Yang Manlun

Suitability Analysis of
Urban Green Space

System
Based on GIS
Yang Manlun
September, 2003
Suitability Analysis of Urban Green Space System
Based on GIS
by
Yang Manlun
Thesis submitted to the International Institute for Geo-information Science and Earth Observation in
partial fulfilment of the requirements for the degree of Master of Science in Geo-information Science
and Earth Observation with specialisation in Urban Planning and Management
Thesis Assessment Board
Prof. Dr. Willem v.d. Toorn (Chairman)

Drs. Fred Toppen (external examiner, University Utrecht)
Mrs. Du Ningrui, MSc. (SUS supervisor)
Drs. P. Hofstee (First ITC supervisor)
Ir. M. Brussel (Second ITC supervisor)
INTERNATIONAL INSTITUTE FOR GEO-INFORMATION SCIENCE AND EARTH OBSERVATION

ENSCHEDE, THE NETHERLANDS
Disclaimer
This document describes work undertaken as part of a programme of study at the International
Institute for Geo-information Science and Earth Observation. All views and opinions expressed
therein remain the sole responsibility of the author, and do not necessarily represent those of
the institute.
Acknowledgements
After six months study, it is the time of harvest. Looking back on the study life in the Netherlands, it
is composed of challenge, excitement and happiness. Not only the technical knowledge we learnt here
but also the different cultures and the friendly people. It is really a precious experience that will be
remembered forever in my life.
Firstly, I owe special thanks to my first supervisor, Drs. Paul Hofstee for his important guidance and
helpful comments from the beginning to the end of this research. I am also very grateful to my second
supervisor, Ir. Mark Brussel. He helped me with the analysis and writing in the final stage.
I would like to thank Mrs. Xiao Yinghui for her guidance to process the data and Mrs. Du Ningrui for
her comments to improve the thesis. I owe many thanks to my Chinese supervisors, Prof. Xu
Zhaozhong and Mr. Zhang Jun of Wuhan University, for their guidance and ideas to write the pro-
posal. It was a pleasure to have the opportunity to meet some Chinese PhDs, Zhan Qingming, Huang
Zhengdong, Cheng Jianquan, Zheng Ding, Tang Xingming, and Zhu Sicai. They provided many in-
spirits for my study.
Special thanks go to Chen Wenbo and Liu Kang, who have been always helpful with their rich experi-
ence and knowledge. Every time’s discussion with them made my thesis improve. I also want to thank
them for their patience of teaching me the GIS technology and correcting my English writing.
Sincere thanks go to one of my best friends, Feng Qiaobing. She greatly helped me do the fieldwork,
collect the data and process the data. Her sincere supports and inspirits made me study here at ease.
Many thanks go to the UPLA2 2003 group of my Chinese MSc classmates, particularly to Wang Yu-
jian for his help with my living. It was a pleasure to study and share the joys with them. I also won’t
forget the kindly help from my Chinese classmates, Cao Chunxia and Zhou Qinghua of Wuhan Uni-
versity, for helping me to do the literature review.
Last but not least, I feel deeply grateful to my parents and brothers, for their sincere love, understand-
ing and support to me. And I want to express my heartfelt thanks to all people who ever helped me.
Abstract
Suitability analysis of green space system is designed to identify and measure the suitability of poten-
tial sites for green space system development. Such analysis can be regarded as a relatively difficult
task partially due to large number of factors and large volume of data that may be required for the de-
termination. The purpose of this research is to develop an approach of GIS-based suitability analysis
to identify suitable sites for urban green space system development. This approach identifies seven
major steps involved in the suitability analysis, which include selecting, scoring, weighting suitability
factors, generating suitability scenarios using GIS, ranking suitability scenarios, making sensitivity
analysis, and output evaluation.
Selecting suitability factors is mainly based on stakeholder analysis and desirable environmental quality.
Four groups including urban planners, environmentalists, local residents and local government officials are
involved in the stakeholder analysis. The desirable environmental quality is proposed from two aspects:
existing situation and greening indices. As such, seven suitability factors including air quality, landscape
quality, surface water quality, historic culture value, water system influence, noise influence, and existing
land use, will be selected to carry out the GIS-based suitability analysis. These seven factors are set as ‘high
suitability’, ‘moderate suitability’, and ‘no suitability’. Ratio values are applied in scoring these three
classes within the suitability factors, and the establishment of certainty factor is introduced to improve the
traditional GIS-based suitability analysis model. After that, three weighting methods including statistic in-
tegration, hierarchic analysis of nine-degree and hierarchic analysis of three-degree are used to define three
sets of weighting systems.
All the above data are integrated into a raster-based GIS software and spatial analysis is performed
using an overlay technique to generate six suitability scenarios. Then weighted summation and electre
method are used to make a ranking among these six suitability scenarios. Sensitivity analysis is car-
ried out to test the validity of scores, weights used and the ranking of the scenarios. As such, the best
suitability scenario comes out and it needs to be evaluated by comparing it with the urban master plan,
with the aim of finding the commons and differences between them and then to validate the proposed
approach.
Suitability analysis is a powerful tool for green space system planning. Continued development and
refinement of suitability analysis, particularly with GIS technology, can enable urban planners to help
local government officials and local residents to create a suitable green space system in the urban en-
vironment. In order to advance the art of the suitability analysis, it is important that not only the suit-
ability output is replicable within a study area, but also the approach is transferable, or at least adapt-
able in other places. This research provides an example of such transferability. In general, GIS is a
toolbox capable of providing support for spatial problem-solving and decision-making, and it should
be integrated with the decision support system (DSS) to make the suitability analysis in a more sys-
tematic way.
Table of Contents
List of Tables
List of Figures
List of Maps
List of Formulas
1. Introduction..................................................................................................1
1.1. Background ................................................................................................................................1
1.2. Problem statement ......................................................................................................................2
1.3. Research objective .....................................................................................................................3
1.3.1. Main objective...................................................................................................................3
1.3.2. Specific objectives ............................................................................................................3
1.4. Research questions .....................................................................................................................3
1.5. Workflow ...................................................................................................................................5
1.6. Structure of the thesis.................................................................................................................6
2. Definition and conceptions of urban green space system ........................7
2.1. Definition of urban green space system .....................................................................................7
2.2. Classification of urban green space system ...............................................................................8
2.2.1 Classification in foreign countries .....................................................................................8
2.2.2. Classification in China .....................................................................................................10
2.3. Comprehensive benefits of urban green spaces .......................................................................13
2.3.1. Ecological benefits ...........................................................................................................13
1. Clean air ..............................................................................................................................13
2. Adjust and improve urban climate ......................................................................................14
3. Prevent and reduce hazard...................................................................................................15
4. Eliminate noise....................................................................................................................15
2.3.2. Social benefits ..................................................................................................................15
1. Recreation ...........................................................................................................................15
2. Landscape aesthetics ...........................................................................................................16
3. Adjust psychology...............................................................................................................17
4. Education.............................................................................................................................18
2.3.3. Economic benefits ...........................................................................................................19
3. Methodology ...............................................................................................21
3.1. Definition of suitability analysis ..............................................................................................21
3.2. Suitability analysis methods.....................................................................................................21
3.2.1. Direct overlay..................................................................................................................21
3.2.2. Weighted score................................................................................................................22
3.2.3. Ecological factors combination.......................................................................................23
3.3. GIS application in suitability analysis .....................................................................................23
3.4. GIS-based traditional suitability analysis model (TSAM) and its improvement ....................24
3.4.1. Traditional suitability analysis model (TSAM) ...............................................................24
1. TSAM procedure......................................................................................................................24
2. Example....................................................................................................................................25
3.4.2. Improved traditional suitability analysis model (ITSAM)..............................................26
3.4.3. Summary .........................................................................................................................29
3.5. Weighting methods ..................................................................................................................29
3.5.1. Statistic integration .........................................................................................................29
3.5.2. Hierarchic analysis of nine-degree..................................................................................31
3.5.3. Hierarchic analysis of three-degree.................................................................................32
3.6. Evaluation methods for ranking ...............................................................................................34
3.6.1. Weighted summation.......................................................................................................36
3.6.2. Electre method.................................................................................................................36
3.6.3. Summary .........................................................................................................................37
3.7. Sensitivity analysis...................................................................................................................37
3.7.1. Uncertainty on scores......................................................................................................38
1. Overall uncertainty of the scores..............................................................................................38
2. Uncertainty of one score ..........................................................................................................38
3.7.2. Sensitivity on weights .....................................................................................................38
1. Changes in all weights..............................................................................................................39
2. Different sets of weights ..........................................................................................................39
3.7.3. Summary .........................................................................................................................39
3.8. Methodology flow chart ...........................................................................................................40
4. Case study in Dongguan ............................................................................42
4.1. Study area: Dongguan municipality.........................................................................................42
4.1.1. Location...........................................................................................................................42
4.1.2. Physical characteristics ...................................................................................................43
4.1.3. Social-economic characteristics ......................................................................................43
4.2. Current green space system analysis in Dongguan ..................................................................44
4.2.1. Existing situation of the green space system...................................................................44
4.2.2. Problems existing in the green space system ..................................................................44
1. Public green space and suburban forestry................................................................................44
2. Residential green space and departmental (work unit) affiliated green space.........................45
3. Road green space......................................................................................................................45
4. Productive and defensive green space......................................................................................46
5. Landscape forestry land ...........................................................................................................46
4.2.3. Greening indices..............................................................................................................46
1. Definitions of three greening indices .......................................................................................46
2. Functions of three greening indices .........................................................................................47
4.2.4. Desirable environmental quality in Dongguan................................................................48
4.3. Suitability analysis of green space system based on GIS.........................................................48
4.3.1. Stakeholder analysis for suitability .................................................................................48
1. Urban planners .........................................................................................................................48
2. Environmentalists.....................................................................................................................49
3. Local residents .........................................................................................................................49
4. Local government officials.......................................................................................................49
4.3.2. Selecting suitability factors.............................................................................................50
1. Available data...........................................................................................................................50
2. Data pre-processing..................................................................................................................51
4.3.3. Scoring ............................................................................................................................52
