Seminar Presentation on

LAND USE LAND COVER MODELLING AND ITS FUTURE PROJECTION

Presented by
Sahana V.
P15WR009

Submitted to:

WATER RESOURCES ENGINEERING

CIVIL ENGINEERING DEPARTMENT
S. V. NATIONAL INSTITUTE OF TECHNOLOGY-SURAT
GUJARAT (INDIA)
(YEAR 2016-17)
Guide
Shree G. D. Kale
Assistant Professor
1
 CONTENTS
Introduction
Descriptive LULC Models
Cellular Automata and Markov Chain Process
Procedure and Algorithms followed for LULC Modelling and Future Land Use
Prediction
Accuracy Assessment Of Land Use Change Models
Closure
References
2
 INTRODUCTION
Land cover indicates the physical land type such as forest or open water
whereas land use documents how people are using the land

Root cause for Land use and Land cover

Change

Anthropogenic activities

Natural process have minor

role in land use change

3
 Land Use Land Cover Models
helpful and
reproducible
tools
Land Use Land Cover
Important variable that decides

Land use complementing

making more
Land cover the existing
informed
mental
decisions Models
• Resource Planning capabilities

• Control Measures

Socio-economic, biophysical and proximate analysis of land

use change

causes that influence land use change can

be modelled 4
 Descriptive LULC Models

Linear or Logistic regression Agent based Modelling

Focuses on simulating the Models do not offer high Highly variable results are
functioning of the land use modelling capabilities. obtained -initial conditions and
Fail to address correlations be- the interaction rules of agents
system and the spatially tween independent variables. are arbitrarily defined

explicit simulation of land

use patterns
Decision Tree Modelling
Fractal Modelling
simple algorithm structure
satisfactory assessment of urban
Cellular Automata(CA) for spatial heterogeneity incorporated complexity is of utmost
into the model is of low degree
LULC modelling is chosen for importance for proper fractal
modelling.
the study. Why?
•Non-linearity of Land use classes
Artificial Neural Networks
•Iterative process leads to produce Modelling
fractal patterns in urban areas problem of overfitting the data
if there is a careless selection
•Spatial changes are pronouncedly taken of training data size.

care 5
 What is Cellular Automata?

Possess a regular infinite discrete lattice of cell in one or two

dimensions
• Possess an internal clock which represents a discrete time and
at each pre-defined temporal step, the cells are updated

Every cell has a finite set of possible states

• Transition rules are applied uniformly through time and space

Cell values of the local neighbourhood decides the new state of

the cells
6
 Definitions involved in CA
A d-dimensional CA can be specified as (𝑆, 𝑁, 𝑓)
S = State set Neighbourhoods
N = neighbourhood vector
𝑓 = 𝑆𝑛 → 𝑆, local update rule

Von Neumann, 3 × 3 Moore neighbourhood and 5 × 5

neighbourhood (Source: Verburg et al., 2004)

Central cell with concentric rings neighbourhood

(Source: Wijesekara, 2013)
7
 Markov Chain Process

Transition probability matrix

It is transition probability
with m states

[ ]
𝑃 𝑖𝑗 ( 𝑡 )= 𝑃 ( 𝑋 𝑡 =𝑎 𝑗| 𝑋 𝑡 −1=𝑎𝑖 ) 𝑝11 𝑝 12 … 𝑝1 𝑚
𝑝 21 ⋱ ⋱ ⋮
𝑃=
⋮ ⋱ ⋱ ⋮
depends only on and not on previous process 𝑝𝑚1 … … 𝑝 𝑚𝑚

𝑛𝑖𝑗
This is called memoryless property
′
𝑝 𝑖𝑗 = 𝑚

∑ 𝑛𝑖𝑗
𝑗 =1

𝑁𝑜 . 𝑜𝑓 𝑐𝑒𝑙𝑙𝑠 𝑡h𝑎𝑡 𝑐h𝑎𝑛𝑔𝑒𝑑 𝑓𝑟𝑜𝑚𝑠𝑡𝑎𝑡𝑒 𝑖𝑡𝑜 𝑗

𝑇𝑜𝑡𝑎𝑙𝑛𝑜 .𝑜𝑓 𝑐𝑒𝑙𝑙𝑠 𝑡h𝑎𝑡 𝑐h𝑎𝑛𝑔𝑒𝑑𝑡𝑜 𝑠𝑡𝑎𝑡𝑒 𝑗
elements in any row of P must sum to unity
8
Flowchart for LULC
modelling
and future projection

9
 Markov Chain Analysis Multi Criteria Evaluation
Best example
Landscape transition models can be expressed as
Choice between a set of waste disposal - the allocations of
land depends on multiple criteria such as distance from
residential and heritage lands, present land use practice,
𝐿𝑈 𝑡 +1= 𝐿𝑈 𝑡 × 𝑃 accessibility to roads
Suitability maps for each class of land use are
derived from this MCE technique
Where P is
𝑛 𝑚
𝑆=∑ 𝑊 𝑖 𝐶 𝑖 × ∏ 𝑟 𝑗

