to be used in the

analysis.
Distance/Units=metres;
* Instructs that the units of

measurement for the distance data,
entered under Sample, are in metres.
Length/Units=Kilometres;

measurement for the length of
transect, entered under Effort, are in
kilometres.
Area/Units=Square
Kilometres;

measurement for the area of habitat,
specified under Stratum, is in square
kilometres.
End;
* Remember to include the End;

command after each section.
Data:
Stratum/Area=44.6/Label=
Chipya Habitat;
Sample/Effort=28.0/Label
=Transect1;
46,3,50,9,42,34,30,2,4,8;
=Transect2;
* Habitat data have been split into

two Strata and each assigned a label.
If the area of habitat under study is
known, a population estimate may be
calculated, using the density estimates
produced, by inserting the Area
command.
* Each transect has been assigned as
a Sample and labelled
correspondingly. The Effort, i.e. the
length of the transect multiplied by
the number of repeats, is also
indicated.
24,48,50,2,140,25,27,53,
15;
Sample/Effort=3.4/Label=
Transect3;
0,17,90,57,6;
Stratum/Area=43.5/Label=
Miombo Habitat;
=Transect4;
2,18,7,150,19,14,5,2,18,
8,0;
Sample/Effort=4.8/Label=
Transect5;
;*
* Note no observations were made on

transect 5 but it must still be included
in the analysis.
=Transect6;
70,25,10,11,15,17,15;
=Transect7;
3.5,20,17,12,10,3,80,13;
End;
Estimate:
* The Estimate section determines

which criteria you wish the density
estimate to be calculated by.
Density by stratum/
* This command instructs for an estimate

of density to be calculated for each habitat
type (stratum). To obtain an estimate of
density for each transect substitute with
sample. A pooled estimate of density will
automatically be provided by the default.
Weight by effort;
* The contribution made by each

transect to the density estimate for
Bird Surveys 133
each stratum, is to be weighted by the

length of that transect.
Detection All;
* If the study species has an equal

chance of being detected in each
habitat type (see ESW), then all
distances can be used together (giving
a larger sample size), when
calculating the detection distance.
End;
The second example is the basis for a VCP method study in cloud forest
habitat, Ecuador.
Options:
Type=Point;
* If point counts were used then this

must be specified as the default
setting is for line transects.
Object=Cluster;
* This command will allow estimates

for observations recorded as clusters.
Select=All;
Distance/Measure=Metres/r
truncate=0.20;
* rtruncate=0.20 instructs
DISTANCE to truncate the data at
the right hand side of the detection
curve. This value can be altered and
recommendations have been made in
section 3.5.2.
Dist/int=6,12,18,24,36;
* This command can be used to

instruct DISTANCE to allocate the
distances measured into certain

intervals. This is a group remedy to
overcome heaping (see section
3.4.2).
Area/Units=Square
Kilometres;
End;
Data:
Sample/Effort=11.2;
34, 3, 18, 6, 2.5, 2, 57,
2, 9, 3;
Sample/Effort=5.4;
* Distances should be entered

followed by the cluster size and
separated by a comma. Cluster sizes
are highlighted in bold.
0, 1, 6, 4;
End;
Estimate:
Density by sample;
Estimator/Key=Hazard;
Estimator/Key=Uniform;
Pick=AIC;
End;
* Detection is to be estimated by
sample, selecting between the two
models without estimating further
parameters. See section 3.5.2 for
guidelines on model selection.
* Instructs DISTANCE to choose the
model with the lowest AIC value.
Bird Surveys 135
Sample output
DISTANCE output begins with a summary of the input data including: the
number of samples, observations and strip width. Check here for any obvious
errors. This is followed by the choice of the best model, with appropriate
adjustments, to fit the data. After various test statistics, the output finishes
with a summary of the final density estimates. Below is a brief outline of the
major sections which will be useful to the interpretation of your density
estimates.
Density Estimation Results look here to see which model
and adjustments have been finally chosen by the programme, using the AIC
values (see below).
The chosen model is illustrated by a graph of a detection
probability curve and a CHI-SQ goodness of fitness
test is also performed. If a reliable chi-square test cannot be achieved,
DISTANCE will call for you to pool some of your data by hand. This may
mean adding data from separate transects or habitat types together to obtain
larger sample sizes within the group.
This section contains several important statistics. In particular check:
ESW Effective Strip Width is the distance beyond which as many
observations are missed as are included within the strip (EDR in point
counts). This value may differ between habitats and species, and can be used
to decide whether two sets of data are suitable for pooling by hand.
AIC Akaikes Information Criterion is used in model selection;
generally the model with the lowest value is chosen for fitting.
Chi-p Probability for goodness of fitness test.
2
Estimate
%CV
dF
95% Confidence Interval
48
48
48
.96070E-02
.16432
49.295
Stratum: HABITAT CHIPYA