1. Scores of suitability factors......................................................................................................52
2. Certainty factor.........................................................................................................................55
4.3.4. Weighting ........................................................................................................................58
1. Calculating weights by statistic integration .............................................................................58
2. Calculating weights by hierarchic analysis of nine-degree......................................................59
3. Calculating weights by hierarchic analysis of three-degree.....................................................60
4. Weighting results for suitability scenarios...............................................................................61
4.3.5. Suitability scenario..........................................................................................................62
4.4. Multi-criteria analysis for ranking ...........................................................................................65
4.4.1. Effects table.....................................................................................................................65
4.4.2. Standardization................................................................................................................66
1. Goal standardization for ‘high suitability’ ...............................................................................66
2. Interval standardization for ‘moderate suitability’...................................................................67
3. Maximum standardization for ‘no suitability’ .........................................................................67
4.4.3. Weight .............................................................................................................................68
4.4.4. Ranking ...........................................................................................................................69
1. Weighted summation................................................................................................................69
2. Electre method..........................................................................................................................71
4.5. Sensitivity analysis...................................................................................................................75
4.5.1. Uncertainty of one score .................................................................................................76
4.5.2. Overall uncertainty of the weights ..................................................................................76
4.5.3. Changes in all weights (rank reversal of two alternatives) .............................................77
4.6. Comparison ..............................................................................................................................78
4.6.1. Commons.........................................................................................................................80
4.6.2. Differences ......................................................................................................................81
5. Conclusion and recommendation ...............................................................83
5.1. Conclusion................................................................................................................................83
5.2. Recommendation......................................................................................................................84
List of Tables
Table 2.1 Definitions of green open space.............................................................................................8
Table 2.2 Classification of parks in America.........................................................................................9
Table 2.3 Classification of urban green space system in Japan .............................................................9
Table 2.4 Classification of urban green space system in China...........................................................11
Table 3.1 Factors and weights in the traditional suitability analysis ...................................................25
Table 3.2 Investigating table of importance order ...............................................................................30
Table 3.3 Information table of statistic induction (%).........................................................................30
Table 3.4 Factors weights by statistic integration................................................................................31
Table 3.5 Importance comparison of nine-degree................................................................................31
Table 3.6 Structural judgment matrix of nine-degree ..........................................................................31
Table 3.7 Factors weights by hierarchic analysis of nine-degree ........................................................32
Table 3.8 Comparison matrix of three-degree .....................................................................................33
Table 3.9 Structural judgment matrix of three-degree .........................................................................33
Table 3.10 Factors weights by hierarchic analysis of three-degree .....................................................34
Table 3.11 Overview of evaluation methods for ranking.....................................................................35
Table 4.1 Suitability classes and scores...............................................................................................53
Table 4.2 Investigating information by statistic induction ..................................................................58
Table 4.3 Structural judgment matrix of suitability factors by nine-degree ........................................59
Table 4.4 Comparison matrix of suitability factors by three-degree ...................................................61
Table 4.5 Structural judgment matrix of suitability factors by three-degree .......................................61
Table 4.6 Weighting results for suitability scenarios...........................................................................62
Table 4.7 Standardized effects table ....................................................................................................72
List of Figures
Figure 1.1 Research workflow ...............................................................................................................5
Figure 2.1 Developmental skeleton of green space system ...................................................................7
Figure 2.2 Hierarchical requirement theory (Abraham H. Maslow)....................................................18
Figure 2.3 Circular ring for education function of green spaces .........................................................19
Figure 3.1 TSAM .................................................................................................................................26
Figure 3.2 ITSAM ................................................................................................................................28
Figure 3.3 Methodology flow chart .....................................................................................................41
Figure 4.1 Data pre-processing ............................................................................................................51
Figure 4.2 Effects table ........................................................................................................................65
Figure 4.3 Standardization for ‘high suitability’..................................................................................66
Figure 4.4 Standardization for ‘moderate suitability’..........................................................................67
Figure 4.5 Standardization for ‘no suitability’.....................................................................................68
Figure 4.6 Standardizations and weights .............................................................................................69
Figure 4.7 Ranking results I by weighted summation..........................................................................69
Figure 4.8 Ranking results II by weighted summation.........................................................................70
Figure 4.9 Ranking results III by weighted summation (Scatter diagram) ..........................................71
Figure 4.10 Concordance table ............................................................................................................72
Figure 4.11 Discordance table..............................................................................................................73
Figure 4.12 Strong graph (0: no ranking, 1: a ranking)........................................................................74
Figure 4.13 Weak graph (0: no ranking, 1: a ranking).........................................................................74
Figure 4.14 Ranking results by electre method....................................................................................75
Figure 4.15 Sensitivity of the ranking for changes in one score..........................................................76
Figure 4.16 Uncertainty analysis on the weights (50%) ......................................................................77
Figure 4.17 Weight combination by rank reversal between ‘scenario 1’ and ‘scenario 2’..................78
List of Maps
Map 4.1 The location of Dongguan municipality (study area)…………………………………..……42
Map 4.2 Air (air quality)………………………………………………………………………..……..54
Map 4.3 Lscape (landscape quality)…………………………………………………….……………..54
Map 4.4 Swater (surface water quality)……………………………………………………………….54
Map 4.5 History (historic culture value)………………………………………………………………54
Map 4.6 Noise (noise influence)………………………………………………………………………55
Map 4.7 Luse (existing land use)……………………………………………………………………...55
Map 4.8 Wsystem (water system influence)…………………………………………………………..55
Map 4.9 Cerlscape (certainty factors for landscape quality)………………………………………….56
Map 4.10 Cerhistory (certainty factors for historic culture value)……………………………………56
Map 4.11 Cerluse (certainty factors for existing land use)……………………………………………56
Map 4.12 Cerwsystem (certainty factors for water system influence)………………………………...56
Map 4.13 Clscape (composite certainty factors for landscape quality)……………………………….57
Map 4.14 Chistory (composite certainty factors for historic culture value)…………………………..57
Map 4.15 Cluse (composite certainty factors for existing land use)…………………………………..57
Map 4.16 Cwsystem (composite certainty factors for water system influence)………………………57
Map 4.17 Draft suitability scenario 1………………………………………………………………….63
Map 4.18 Final suitability scenario 1………………………………………………………………….64
Map 4.24 Master plan of Dongguan municipality (2000-2015)………………………………………79
Map 4.25 Comparison map……………………………………………………………………………79
List of Formulas
Formula 3.1 Equal-weight summation .................................................................................................22
Formula 3.2 Weighted score ................................................................................................................22
Formula 3.3 Certainty factor function..................................................................................................27
Formula 3.4 Composite certainty factor...............................................................................................28
Formula 3.5 Statistic integration ..........................................................................................................30
Formula 3.6 Quantitative comparison of three-degree.........................................................................33
Formula 3.7 Hierarchic analysis of three-degree .................................................................................33
Formula 3.8 Weighted summation .......................................................................................................36
Formula 3.9 Concordance index ..........................................................................................................37
Formula 3.10 Discordance index .........................................................................................................37
Formula 4.1 Goal standardization……………………………………………………………………..66
Formula 4.2 Interval standardization………………………………………………………………….67
Formula 4.3 Maximum standardization……………………………………………………………….68
Formula 4.4 Expected value method…………………………………………………………………..68
SUITABILITY ANALYSIS OF URBAN GREEN SPACE SYSTEM BASED ON GIS
1. Introduction
1.1. Background
Land suitability analysis is the process of determining the fitness of a given tract of land for a defined
use (Steiner, McSherry et al. 2000). In other words, it is the process to determine whether the land
resource is suitable for some specific uses and to determine the suitability level. In order to determine
the most desirable direction for future development, the suitability for various land uses should be
carefully studied with the aim of directing growth to the most appropriate sites. Establishing appropri-
ate suitability factors is the construction of suitability analysis.
Initially, suitability analysis was developed as a method for planners to connect spatially independent
factors within the environment and, consequently to provide a more unitary view of their interactions.
Suitability analysis techniques integrate three factors of an area: location, development activities, and
biophysical/environmental processes (Miller, Collins et al. 1998). These techniques can make plan-
ners, landscape architects and local decision-makers analyse factors interactions in various ways.
Moreover, such suitability analysis enables elected officials and land managers to make decisions and
establish policies in terms of the specific landuses.
Even though suitability analysis is a well-known tool among planners, landscape architects and local
decision-makers, there are relatively few examples where a process used in one place has been trans-
ferred or adapted in another place (the few examples include the work of McHarg, 1969 and Lyle,
1985). Applications of suitability analysis can be found in many fields, such as site selection for crop-
land (natural resource management field), flooding control, sustainable development (environment
management field), etc. This method covers broad topics and develops continuously. However, spe-
cific applications on the green space system cannot be found very often. This research provides such
an example that uses seven factors to carry out the suitability analysis of urban green space system, as
will be critically explained in Chapter 4.
Since suitability analysis came into being, there have been many analytic methods that primarily in-
clude the method of sieve mapping, landscape unit method, grey tone method (map overlay) and com-
puter method (GIS). The method of sieve mapping is to use a series of ‘sieves’ (factors) to exclude
those areas that are not suitable for the specific landuse. Once passing all the ‘sieves’, it is easy to
eliminate all the assumed unsuitable areas, and what is left is suitable for some specific uses. The
landscape unit method is absolutely different from sieve mapping. First it needs to classify landscape
units according to a set of geographic characteristics, the land’s potentials and limitations are then
identified in each landscape unit. Finally the suitability analysis is finished after all the landscape
units are identified. Grey tone, also named map overlay, is created by professor McHarg (1996). This
American landscape architect has systematically expatiated on such method in his book Design With
1
Nature. Grey tone wants to make use of gradual colours to represent the suitability levels in the same
scale, and overlay all the single factor maps in a certain order. As such those supposed useful areas
would be displayed after the above process.
Grey tone method has made some excellent effects in North America, even all over the world. But it
also has some disadvantages: (1) It neglects the relative influence among the factors; namely, it as-
sumes that each factor is independent. (2) If large number of factors must be involved, it is a time-
consuming task to do the analysis by manual operations. (3) Worse is that grey tone method cannot
carry out arithmetic operations. However, computer methods were developed to solve these problems,
particularly the analytic method depending on GIS. The GIS technique can transfer the suitability
level into numerical value, and assign the weight to each factor according to their relative importance.
So finally we can achieve the composite suitability levels by summing up the multiplication.
The limitation of GIS-based computer method is that it needs a complicate expert system, which can
precisely select, assess the suitability factors and set up a weighting system. This is the most impor-
tant and difficult step in the suitability analysis. In general, GIS-based computer method can overcome
those difficulties that other methods can’t. It enables landscape architects and urban planners to use
and to process more information, to plan more complicate landuses, and then to push the suitability
analysis method to a new stage.
1.2. Problem statement