[ ]
Constraint
𝑝11 𝑝 12 … 𝑝1 𝑚 𝑖=1 𝑗=1 0 0r 1
Suitability Score
𝑃 = 𝑝 21 ⋱ ⋱ ⋮
0 to 1 Criterion score of factor
⋮ ⋱ ⋱ ⋮
0 to 1
𝑝𝑚1 … … 𝑝 𝑚𝑚 Weight assigned to factor
1 to 100%

10
 Eg. Suitability map for water class

i=1 i=2 i=3

: Criteria for suitability Bathymetric zones map The Water class Slope map derived from
Bathymetric zones map rated from 0.1 to 1 as very probability map of Dead DEM map
rated shallow to very deep Sea derived from rated from 0.1 to 1 as the
water respectively Markov process. highest slope to the
lowest slope respectively
: Weight for a criteria i 34% 33% 33%

: Restriction Area boundaries

11
 CA-Markov Analysis

Future projection of LULC in IDRISI

𝐿𝑈 𝑖 , 𝑗 (𝑡 +1)= 𝑓 ( 𝐿𝑈 𝑖, 𝑗 ( 𝑡 ) . 𝑆𝑖 , 𝑗 (𝑡 ) . 𝑃 𝑥 , 𝑦 ,𝑖 , 𝑗 (𝑡 ) . 𝑁 𝑖, 𝑗 (𝑡 ) )
neighbourhood index
state of cell Moore or Von Neumann
at t+1
state of cell probability of cell
at t to change from x to y ∑ 𝑁 𝑖 , 𝑗 (𝑡 )
𝑡
𝑁 𝑖 , 𝑗 (𝑡 )=
¿ 𝑜𝑓 𝑎𝑑𝑗𝑎𝑐𝑒𝑛𝑡 𝑐𝑒𝑙𝑙𝑠

suitability indexes of cell

for that state

12
 ACCURACY ASSESSMENT OF LAND USE CHANGE MODELS
𝑐

•In discrete multivariate analysis, a statistic called

Observed fraction of agreement
𝑝 0= ∑ ¿ ¿
𝑖 =1
Kappa coefficient of agreement is used 𝑐
Expected fraction of agreement 𝑝 𝑒=∑ ¿ ¿
•This coefficient, as the expected agreement can be 𝑖 =1
interpreted in terms of map comparisons provided the 𝑐

given classes are allocated randomly all over the map.

Maximum fraction of agreement 𝑝 𝑚𝑎𝑥 = ∑ ¿ ¿
𝑖=1

Map S categories
Total
1 2 ... c map A
Generic form of contingency table
Map A categories

1 p(a=1∧s=1) p(a=1∧s=2) ... p(a=1∧s = c) p(a=1)

2 p(a=2∧s=1) p(a=2∧s=2) ... p(a=2∧s = c) p(a=2)
fraction of cells that have ... ... ... ... ... ...
land use i in map A and
p(a = p(a = p(a =
land use j in map S. c p(a = c∧s = c)
c∧s=1) c∧s=2) c)
13
Total map S p(s=1) p(s=2) ... p(s = c) 1
 In context of map comparison • Agreement between the allocation of land uses is a not a
meaningful measure

• Vliet et l., 2011 achieved this by considering distribution of class

𝑝0 − 𝑝𝑒 transitions that can be interpreted as conditional probabilities
𝐾𝑎𝑝𝑝𝑎= -1 to 1
1− 𝑝 𝑒 using the original map

𝑝 𝑚𝑎𝑥 − 𝑝𝑒 • chance of finding a certain class at a location will depend on the

𝑘h𝑖𝑠𝑡𝑜𝑔𝑟𝑎𝑚= 0 to 1
1− 𝑝 𝑒 class that was originally there

𝑘𝑙𝑜𝑐𝑎𝑡𝑖𝑜𝑛=
𝑝0 − 𝑝𝑒
𝑝 𝑚𝑎𝑥 −𝑝 𝑒
-1 to 1
𝑝¿¿
𝐾𝑎𝑝𝑝𝑎=𝑘h𝑖𝑠𝑡𝑜𝑔𝑟𝑎𝑚 ×𝑘𝑙𝑜𝑐𝑎𝑡𝑖𝑜𝑛
𝑝¿¿
fraction of cells that changed from land use j in the original map
to land use i in the simulated land use map
14
 Coefficient of Kappa for simulation and its components
Coefficient of agreement
between the simulated land use
𝐾 𝑆𝑖𝑚𝑢𝑙𝑎𝑡𝑖𝑜𝑛=𝑝 0 −𝑝 ¿¿ ¿ -1 to 1
transitions and actual land use
transitions.