Half-normal/Cosine
m
3.0000
AIC
519.12
Chi-p .29314
f(0)
p
ESW
13960E-01
.23877
71.632
18.75
18.75
18.75
The last section of output presents the final density estimates.
.20286E-01
.34697
104.09
Estimate
%CV
dF
95% Confidence Interval
4
5
2.5770
5.0403
10.572
15.649
34.25
35.36
23
27
8.18.76
13.118
32.488
53.654
22.55
23.59
25
31
5.4160
8.4354
13.555
21.801
Stratum: HABITAT CHIPYA

Half-normal/Cosine
DS
D
5.2197
7.9515
25.84
26.80
Stratum: HABITAT MIOMBO

Half-normal/Cosine
DS
D
16.310
26.529
Pooled Estimates
DS
8.5682
D
13.561
Key points are:
DS the density of clusters (if appropriate).

D the density of individual animals per unit of measurement specified.
95% confidence interval indicates that there is 95% probability that the
density estimate falls between the two values specified. These values are
useful for calculating maximum/minimum population sizes within specified
areas.
The website for DISTANCE includes software that is free to download. The
address is:
http://www.mbr.nbs.gov/software.html
Bird Surveys 137
The BP Conservation Programme

The BP Conservation Programme, organised by BirdLife International, the
British Petroleum Company plc (BP) and Fauna & Flora International (FFI),
aims to encourage long term conservation projects which address global
priorities at a local level. Each year the Programme gives out advice, training
and financial awards to teams of students all over the world building projects
which fulfil the following criteria:
address a conservation priority of global importance;

have a strong association with the country where the project will take
place (local people participating in all stages of the project);
the majority of the team must be university students (under or postgraduates in full- or part-time study).
These specific criteria aim to increase the long-term, sustainable conservation

achievements of a project by focusing the research objectives and building
vital links between personnel at all levels, from project team members and
local people to government staff.
Further information about this Programme is available on the web:
http://www.bp.com/conservation/
or from:
Programme Manager
BirdLife International/FFI
Wellbrook Court
Girton Road
Cambridge CB3 ONA
UK
Tel: +44 1223 277318
Fax: +44 1223 277200

Bird Surveys-Expedition Field Techniques.

Uploaded by

Copyright:

Available Formats

Bird Surveys-Expedition Field Techniques.

Uploaded by

Document Information

Original Description:

Copyright

Available Formats

Share this document

Share or Embed Document

Sharing Options

Did you find this document useful?

Is this content inappropriate?

Copyright:

Available Formats

Bird Surveys-Expedition Field Techniques.

Uploaded by

Copyright:

Available Formats

Expedition Field Techniques

Colin Bibby, Martin Jones and Stuart Marsden

Published by the Expedition Advisory Centre

Fax +44 (0)171 591 3031

Expedition Field Techniques

Section One: Why Count Birds?

Section Two: Study Design

Expedition Field Techniques

Section Three: Estimating bird densities using distance sampling

birds of prey, migrants, etc.)

Section Five: Assessment of sites: measurement of species richness

Expedition Field Techniques

6.3.1 Different approaches/survey designs

Section Seven: Maximising the impact of the work

Section Eight: References and further reading

Section Nine: Sample input and output

The BP Conservation Programme

2 Expedition Field Techniques

4 Expedition Field Techniques

WHY COUNT BIRDS?

6 Expedition Field Techniques

* These criteria do not categorise species alone but have to be met in

8 Expedition Field Techniques

1.3 Important Bird Areas

which a sites-based approach is appropriate. Sites in Europe (Grimmett and

globally threatened species

Population estimates are required to identify IBAs for globally threatened

10 Expedition Field Techniques

Determining boundaries of IBAs can be tricky. Sites should, as far as

been written to indicate some methods for measuring and describing

1.5 Choice of methods

what is the question?

12 Expedition Field Techniques

conservation organisations or universities. Government departments might

14 Expedition Field Techniques

unfinished job of limited value to anyone. As broad guidance, consider that

16 Expedition Field Techniques

exercise. Another example of bias is comparing results from a noisy

Figure 6. The relationship between

precision and accuracy.

c) Precise and inaccurate

b) Imprecise and accurate

d) Precise and accurate

18 Expedition Field Techniques

2.2 Study Design

20 Expedition Field Techniques

22 Expedition Field Techniques

information and prepare your own version of a guide. It is often worthwhile

24 Expedition Field Techniques

26 Expedition Field Techniques

Replication of your sampling is a way of increasing the reliability and

28 Expedition Field Techniques

initial siting of a path is also not likely to be random with respect to

Figure 9. Positioning point count sites or the beginning of transects.

a) Along paths or rivers

2.3 Pilot surveys and training

obtain as much local knowledge as possible on the distribution of key

30 Expedition Field Techniques

start tree species identification or classification (if appropriate);