City is a multiplex ecological system made up of social, economic and natural these three sub-systems
(Huang and Chen 2002). Green space system is the foundation of the natural system. It is also the
principal part of the natural productivity in the urban structure. A suitable green space system can play
an effective role in cleaning air, adjusting climate, eliminating noise, beautifying surroundings, etc. It
is dispensable for constructing a high quality human settlement and a high standard ecocity.
A number of studies proved that increasing population and enhancing urbanization processes are con-
verting more and more soft green spaces into impermeable hard concrete surface. Particularly in a de-
veloping country, this trend is more serious (Shi 2002). China is a large country with almost 1.3 bil-
lion population in East Asia. With the fast economy growth in the past two decades, China is facing a
rapid urbanization, especially due to the rural-urban migration. The growing urban population wishes
a better living environment, and puts an enormous pressure on the demand for green spaces. At the
same time, rapid economy growth has resulted in the loss of valuable land resources. This does not
only destroy sustainable economy and human settlement, but also lead to environmental degradation
and reduction of green spaces.
In Dongguan municipality, some green spaces are being converted to other land uses every year. This
has caused some serious environmental consequences: increased soil temperature, local climate
change, instability in hydrological regime, and the loss of important species, all of which ultimately
have negative effects on the ecological environment and human settlement. In order to reduce such
harm, the Dongguan government has taken some activities to increase green spaces in the urban areas
such as ‘Greening Dongguan’, ‘Horticultural city with water and mountains’, ‘Sustainable Dong-
2
guan’, with the aim of improving the environmental conditions. It is generally believed that such ac-
tivities can bring more green spaces and make the integration of trees, parks, lawns, etc., as an ele-
ment of urban landscape. However, they are affected by many factors including natural conditions,
social-economic conditions, technical factors and so on. The result is that these activities cannot play
a good ecological function to the urban environment.
In this research, an approach that integrates suitability analysis with geographic information system
(GIS) technology will be developed and implemented to identify suitable sites for the urban green
space system development, in order to play a good ecological role and create an elegant landscape in
the study area of Gongguan municipality. Now the GIS-based traditional suitability analysis model is
not very precise for some specific factors analyses. It can’t meet the needs of new ecological planning.
Therefore, an approach to establish certainty factors is introduced to improve this GIS-based tradi-
tional suitability analysis model. After that, some suitability scenarios are generated and a ranking is
made among them. Sensitivity analysis is used to test the validity of this ranking to find the best suit-
ability scenario. Finally, the research compares this best suitability scenario with the urban master
plan and analyses their commons and differences.
1.3. Research objective
1.3.1. Main objective

The main objective of this research is to develop an approach of GIS-based suitability analysis to
identify suitable sites for urban green space system development.
1.3.2. Specific objectives

1. To understand the definition and conceptions of the urban green space system.
2. To analyse the strengths and weaknesses of current suitability analysis methods.
3. To evaluate the GIS-based traditional suitability analysis model.
4. To generate suitability scenarios of the urban green space system by integrating suitability
analysis with geographic information system (GIS) technology.
5. To carry out the ranking and sensitivity analysis to find the best suitability scenario.
6. To compare the best suitability scenario with the urban master plan.
1.4. Research questions

To realize the above stated objectives, the following research questions shall be answered:
1. Understand the definition and conceptions of the urban green space system.
• What is the urban green space system?
• What are the classifications of the urban green space system and what are the com-
prehensive benefits of the urban green spaces?
3
2. Analyse the strengths and weaknesses of current suitability analysis methods.

• What are the current suitability analysis methods and their strengths and weaknesses?
3. Evaluate the GIS-based traditional suitability analysis model.

• What is the GIS-based traditional suitability analysis model? What are the strengths
and weaknesses?
• Which method can be used to improve the GIS-based traditional suitability analysis
model?
4. Generate suitability scenarios of the urban green space system by integrating suitability
analysis with geographic information system (GIS) technology.
• How to select factors for the suitability analysis and how to determine their weights
and certainty factors in the study area?
• How to overlay all single factor maps to generate the suitability scenarios of the urban
green space system based on GIS?
5. Carry out the ranking and sensitivity analysis to find the best suitability scenario.
• What are the evaluation methods for ranking of the suitability scenarios?
• What is the sensitivity analysis and how to use it to test the validity of the ranking?
6. Compare the best suitability scenario with the urban master plan.
• What are the commons and differences between the best suitability scenario and the
urban master plan? What are the reasons?
4
1.5. Workflow
Literature - Urban green space system

review - Suitability analysis
Selecting 7 factors for

suitability analysis
Defining 3 sets of Using ILWIS to do the Getting alternatives

weighting systems suitability analysis (6 suitability scenarios)
Using Definite to make

the ranking and
Need to: sensitivity analysis
- Quantifying each factor class to a value (Scoring)
- Calculating each factor's weight (Weighting)
- Getting composite scores
- Classifying suitability
Comparison
Figure 1.1 Research workflow
The workflow above summarizes the main idea of this research. First, the literature review will focus
on the urban green space system and suitability analysis, such as their definitions, study methods, de-
velopment processes, etc. Then after a combination of the stakeholder analysis, published literature,
and fieldwork, this research will select seven factors for the suitability analysis, including air quality,
landscape quality, surface water quality, historic culture value, water system influence, noise
influence and existing land use (see section 4.3.2). Nowadays there are many methods used to calcu-
late weights for the suitability factors. This research wants to use three typical and efficient methods
to define three sets of weighting systems. These three weighting methods are statistic integration,
hierarchic analysis of nine-degree and hierarchic analysis of three-degree (see section 3.5). Af-
terwards, ILWIS, a GIS software is used to carry out the suitability analysis. And then six suitability
scenarios will be generated according to different sets of scores, weights and certainty factors (see
section 3.4.2). Each suitability scenario can be regarded as an alternative, so a ranking will be carried
out among these six alternatives (scenarios), and sensitivity analysis is carried out by Definite (a mul-
tiobjective decision support system software) to test the validity of scores, weights used and the rank-
ing of alternatives (see section 4.5). Finally, a best suitability scenario will come out and we can com-
pare this best scenario with the urban master plan, in order to find the differences and commons be-
tween them.
5
1.6. Structure of the thesis

This thesis focuses on the suitability analysis of urban green space system based on GIS, and sensitiv-
ity analysis that is to test the validity of scores, weights used and the ranking of alternatives by using
DSS (decision support system), with the aim of developing an approach of GIS-based suitability
analysis to identify suitable sites for urban green space system development. This thesis is structured
into five chapters:
Chapter 1 states the research background, problem, objectives and questions as well as a workflow.
Chapter 2 presents a literature review about the definition, classifications of the urban green space
system and comprehensive benefits of the urban green spaces.
Chapter 3 states the research methodology including the definition, methods of suitability analysis,
GIS-based traditional suitability analysis model and its improvement. In addition, weighting methods,
evaluation methods for ranking, and sensitivity analysis are involved in this chapter. Based on the
above analysis, a methodology flow chart will come out to direct the case study in Dongguan munici-
pality.
Chapter 4 describes the case study in Dongguan municipality, which includes the introduction in the
study area, data collection, and data processing as well as data analysis. Afterwards, six suitability
scenarios will be generated by using some GIS techniques; a ranking and sensitivity analysis is used
to get the best suitability scenario and test its validity.
Chapter 5 gives a conclusion about the suitability analysis. Some recommendations are provided as
well in this chapter.
6
2. Definition and conceptions of urban

green space system
2.1. Definition of urban green space system

Green spaces refer to those land uses that are covered with natural or man-made vegetation in the
built-up areas and planning areas (Wu 1999). It has been long argued about the definition of green
space system. Different disciplines have proposed different definitions from their own professional
angles, such as Horticultural Greenland System, Urban Greenland System, Ecological Greenland Sys-
tem, Urban Green Space, Green Open Space. The meaning of green space system has also been con-
tinuously developing with the development of city theory, which mainly involves horticultural, eco-
logical and spatial these three meanings. Figure 2.1 obviously shows the developmental skeleton of
green space system. It is a process gradually developing from non-existence to existence, from sim-
plicity to complexity.
Name Discipline Meaning
Horticultural
Landscape
Greenland Horticultural
Horticulture
System
Ecological Urban Planning