agreement in the quantity of

𝐾 𝑇𝑟𝑎𝑛𝑠𝑖𝑡𝑖𝑜𝑛=𝑝 ¿¿ ¿ land use transition.
0 to 1

expresses the degree to which

𝐾 𝑇𝑟𝑎𝑠𝑙𝑜𝑐=𝑝 0 − 𝑝¿ ¿ ¿ the transitions agree in their
-1 to 1 allocations
15
 CLOSURE
• The satellite derived LULC maps in conjunction with biophysical and socio-economic data can
be used to estimate the past and present changes
• The CA-MC process aids in prediction of land use change which takes care of the spatial and
temporal allocation of a particular land use taking into account the transition probabilities
• The factors and constraints should be carefully selected for different areas.
• Suitability maps based on MCE technique aids in the change analysis of land use and
projection of land use.
• The validation and sensitivity analysis of the projected LULC maps making use of Kappa
simulation and Kappa coefficients which gives a comparison of simulated and original maps is
explained

• Future LULC maps are in great demand for

o Hydrologic modelling , in which future runoff estimation of a river basin or catchments are
invariably needed for resource planning, flood management and construction of hydraulic
structures
o The growth modelling of an expanding city or urban area are analysed 16
 References
Costanza, R., and Ruth, M. (1998). Using dynamic modeling to scope environmental problems and build consensus. Environmental management,22(2), 183-195.
 
Couclelis, H. (1997). From cellular automata to urban models: new principles for model development and implementation. Environment and planning B: Planning and
design, 24(2), 165-174.
 
Crooks, A. T. (2006).  Exploring cities using agent-based models and GIS. Working/discussion paper. <http://eprints.ucl.ac.uk/3341/> (Retrieved 06.11.12).
 
El-Hallaq, M. A., andHabboub, M. O. (2015). Using Cellular Automata-Markov Analysis and Multi Criteria Evaluation for Predicting the Shape of the Dead
Sea. Advances in Remote Sensing, 4(01), 83.
 
Haan, C. T. (2002). Statistical Methods in Hydrology.Iowa State University Press, Ames, IA.
 
Hagen, A. (2002, April). Multi-method assessment of map similarity. InProceedings of the fifth AGILE conference on geographic information science, Palma,
Spain (pp. 171-182).
 
Hasbani, J. G. (2008, January). Semi-automated calibration of a cellular automata model to simulate land-use changes in the Calgary region. InMasters Abstracts
International (Vol. 46, No. 06).
 
Kari, J. (2005). Theory of cellular automata: A survey. Theoretical Computer Science, 334(1), 3-33.
 
Lambin, E. F., Turner, B. L., Geist, H. J., Agbola, S. B., Angelsen, A., Bruce, J. W., ... and George, P. (2001). The causes of land-use and land-cover change: moving
beyond the myths. Global environmental change, 11(4), 261-269.
 
Mandelbrot, B. B. (1983). The Fractal Geometry of Nature (Vol. 173). Macmillan.
 
17
McDonald, R. I., and Urban, D. L. (2006). Spatially varying rules of landscape change: lessons from a case study. Landscape and Urban Planning, 74(1), 7-20.
 …References
Power, C., Simms, A., and White, R. (2001). Hierarchical fuzzy pattern matching for the regional comparison of land use maps. International Journal of Geographical
Information Science, 15(1), 77-100.
 
Rumelhart, D. E., Hinton, G. E., and Williams, R. J. (1985). Learning internal representations by error propagation (No. ICS-8506). California University San Diego
La Jolla Inst for Cognitive Science.
 
Srivastava, P. K., Mukherjee, S., and Gupta, M. (2010). Impact of urbanization on land use/land cover change using remote sensing and GIS: a case study.International
Journal of Ecological Economics and Statistics™, 18(S10), 106-117.
 
Triantakonstantis, D., and Mountrakis, G. (2012). Urban growth prediction: A review of computational models and human perceptions.Earth and Environmental
Sciences, 4(6), 555-587.
 
Van Ittersum, M. K., Rabbinge, R., and Van Latesteijn, H. C. (1998). Exploratory land use studies and their role in strategic policy making.Agricultural Systems, 58(3),
309-330.
 
van Vliet, J., Bregt, A. K., and Hagen-Zanker, A. (2011). Revisiting Kappa to account for change in the accuracy assessment of land-use change
models.EcologicalModelling, 222(8), 1367-1375.
 
Verburg, P. H., Schot, P. P., Dijst, M. J., andVeldkamp, A. (2004). Land use change modelling: current practice and research priorities. GeoJournal, 61(4), 309-324.
 
Wijesekara, G. N. (2013). An integrated modeling system to simulate the impact of land-use changes on hydrological processes in the Elbow River watershed in
Southern Alberta (Doctoral dissertation, University of Calgary).
 

  18
Thank You

19