Greenland and Ecological
System Urban Ecology
Green Landscape
Open Planning and Spatial
Space Urban Design
Figure 2.1 Developmental skeleton of green space system
Echoing the opinions of A.R Beer (1997), green spaces are: ‘Places where contact with animals and
birds and the more attractive insects like butterflies’, ‘Places with visual variety’, ‘Places are children
can learn about nature and social life through contact with animals’, ‘Places to loiter in and watch the
world go by’, ‘Places to chat while children play’ (Mugenyi 2002). Referring to some definitions from
other countries such as Britain, America, Japan (Table 2.1), some scholars have proposed the defini-
tion of Green Open Space from the angle of landscape planning and urban design. Lingzhang (2001)
defines green open space as all the areas within the city and its surrounding regions, enabling people
to contact the nature. Thus green space system is endowed with spatial meaning.
7
Table 2.1 Definitions of green open space

Country Form Purpose or function
Residential land that the area for architectural-use Courtyard, recreational land
Britain is lower than 1/20 of the whole area (excluding
wasteland)
America Land with natural environment Recreational land, land for adjust-
(1) ing urban construction
America Non architectural-use land (e.g. air, land, water) Recreational land, landscape area,
(2) national forestry, roadside green
belt
Japan Non architectural-use land Park, square, gym, zoo, botanic
(1) garden (excluding road and canal)
Japan Non architectural-use land Park, game land, gym, graveyard,
(2) farmland, forestry land
Source: Gaoyuan Rongzhong, Yang Zhengzhi et al. translate, 1983, Urban green space planning, P5,
Table 1-1.
This study wants to propose its own definition of green space system by referring to some published
literatures, with the aim of integrating horticultural meaning, ecological meaning with spatial mean-
ing. The detailed definition of urban green space system in this study is that, in the urban spatial
environment, there are some good green areas (green space per capita must be over 9.0 square me-
ters), which are mainly covered with natural or man-made vegetation and can function as ecological
balance, playing an active role to urban environment, landscape, and residents recreation. They also
include those water areas enabling people to contact the nature and those greenways that can connect
parks, productive and defensive green spaces, residential green spaces, landscape areas and suburban
forest.
2.2. Classification of urban green space system

Green space system can be grouped in different classes according to different classifying standards.
As to the element of terrain, it can be classified into mountain, water, forestry, farmland and road
these five classes. Green space system can also be classified into patch, area, line and point by its
forms. However, the most practical and efficient method to classify green space system is based on its
functions. Both China and other foreign countries adopt this method to classify their national green
space systems.
2.2.1 Classification in foreign countries

There is not a uniform method to classify green space system in the world till now. Different countries
have proposed different classifications based on the function, size, and physical characteristics of
green space system. America classifies the park according to its service radius (Table 2.2).
8
Table 2.2 Classification of parks in America

Class Area Serving population Service radius
2
Children’s park 200 ~ 400 m 500~2500 Neighbourhood
Small pleasance 200 ~ 400 m2 500~2500 Neighbourhood
Neighbouring park 2 ~ 8 ha 2000~10000 400 ~ 800 m
District park 8 ~ 40 ha 10000~50000 800 ~ 5000 m
Large urban park ≥ 40 ha ≥ 50000 Riding distance within
half an hour (by car)
Regional park ≥ 100 ha Serving a larger region Riding distance within
an hour (by car)
Specific facility Including avenue, seashore, square, historic relic, floodplain, small park,
lawn, forestry land, etc.
Modified from: Jia Jianzhong, 2001, Planning and design of green space system, P17, Table 2-3.
Japan has carried out the Establishment Of Green Comprehensive Planning since April 1977, in order
to apply a green comprehensive planning in the urban planning areas, construct and protect urban
parks, green lands and public spaces. This planning was modified every five years. In addition, with
the help of those subsidiary laws such as Natural Parks Law, Metropolitan Parks law, Children Parks
law, etc, an integrated urban green space system was formed. (Table 2.3)
Table 2.3 Classification of urban green space system in Japan

First class Second class Third class
Park green land, playground,
Public green space park road, footway, bikeway
Square
Park graveyard
River, lake, waterway
Public Natural green space Seashore, riverside, lakeside
green Mountain forestry, weald, farm-
space land
Churchyard, graveyard and its
affiliated land
Gaoyuan Open green space Affiliated garden plot of com-
Rongzhong monweal facility
classification Garden plot of individual facility
Sharing residential garden plot
Sharing recreation facility
Sharing green space Enterprise welfare facility
Private School playground, other garden
green plots
space Individual garden plot
Specific green space Testing land of nursery
Water supply, drainage and other
facilities
9
Park for ordinary use

Park for district use
Park Park for special use
Park for regional use
Spatial Park with special forms
classification Graveyard
of Street tree
urban Traffic Pavement tree
green space Park avenue
space Expressway
system Sharing road
Other Pleasance
green Golf course
space Industrial green space
Public Park and green land
green Square and playground
space Graveyard
Classification Other similar green spaces
of green Including water area, riverside
comprehensive belt, farmland, forestry weald,
planning Other churchyard, public affiliated
green green space, pleasance, school,
space agricultural experimental land,
etc. (their areas must be over
1000 square meters)
2.2.2. Classification in China

Chinese classification of green space system is also developed step by step. Urban And Rural Plan-
ning (1961) classified green space system into four classes: public green space, quarter and street area
green space, specific green space, landscape and recuperation green space. In 1973, National Con-
struction Committee classified green space system into public green space, courtyard green space,
street tree, suburban green space and defensive green space these five classes. Urban Horticultural
Green Space Planning (1981) had six classes: public green space, residential green space, affiliated
green space, traffic green space, landscape area green space, productive and defensive green space.
There are seven classes in Urban Greening Byelaw (1992), which included public green space, resi-
dential green space, departmental affiliated green space, defensive forestry, productive green space,
landscape forestry and main road green space. Urban Landuse Classification And Standard, a national
standard, only has two classes: public green space, productive and defensive green space.
In the past few years, some scholars have proposed different practical classifications of green space
system to meet the new needs of urban constructions. Meanwhile, the government has also established
the Classification Standard Of Urban Green Space System as a national standard since 1993. (Table
2.4)
10
Table 2.4 Classification of urban green space system in China

First class Second class
Municipal comprehensive park, district comprehensive park,
Park residential comprehensive park, botanical garden, zoo, chil-
dren park, etc.
Street side green Small pleasance, avenue, garden belt, square green space, etc.
space
Residential green Green space in residential district, green space in residential
space quarter, green space in street area, etc.
Departmental affili- Affiliated green space in the factory, school, hospital, hotel,
Jia Jian- ated green space warehouse, municipal public facility, etc.
zhong Roadside green space Roadside tree, affiliated green space of road.
(2001) Defensive forestry of health, industry, railway, etc, wind-
Defensive green defensive forestry, cuneal green space, water and soil conser-
space vation forestry, etc.
Productive green Nursery, flower garden, grass garden, etc.
space
Landscape green Landscape forestry, forestry parcel, and other independent
space forestry parcels.
Suburban ecological Landscape area, forestry garden, natural conservation forestry,
green space waterhead conservation forestry, farmland forestry network,
orchard, and other forestry lands.
Park G1 Skeleton park G11, specific park G12, historic relic park G13,
park belt G14, street corner green space G15, square green
space G16.
Productive green Nursery and flower garden G21, orchard and forestry land G22.
space G2
Defensive green Urban wind-defensive forestry belt G31, health-defensive for-
space G3 est belt G32, safety-defensive forest belt G33.
Residential green
space R0
Roadside green space
Wu Renwei S0
(1999) Industrial green space M0, warehouse green space W0, public
Affiliated green facility green space C0, municipal sharing facility green space
space G0 U0, external traffic green space T0, affiliated green space for
specific landuse P0.
Suburban ecological
landscape conserva-
tion B1
Suburban ecological
forestry land B2
Suburban defensive
forestry B3
11
Park G1 Comprehensive park G11, specific park G12, park belt G13,
street side pleasance G14.
Productive green
space G2
Defensive green
Draft space G3
national Residential green
standard space G4
(1999) Public facility green space G51, industrial green space G52,
Affiliated green warehouse green space G53, external traffic green space G54,
space G5 roadside green space G55, municipal facility green space G56,
specific green space G57.
Ecological landscape
green space G6
Table 2.4 has presented three typical classifications of urban green space system in China. It can be
showed that there are two obvious tendencies in this table:
(1) These three classifications use the name of Park (green space) instead of the name of Public green
space that has been used in urban planning and green space system planning for a long time, in order
to combine it with international terminology. Meanwhile, this name can be better to embody the green
space functions rather than only represent its affiliated relation and serving object. As such, the green-
ing index of “Public green space per capita” used in the past ten years will also be replaced by “Park
area per capita”.
(2) These three classifications regard the urban green space system from the regional perspective.
They concern more on those suburban green spaces that can play a good ecological role to the city
(e.g. Suburban ecological green space proposed by Jia Jianzhong, Ecological landscape green space in
the Draft national standard). Wu Renwei classifies these suburban green spaces into Suburban eco-
logical landscape conservation, Suburban ecological forestry land and Suburban defensive forestry.
The common is to elicit such a conception to integrate suburban green spaces into planning system,
but they don’t be counted in urban landuse balance and urban greening indices.
There are some differences in the detailed classes because these three classifications are based on dif-
ferent perspectives. It is obvious that the classification of Jia Jianzhong (2001) has a good link with
the traditional landscape horticultural classification, but it has no detailed classes and indices for ex-
planation. The classification of Wu Renwei (1999) and draft national standard pay more emphasises
on the practice, and separately has detailed functions and indices for explanation. From the view of
authority, it is more practical to take the draft national standard into application. However, it is feasi-
ble to adopt other two classifications for the second class. For example, it can adopt the classification
of Wu Renwei for the second class of Productive green space G2 and Defensive green space G3. As to
the Residential green space G4, it can use the classification of Jia Jianzhong (2001) and explain it with
corresponding indices. Ecological landscape green space G6 can also be replaced by B1, B2, and B3
from the classification of Wu Renwei. In addition, Allowing for the function of green space system, it
12
is feasible to change Square green space belonging to the second class of Street side pleasance G14
into a new second class of G15. Likewise, it can cite the foreign conception of Greenway to convert
the External traffic green space G54 and Roadside green space G55 into a new class of G17.
2.3. Comprehensive benefits of urban green spaces

Green space system has a great effect on the urban feature. Only a good concordance between the
man-made environment and natural environment can generate a suitable human settlement. As a re-
cycling organization of urban ecological system, green space system has been prevalently concerned
by the society. People instinctively have intimate psychology to green spaces at the beginning, and
now they have transferred to rationally study the benefits of green spaces. Western scholars concern
more on the quality of green space benefits. W.Miller (1996) has grouped the functions of urban green
spaces in three classes: architecture and aesthetics function, climate function, engineering function. In
our country, the scholars are affected by the theory of sustainable development. They emphasize on
ecology, society, and economy these three aspects (Ping 1994). This research will expatiate on the
comprehensive benefits of urban green spaces by using the method of Chinese three classifications,
including ecological benefits, social benefits and economic benefits.
2.3.1. Ecological benefits
1. Clean air
(1) Balance carbon and oxygen: Vegetation can release O2 and absorb CO2 in the photosynthesis,
which play an important role in balancing carbon and oxygen. In the urban environment, such balance
needs to be maintained much more by green spaces because of the more oxygen consumptions. It has
been measured that 1 hectare broadleaves can consume 1 ton CO2 and release 0.75 ton O2 everyday in
the growing season. If an adult resident absorbs 0.75 kilogram O2 and releases 0.90 kilogram CO2
every day, the balance between carbon and oxygen for one person will need 10 square metres forestry
or more than 25 square metres lawns to maintain (Lingzhang 2001). Some German experts have
proved that, as to people’s breath plus fuel’s burning, only 30~40 square metres green spaces for
every resident can keep the balance between O2 and CO2 within the city. Based on this theory, some
countries determined that green space per capita should be 40 square metres when planning the urban
green space system.
(2) Absorb toxic gas: There are more and more toxic gases existing in the air with the improvement
of industrial level, which mainly include SO2, NOx, Cl2, HF, NH3, Hg, etc. Under some concentra-
tions, however, many kinds of vegetation can absorb toxic gases into their bodies through the laminas’
pores and tresses’ lenticels, and use redox to transfer them into non-toxic gases, or exclude those toxic
gases out of their bodies by the root system or get them together in some organs. As such, vegetation
can play a cleaning function to air pollutions. Some researches have showed that 1-hectare Japan ce-
dars can absorb 720 kilogram SO2 every year. The concentration of HF will be reduced to 47.9%
when going through a green belt of 40-metre width.
13
(3) Trap dust: Dust is one of the main air pollutions besides toxic gases. Vegetation, particularly
trees, can effectively hold up, filtrate and absorb dusts. This is because trees have strong crowns and
their leaves are covered by hairs and excretive greases, which enable trees to play an active role in
trapping dusts. For example, it can slow down the wind through the trees shielding function. The pol-
luting dusts particles will be eliminated after they fall down to the ground. It was reported that in
Hamburger (1966), the annual average value of dusts was over 850 milligram per square metres in the
urban areas almost with no trees. While in the suburban areas, this average value around the parks
with flourishing trees was lower than 100 milligram per square metres. It has been measured in Bei-
jing, when the greening coverage rate (see section 4.2.3) was 10%, the total number of suspending
dusts particles was reduced 15.7%. While the greening coverage rate was 40%, this number was re-
duced 62.9%.
2. Adjust and improve urban climate

(1) Adjust “Urban Heat Island”: Large areas of paved surfaces dissipate the heat of the sun very
slowly. This results in the urban heat island effect where a city heats up rapidly and then maintains a
high temperature (World Bank Report on green space use, May 1997). Trees and other vegetation can
use their transpiration to dissipate steams into the air. During this process, the temperature on the
leaves and surrounding temperature will drop because of consuming heat. At the same time, trees can
slow down the wind and play a shielding function to reduce the energy requested by the buildings.
Thus green spaces can effectively reduce the urban energy consumptions (W.Miller 1996). In Phoenix
(1992), America, Akbari used computer to simulate and predict that when the greening coverage rate
reached 25%, the temperature would drop 6~100F at 14:00PM in summer (July). In China, The Minis-
try of Land and Resources has measured that, when the greening coverage rate is lower than 20%, en-
ergy consumption in the vegetation transpiration is lower than the energy attained from the sun radia-
tion. While the greening coverage rate reaches 37.38%, this situation is on the contrary. This time
green spaces absorb energy from the urban environment, so they will have a good effect on the envi-
ronment.
(2) Improve urban climate suitability: According to W.Miller (1996), there were four elements in-
fluencing urban climate: sun radiation, air temperature, humidity, and airflow. The frontal two ele-
ments have been mentioned before. Here it primarily concerns about the latter two elements. Vegeta-
tion leaf surface can play a transpiration function that can not only drop down the temperature but
also increase the humidity. Some researches have proved that 1-hectare forestry can transpire 8000-
ton water and absorb 4 billion calorie heat every year. So green spaces can improve 4%~30% air hu-
midity. Generally the range where massive green spaces can adjust the humidity, is equal to the dis-
tance around the green spaces that is 10~20 times than the tree height, even enlarging to the
neighbouring districts of 500-metre service radius. Moreover, green spaces can hold back, lead, rotate,
and filtrate airflow (Li 1999). In order to prevent the wind hazard, it can use green belts that are verti-
cal to the main wind direction to form a barrier. The density, highness of green belts and the distance
of conserved areas are the most important to influence on the wind speed. Those green spaces in riv-
erside and lakeside can be used to lead the natural airflow from suburbs to the inner city. As such, the
air convection is improved.
14
3. Prevent and reduce hazard

(1) Prevent earthquake and fire
Urban green spaces can be used to evacuate the residents when earthquake or fire takes place. Accord-
ing to the needs for protecting environment and preventing hazards, urban green spaces area should be
higher than 30% of the total urban area. In addition, vegetation leaf contains plenty of water and can
slow down the wind, so it can play an effective role in preventing fire. Green spaces should be more
than 3 hectare as a refuge location, and it will be more effective to plant those non- inflammable trees
around the refuge location. In 1970, Japanese construction bureau has investigated in the Report for
Earthquake and Fire in Tokyo, 63% of the flameout reason was the suitable distribution of green
spaces and the existence of rivers.
(2) Water and soil conservation

Vegetation dense leaf surface can prevent spate directly impacting the soil. The fallen leaves cover the
soil and reduce the impact that flowing water exerts on the ground. Also, the root system can tightly
complect in the soil and fix the sand, stone to prevent erosion. Thus green spaces have a good effect
on adjusting flood and preventing soil being eroded. It has been measured that, when the slop is 300
and rainfall is 200 millimetres per hour, the erosion ability of soil is respectively 0%, 11%, 49% and
100%, according to the lawn’s coverage rate of 100%, 91%, 60% and 31%. Therefore, urban green
spaces can play a good function to the water and soil conservation, through holding back rain, slowing
down the wind and fixing the soil by their root systems.
4. Eliminate noise
As one kind of the environmental pollution, noise will have a bad effect on residents’ health when it is
over 70-decibel. The most effective method to eliminate noise is to make a suitable green space sys-
tem. The surface of tree’s stem and leaf is very rough. Its numerous tiny pores and dense hairs can
prevent the sound wave from transmitting, all of which can function as eliminating the noise. It has
been proved that a 4.4-meter width green belt can eliminate 6-decibel noise. 40-meter width multiple
hierarchical green spaces combined by arbors, shrubs and grasses can eliminate 10~15 decibel noise.
Noise will be eliminated much better if green spaces are closer to the noise source. Likewise, the more
flourishing the green spaces, the better the effect of eliminating noise. A denser and wider green belt
of 19~30 meter integrated with a soft soil surface can eliminate 50% of the total noise. Actually it is
impossible to construct very wide green belts in the city because of the limited spaces. If it is ration-
ally designed, however, even a 6-meter width green belt can have some good effects on eliminating
the noise. Furthermore, the barrier function of green spaces can give people a kind of psychological
relief that they can eliminate noise.
2.3.2. Social benefits
1. Recreation
Recreation is one of the four urban basic functions in the Athens Charter. Active life itself is recrea-
tion and recreation is the essence of life (Wu 1998). Arousal-seeking theory says recreation is a kind
of behaviour that can improve the individual arousal level. It is created by the interactive requirements
between the individuality and environment, or between the individuality and society. The task of rec-
15
reation is to provide people with opportunities to optimise the arousal level. When the urban planners
and landscape designers regard green spaces as an important designing element and take it into appli-
cation, they will virtually create the active open spaces where people can have a rest and play. With
the improvement of material and culture, urban residents put their more emphasis on pursuing outdoor
recreations. They are fond of carrying out all kinds of recreations in the green space system just be-
cause of its diverse representations, multiple functions and intimate characteristics. For example, the
green space system in Guangzhou is an important resource and base for tour. Its greening rate (see
section 4.2.3) in some sense determines the attractiveness for tourists, and this attractiveness is able to
effectively promote the sale and production of urban tour products.
In order to make the residents have a good recreation in the green space system, it needs to ensure
every one can have some suitable areas of green spaces. Thus a standard was made that green space
per capita should be over 60 square metres. If 10% of the total urban residents spend the holiday in
the green space system at the same time, then every one should have at least 6 square metres public
green spaces. According to the research that Chinese Urban Planning and Designing Institute has
made on the requirements for recreation green spaces, public green space per capita F=P*f/e. Where,
P is the residents travelling rate (%) in the holiday. This rate was 8% in 1988 and it will increase 1%
every four years. f represents the area that every traveller should have. It is 60 square metres per cap-
ita in the large park while 30 square metres per capita in the residential park. e is the turnover coeffi-
cient representing the percentage that current travellers divide by total travellers in the rush hour. The
calculated result is public green space for planning is 7~11 square metres per capita applied from
2000 to 2010.
For the majority of the middle and low-income class people, a public green space serves as an
essential meeting place, a place where they can go and spend their time while relaxing. It is true that
green spaces are centres of recreation. This is more prevalent if the green space is within 10-minute
walking distance. In many developing countries, not many amenities are offered by green spaces.
Thus a proportion of people occupy themselves by playing games, others go walking and the rest
simply take view of the green space surroundings from a distance. The urban poor generally have few
affordable options for recreation, and thus place a high value on green areas (Mugenyi 2002).
2. Landscape aesthetics
From the view of architects, green space is the “soft component” that composes an integrated urban
space with ‘hard space’ enclosed by entities. The landscape function of green spaces mainly reflects
on space, time and location these three elements (Hesheng 1999). Green vegetation can enrich the ur-
ban architecture complex skyline and intenerate the hard space through their different forms, colours
and styles. So green spaces can not only beautify the urban feature and set off architectures, but also
improve aesthetic effects, which makes the urban environment more uniform and more diverse.
Meanwhile, in order to embody the landscape value of symbolic aesthetics, it can combine different
kinds of green spaces to enclose, plot out spaces and then to create a good urban space image.
In his bookmaking The Image of City, Lynch (1961) has proposed five kinds of image elements from
the perspective of landscape sense: path, edges, district, nodes and landmarks. He pointed out that im-
age is resulted from the interaction between environment and observer. Urban green space is the criti-
16
cal element for people to recognize and grasp the landscape structure. And it has a strong imageability
because of its tuneful colours, integrated shape, intimate scale and obvious greenness. On the other
hand, green space has become an important element to embody the urban culture and reconstruct the
urban feature. This is because more and more people have felt city has its own characteristics, also
called ‘Local Spirit’ (Shi 2002). It means that thinking and emotion, which are based on the local
natural characteristics, can create a specific cultural landscape with those natural landscapes such as
local terrain, soil, vegetation, water body, etc. as the urban green landscape line, green spaces gener-
ally occupy 25%~30% of the urban landuse, which will be the main element influencing on the urban
feature. In addition, every green space has its specific form, colour and style. All of these characteris-
tics will have a good expression of the ‘Local Spirit’. A good example is Lincoln Park, Grand Park
and Jackson Park connected by the green belts, have their own playground, botanic garden, gallery,
museum and other facilities, which has endowed the city with more cultural meanings and creates a
large-scale, impressive green space system in the world.
3. Adjust psychology
Green space can play a psychological role to people. M.J.Cohen (1993) has summarized the psycho-
logical effects on the nature. He concluded there are 97 kinds of human activities related to the nature,
which can produce 49 kinds of satisfactions. From the view of chromatics, lakes’ blueness and vegeta-
tion’ greenness belong to impassive colours that can make people calm down. If there are not enough
blueness and greenness but full of exciting redness in the city, there will be no peaceful environment
for the residents (Shi 2002). Thus it can be showed that people must live together with the nature. In
general, social benefits of green spaces originate from the potential influence that green spaces act on
people’s psychological behaviour model. This research tries to explicate it from the perspective of
environmental behaviour.
People, as a single entity, have a mutual relationship with the existing environment. Kurt Lewin, a
German psychologist, described such relationship as the following formula, which constitutes his ba-
sic frame of “Field Theory” (Shi 2002) :
B=f (P, E)
B—behaviour P—personality E—environment
The above three parameters can transfer with each other. It means that people’s behaviour is the
mutual result of realistic nature and social environment. Based on the essential grasp to people’s
behaviour, here it adopts the explanation about the basic requirement and inner motivity in the
hierarchic requirement theory (Figure 2.2), to analyse people’s behaviour within the city. In contrast
to the functions of green space system, it can be found that green spaces with a beautiful environment
help to eliminate physical tiredness and mental oppression, and satisfy people’s physiological re-
quirements. Also, green spaces in a good layout condition can create some relatively private and pri-
vate spaces, not only making people feel homelike and relaxed, but also satisfying people’s safety re-
quirement. Moreover, green spaces can maintain a beautiful, clean, comfortable environment for
working and study, and provide spaces for rest and outdoor communication, which can satisfy peo-
ple’s requirement for going back to the nature and the need of ownership and love. In addition, recrea-
tion can satisfy the upper two requirements (in Figure 2.2).
17
Requirements for ego reality: knowledge, idea, ambition, etc.
Requirements for respect: achievement, power, right, fame, etc.
Requirements for ownership and love: recreation, ownership, friends’ love, etc.
Safety requirements: avoiding danger and horror, etc.
Physiological requirements: hunger, thirst, air, sex, etc.
Figure 2.2 Hierarchical requirement theory (Abraham H. Maslow)

Note: The above five requirements are step-up in turn, namely the upper hierarchic requirement will
be considered only after the lower hierarchic requirement is satisfied.
Source: Zhu Zhixian (1989), Psychology gradus, P808, Figure a.
4. Education
Generally it is easy to ignore the education function in the social benefit of green spaces. Actually just
because of the functions of recreation, landscape aesthetics and creating urban feature, green space
has been a powerful media to potentially transmit all kinds of information and emotion in the city,
which will influence on the residents’ personality. Here it adopts the model of individual behaviour
(Shi 2002) to explain the above process. The model is:
B=HELP
B—Behaviour, H—Heritage, E—Environment, L—Learning, P—Pursuit
Where, the element of H cannot be changed but other elements can be adjusted gradually. It is neces-
sary to emphasize the active effects of green spaces exerting on E (environment), L (learning), and P
(pursuit). As an external environment, green space itself is a nature museum and outdoor classroom. It
is also a good place for disseminating and spreading the science. As such, people can have chances to
directly approach the green spaces, learn more about them and consciously cherish them. In a fresh
and quiet living environment, people’s life style and behaviour will be potentially affected. A suitable
green space system can arouse residents’ environmental consciousness, make them produce pursuit,
and desire for learning the green spaces environment. After the step-by-step learning and conscious
observation about the nature’s vicissitude, people will have more profound comprehensions on their
own surroundings. Thus a good circular ring is formed (Figure 2.3).
18
Environment
stimulating understanding
Education
Pursuit Learning
promoting
Figure 2.3 Circular ring for education function of green spaces
In order to stimulate the working of the above circular ring, it is critical to create an elegant environ-
ment and make a suitable green space system, which can provide people with more chances to ap-
proach the nature, particularly for the youth. It is impossible; otherwise, to appeal that people should
care about their surrounding ecological environment and control their improper daily behaviours.
Nowadays there are many foreign schools locating in the good green space system. They entitle stu-
dents to manage the vegetative districts in parcels. This helps to improve the children’s commonweal
conception and make up for those disadvantages in the classroom. It is beneficial for children to learn
more about the nature, and it can improve their consciousness, creativity, imagination, the spirit of
loving life and go-ahead. Therefore, it is necessary to attach more importance to the social benefits of
green spaces, and the more important is to stimulate the working of the above education circular ring.
2.3.3. Economic benefits

People often concern about the economic benefits of urban green spaces, but it is relatively difficult to
be measured. The commercial value of urban green spaces includes shadow price and market price
(Wu 1999). Shadow price is attached to some public products. For example, producing oxygen, ab-
sorbing toxic gases and trapping dusts can save energy. Those safety preventions such as earthquake
and fire prevention, water and soil conservation can reduce some loss. In 1995, when the temperature
was over 35 centigrade in Tokyo, the electricity consumption of air-conditions would be 1200 thou-
sand kilowatt as long as the temperature rose 1 centigrade. It has been estimated that, 100 million
adult trees can save 30 billion kilowatt electricity every year in American cities, which is equal to
save 2 billion dollars energy consumption. Thus it can be showed that it is efficient to make use of
green spaces to save energy through dropping down the temperature.
Market price is composed of three parts. One is some tangible products can directly generate the mar-
ket price, such as the productions of lumbers, drugs, nurseries, fruit gardens, etc. The production value
of flowers is 13 billion dollars every year in the Netherlands, German and many other countries. An-
other part is some intangible products can also generate the market price, such as the increase of sur-
rounding land price, the increment of service, etc. In the new district of Pao Ya in Dalian city, the
housing price rises from original 800~1000 RMB per square meters to 2000 RMB per square meters
after the green spaces are suitably distributed. The last part of market price is also attached to some
intangible products, which can generate the market price but are not realized by substance exchange.
19
For example, the ticket prices in some landscape areas and the prices of travelling service, construct-
ing urban feature for improving the investing environment belong to this part. So it is very important
to create an elegant environment to attract more investors, and then it will have good effects on the
urban economy. In Dalian, China, the newly increased public green spaces are 6580 thousand square
metres from 1994 to 1999, and the greening coverage rate reaches 40%. Meanwhile, the number of the
joint, cooperative and single-foreign-invested enterprises has been developed from 1400 to 7000. This
is the economic benefits resulted from the environmental benefits.
There are two conceptions of broad-sense and narrow-sense to calculate the economic benefits of ur-
ban green spaces. Broad-sense means it only calculates the substantial outputs represented by value
and the benefits attained through management. Narrow-sense tries to convert the ecological, social
and market value into currency unit, with the aim of reflecting the real value within the green spaces.
At present, it is feasible to calculate the economic benefits but difficult to calculate the social benefits.
Some scholars have used ecological benefits to reflect the social benefits, namely ecological benefits
contain social meanings. In general, it is an arduous task for us to calculate the total benefits in the
urban green spaces. How to calculate and which is the best method? It is a burning issue.
20
3. Methodology
3.1. Definition of suitability analysis

Land suitability analysis is the process of determining the fitness of a given tract of land for a defined
use (Hopkins 1977; Steiner 1983). In other words, suitability analysis is the process to determine
whether the land resource is suitable for some specific uses and to determine its suitability level. It is
an important analytical method for ecological planning. Land suitability refers to the inherent suitabil-
ity of the land for some specific, persistent uses. This land is determined by such characters as hy-
drology, geography, topography, geology, biology, sociology, etc. Land suitability will have no mean-
ings unless it is relevant to some specific uses, and it is very important for making good use of land
and promoting the land’s social value.
3.2. Suitability analysis methods

There have been many analytical methods since suitability analysis came into being, which primarily
include the method of sieve mapping, landscape unit method, grey tone method (map overlay) and
computer method (GIS). The frontal three methods have been explained in section 1.1. Here it puts
the emphasis on the GIS method, which can also be divided into three classes: direct overlay,
weighted score, and ecological factors combination.
3.2.1. Direct overlay

The method of direct overlay includes map overlay and equal-weight summation. Map overlay can be
traced back to the beginning of 20th century. According to McHarg (1969), this method can be suc-
cessfully applied in land use suitability, which enables urban planning efficiently and comprehen-
sively to allow for the social and environmental factors. The main steps of map overlay can be con-
cluded as: (1) Defining the planning purpose and identifying the factors contributed to the planning.
(2) Investigating each factor’s situation and distribution (forming ecological purpose), making a clas-
sification according to the suitability for some specific land uses, and using some gradual colours to
identify each factor’s suitability class in a single factor map. (3) Overlaying two or more single factor
maps to get a composite map. (4) Analysing the composite map and finally making the land use plan-
ning. In the planning of Staten Island, McHarg and his colleagues applied this method to analyse land
use suitability of natural conservation, passive recreation, active recreation, housing development,
commerce development and industry development, etc, which has made a great effect.
Map overlay is a kind of visual and intuitionistic method. It can integrate environmental factors with
social-economic factors to make the suitability analysis. The disadvantage of this method is that it is
essentially a kind of equal-weight additive method. Actually each factor’s function is different and
sometimes the same factor may be considered repeatedly. Another advantage is that while the factors
increase, it is rather complicate to use the gradual colours to represent different suitability classes and
to make the overlay. Moreover, it is difficult to identify the little difference from the gradual colours
in the composite map. Anyway, map overlay plays an important historical role in the development of
21
ecological suitability analysis. Afterwards, many new methods are developed primarily based on this
method.
The method of equal-weight summation is first to quantify the factor’s class, then to make a direct
addition and finally to get a composite evaluation value. This method takes advantage of different
numerical values (map overlay uses gradual colours) to represent the suitability class, which can over-
come the inconvenient map overlay and the difficulty to identify the gradual colours. The formula of
equal-weight summation is presented below (the premise of such direct overlay method is that each
factor’s influence on the specific land use is similar and independent):
n
Vij = Bkij
k =1
Formula 3.1 Equal-weight summation
Where, i represents the parcel number or gird number; j represents the land use number; k repre-
sents the number of the ecological factor influencing the j th land use; n represents the total of the
ecological factors; Bkij represents the suitability evaluation value of the k th ecological factor in the
in the i th parcel of the j th land use ( single factor evaluation value); Vij represents the composite
evaluation value in the i th parcel of the j th land use (composite ecological suitability of the j th land
use).
3.2.2. Weighted score

When all kinds of ecological factors’ influences on the specific land use are very obvious, it can’t
make a direct overlay to get the composite suitability. It must take advantage of the method of
weighted score. The principle of this method is similar to that of the equal-weight summation. The
difference is that it needs to identify each factor’s relative importance (weight) in the weighted score.
The more influence on the specific land use, the higher weight for the factor. On the basis of scoring
each single factor class, it will carry out the weighted summation for the evaluation result of each sin-
gle factor. Finally we will get the total scores of the corresponding parcels or grids of the specific land
use. Generally a higher score represents the more suitability. The formula of weighted score is showed
as follow. Where, Wk is the weight of the k th factor for the j th land use. Other symbols are the same
as the above method of equal-weight summation.
n
Vij = BkijWk
k =1
Formula 3.2 Weighted score
The method of weighted score overcomes those disadvantages in the method of equal-weight summa-
tion. Another important advantage of this method is able to make a girding, classification and
22
quantification in the map, which is suitable for the computer application. This is why this method is
so widely applied in the past few years. Whether the method of direct overlay or the method of
weighted score, however, the mathematics theory requires that each factor should be independent. Ac-
tually many factors have mutual relationships and mutual influences with each other. In order to over-
come this disadvantage, ecological planning experts create a new method of ‘ecological factors com-
bination’.
3.2.3. Ecological factors combination

As mentioned above, direct overlay and weighted score require that each factor should be independ-
ent. Actually many factors depend on each other. For example, it is unsuitable to construct an ex-
pressway when the slope is over ‘30%’, no matter how the drainage condition is. But according to the
weighted score or direct overlay, when the slope is over ‘30%’ and the drainage condition is very
good, perhaps it will get the moderate suitability. The method of ecological factors combination ac-
knowledges that different combinations of the dependent factors determine the suitability of the spe-
cific land use.
This method can be classified into hierarchical combination and non-hierarchical combination. The
method of hierarchical combination is first to use a set of dependent factors to identify the suitability
level, then to regard these dependent factors as a new factor and to combine it with other dependent
factors to identify the final suitability level. The method of non-hierarchical combination is to com-
bine all the dependent factors to identify the suitability level at the same time. Obviously, this method
is suitable for the analysis with a few factors. And it is useful to apply the hierarchical combination in
the analysis with large number of factors. Whether the method of hierarchical combination or non-
hierarchical combination, first it is necessary for experts to establish a set of complicate and inte-
grated dependent factors and an evaluation standard. This is the most critical and difficult step to ap-
ply the method of ecological factors combination in the suitability analysis. Allowing for the data
available and the limited time for the fieldwork, the method of weighted score was selected to carry
out the GIS-based suitability analysis in this research (see section 4.3.5).
3.3. GIS application in suitability analysis

GIS (Geographic Information System) is a computerised system that facilitates the phases of data
entry, data analysis, and data presentation especially in cases when we are dealing with georeferenced
data (By, Knippers et al. 2000). Generally it includes six parts: data source selection and standardiza-
tion, data pre-processing, data entry, data management, data analysis and displaying, mapping. The
core is the analysis function containing overlay processing, neighbourhood comparison, grid analysis,
measurement statistics, etc. The geographic information can be collected from field investigation,
map, remote sensing, environmental monitoring, and from social-economic data sources.
One of the burning issues in the GIS application is to establish the application and analysis models
according to different requirements, such as land suitability evaluation, ecological sensitive areas
analysis, ecological benefits analysis, and so on. In the suitability analysis, GIS offers spatial overlay
capabilities and grid-cell processing methods that enable us to analyse spatial factors (Shuaib 1998).
23
Its development and capability to overlay digital maps has made suitability mapping easier and
quicker. Its potential for being linked with the planning process through the development and
application of relevant models, can be realised by its functionality. Since suitability analysis deals
with the analysis of several data sets, GIS can effectively be used in looking at the characteristics of
land from a number of layers for each location to solve a problem. GIS-based models can be used to
create simplified representations of phenomena. This is done in GIS by combining different sets of
map layers to analyse the relationships between them.
GIS-based suitability analysis models generally take advantage of the method of weighted score and
the method of ecological factors combination (see section 3.2). IGIS (Intelligent Geographic Informa-
tion System) is the latest development of GIS technology. It can integrate the expert knowledge with
the immense functions in the computer system, which will have great potentials later on. Just because
GIS can process enormous data and have the powerful functions of displaying and outputting maps, it
will be the main tendency to apply GIS in the suitability analysis.
3.4. GIS-based traditional suitability analysis model (TSAM) and its

improvement
Land suitability analysis is very important for the land planning and management. With the advent
and rapid development of GIS, land suitability analysis has been more and more applied in the urban
planning. One of the most important spatial analysis functions in GIS, overlay, is primarily designed
for the multiple factors evaluation (e.g. land suitability analysis). Nowadays land suitability analysis
has been developed into those evaluations such as agriculture suitability, graze suitability, forestry
suitability, suitability for urban expansion, site selection for specific landuse, and so on. In contrast to
the traditional pure-mathematics evaluation method, GIS-based suitability analysis can systematically
integrate mathematical calculation with map processing. This method is so intuitionistic, easily opera-
tional and quick that it can greatly improve the evaluation efficiency. This research will expatiate on
such GIS-based traditional suitability analysis model (TSAM), and will take advantage of the cer-
tainty factor to make an improvement and propose the improved traditional suitability analysis model
(ITSAM).
3.4.1. Traditional suitability analysis model (TSAM)
1. TSAM procedure
Land suitability analysis is to determine the suitability for some specific landuses by scoring the land.
In other words, the higher the score, the more suitable the land for some specific landuses. The tradi-
tional suitability analysis includes the following three steps:
(1) Selecting suitability factor. Each factor is represented by a thematic map in GIS. (2) Single factor
analysis. Namely according to the single-factor evaluation standard, score is given to the map unit of
each factor and then the single factor suitability map is generated. The score is normally grouped into
‘1’, ‘2’, ‘3’ (ratio value) these three classes, which respectively represents ‘no suitability’, ‘moderate
suitability’ and ‘high suitability’. Sometimes the score can be grouped into five or six classes based on
different requirements. (3) Multiple factors overlay. First weights are assigned to the suitability fac-
24
tors according to their relative importance. The weights are determined by statistic integration and
hierarchic analysis in this research (see section 3.5). Then it takes advantage of Formula 3.2 to calcu-
late the composite score (final score).
It should be noted that in the vector-based TSAM, the map unit after overlay is the most basic unit
produced by multiple factors overlay. While in the raster-based TSAM, the map unit after overlay is
the same as the original map unit in the single-factor map, which is composed of regular-arranged
grids. Actually the principle of the vector-based TSAM is the same as that of the raster-based TSAM.
The difference between them is the map processing. In this research, raster-based structure is used to
carry out the GIS-based suitability analysis.
2. Example
The following is a simple example to select a suitable site for new green spaces. In order to explain
the TSAM more clearly, here we only select two factors: existing land use, slope. We assume there
are only two kinds of land uses in the single factor map of existing land use: existing green space,
built up area. The existing green space includes an old park and a new park, and only residential land
is in the built up area. Their codes are ‘26’, ‘25’, ‘24’. The single factor map of slope is grouped into
three classes: 0% ≤ slope<1%, 1% ≤ slope<10%, 10% ≤ slope<20%. Here the slope is ‘0.1%’, ‘9.99%’,
‘10%’, and their codes are ‘16’, ‘15’, ‘14’. Scores given to these two single factors are presented in
Table 3.1. For simpleness we assume these two factors have the same weights, namely W1=W2=0.5.
According to the single factor analysis and the multiple factors overlay, this TSAM procedure can be
realized in Figure 3.1. The final overlay results show that, the higher the composite score, the more
suitable the land use for the new green spaces.
Table 3.1 Factors and weights in the traditional suitability analysis

Factor Suitability class Code Score Weight
Existing green space 26 3
(old park)
Existing land use Existing green space 25 3
(new park) 0.5
Built up area 24 1
(residential land)
0% ≤ slope<1% (0.1%) 16 3
Slope 1% ≤ slope<10% (9.99%) 15 2 0.5
10% ≤ slope<20% (10%) 14 1
25
16 (0.1%)
26 25 24 15 (9.99%)
14 (10%)
3
3 3 1 2
1
× W1 × W2
3.0 3.0 2.0

2.5 2.5 1.5
2.0 2.0 1.0
Figure 3.1 TSAM
3.4.2. Improved traditional suitability analysis model (ITSAM)

From the above TSAM, we can see that the single factor analysis has two disadvantages: (1) One is
that no difference is identified among the different subclasses within the same suitability class. For
example, an old park (subclass) and a new park (subclass) belong to the same suitability class of ‘ex-
isting green space’ and they get the same score of ‘3’. It is obvious that, however, the new park is less
suitable for the site selection of new green spaces because its acquisition is more expensive than the
old park. The slope of ‘1.01%’ and ‘9.99%’ are in the same suitability class of ‘1% ≤ slope<10%
(9.99%)’. The suitability in these two subclasses is obviously different, because the expense for level-
ling off the ground of ‘9.99%’ is much higher than that of ‘1.01%’. But these two subclasses get the
same score of ‘2’ in the TSAM that cannot identify their difference. (2) The other disadvantage is that
the difference between the two different suitability classes is exaggerated around the division of
scores. For example, there is very little difference between the slope of ‘9.99%’ and ‘10%’, but they
are divided into ‘moderate suitability’ and ‘no suitability’ and they respectively get the score of ‘2’
and ‘1’. Actually ‘moderate suitability’ and ‘no suitability’ are all fuzzy conceptions. One of the most
efficient methods to solve this problem is to use the membership degree in the fuzzy set theory, but
the membership degree is complicate to be integrated with GIS (Huang 1997). Thus in this research,
certainty factor is used to solve this problem instead of the membership degree. The certainty factor
obeys such a rule:
IF A THEN B
It means that if A holds then B holds. For example, if the land use belongs to the existing green space,
then it gets a score of ‘3’. As mentioned above, the old park and the new park belong to this class of
26
‘existing green space’, but they are obviously different. In order to exactly identify the difference be-
tween them, the user has to give a certainty factor (CF) to this rule:
IF A THEN B (CF=0.8)
It means the certainty that the user is certain of the rule ‘if A holds then B holds’ is 80%. For example,
we can give a certainty factor of ‘0.8’ to the old park, which means the certainty of the rule ‘if the
land use is an old park then it gets the score of 3’ is 80%. Likewise, we can give the new park a cer-
tainty factor of ‘0.3’, to reflect the certainty that the new park gets the score of ‘3’ is 30%. We also
can give the certainty factor of ‘0.2’, ‘0.4’, or other values less than ‘0.8’ to the new park. The value
of the certainty factor should be based on the current situation and the practical requirements. Thus it
can be showed that the certainty factor arranges from ‘0’ to ‘1’. There are two methods to determine
the certainty factor:
(1) The certainty factor is determined by the user. This method is suitable for the qualitative evalua-
tion. For example, the suitability of the old park and the new park are difficult to be described by a
mathematical formula. So it has to be determined by the user based on the real condition. As men-
tioned above, the certainty factor of the old park to get the score of ‘3’ is ‘0.8’, while that of the new
park to get the score of ‘3’ is ‘0.3’. As such, the suitability difference between the subclass of the old
park and the new park is identified.
(2) The certainty factor is determined by establishing a certainty factor function. This method is suit-
able for the quantitative evaluation (e.g. the continuous variable). For example, the suitability class of
‘0% ≤ slope<1%’ gets the score of ‘3’ in the slope map. We assume the slope between ‘0%’ and ‘1%’
is linear. Thus a certainty factor function can be established to identify the difference between the
slope of ‘0%’ and ‘1%’. The other two certainty factor functions for the class of ‘1% ≤ slope<10%’
and the class of ‘10% ≤ slope<20%’ are established in the same way (Formula 3.3):
x−0 x−0
CF ( x ) = 1 − = 1− (0% ≤ x% < 1%)
1− 0 1
x −1 x −1
CF ( x ) = 1 − = 1− (1% ≤ x% < 10%)
10 − 1 9
x − 10 x − 10
CF ( x ) = 1 − = 1− (10% ≤ x% < 20%)
20 − 10 10
Formula 3.3 Certainty factor function
0. 1 − 0
Where, x represents the slope value. If x% = 0.1% , its CF = 1 − = 0.9 , which means the
1
certainty that the slope of ‘0.1%’ can get the score of ‘3’ is 90%. If x% = 9.99% , its
9.99 − 1
CF = 1 − = 0.001 , which means the certainty that the slope of ‘9.99%’ can get the score of
9
27
10 − 10
‘2’ is 0.1%. Likewise, if x% = 10% , its CF = 1 − = 1 . The original score and the certainty
10
factors for each single factor have been determined, now we can use Formula 3.4 to calculate the
composite certainty factor CF '( x) (composite single factor score).
CF '( x) = S ( x) − 1 + CF ( x)
Formula 3.4 Composite certainty factor
Where, S ( x) represents the original single factor score, namely S ( x) = ‘1’, ‘2’, or ‘3’. According to
the above principles and the multiple factors overly theory, the procedure of the improved traditional
suitability analysis model (ITSAM) is realized in Figure 3.2. It is showed that this ITSAM encom-
passes more enriched and precise information than the TSAM (Figure 3.1). In other words, the differ-
ence among the different subclasses within the same suitability class is identified (e.g. the difference
between the old park and the new park), and the difference between the two different suitability
classes around the division of scores is not exaggerated but reduced (e.g. the difference between the
slope of ‘9.99%’ and ‘10%’). Therefore, the ITSAM can provide more options for the decision mak-
ers.
16 (0.1%)
26 25 24 15 (9.99%)
14 (10%)
3 0.9
3 3 1 0.8 0.3 1 2 0.001
1 1
2.900
2.8 2.3 1.0 1.001
1
× W1 × W2
2.90 2.60 1.95

1.90 1.65 1.00
1.90 1.65 1.00
Figure 3.2 ITSAM
28
3.4.3. Summary
This research has proposed two land suitability analysis models including TSAM and ITSAM. They
are all designed based on the multiple factors evaluation theory and the GIS technology. According to
the different practical requirements, users can change the weights and the single factor score, and
quickly get the final evaluation results. ITSAM, which takes advantage of the certainty factor, can
greatly simplify the operation in GIS, and has a more powerful application than the membership de-
gree in the fuzzy set theory. Determining the membership degree needs to consider the certainty that
the evaluation object belongs to each suitability class. If the land suitability analysis involves multiple
hierarchies, the evaluation procedure will be made rather complicate. The determination of the cer-
tainty factor, however, only needs to consider the certainty that the evaluation object belongs to one
suitability class, and it doesn’t need to assign the uncertainty to the other suitability classes. There-
fore, the certainty factor is easier integrated with GIS than the membership degree. The major disad-
vantage of using the certainty factor is that it has some subjectivity to determine the certainty factor in
the qualitative evaluation (e.g. the certainty factor ranges from ‘0’ to ‘1’, but it is somewhat subjective
to give the certainty factor of ‘0.8’ and ‘0.3’ to the old park and the new park). And if the values in the
single factor map are not linear, the certainty factor function is not very suitable for this quantitative
evaluation.
3.5. Weighting methods

“Weighted suitability is more complex, because we do not only consider the binary
suitable/unsuitable for each aspect but we compare suitability scales for all aspects and we give them
a weighting factor that should reflect their relatively importance. ” (Bruijn 1991) “Weighting should
be applied when not all aspects have an equal importance. It should be realized that the choice of a
weight is most important, as it has a great effect through multiplication of the scores.” (Hofstee and
Brussel 1999) Assigning the weights to factors is a critical element in suitability analysis. Weighting
factors are often based on a mixture of implicit knowledge, personal experience and individual opin-
ions. In order to reduce the subjective bias as much as possible, this research will use three weighting
methods based on three different considerations in the study area.
3.5.1. Statistic integration

The method of statistic integration is a kind of special ‘expert-assess’ method. It is based on a large
number of investigations with many experts. Then it carries out the statistic induction and uses the
method of integration to get the final weighting values. The following is an example using statistic
integration to calculate the factors weights in the environmental quality evaluation. Four factors in-
cluding air pollution, surface water pollution, groundwater pollution, and noise pollution are selected
out for this evaluation. First an investigating table is sent to some experts for comparing the factors
importance (Table 3.2). Secondly according to the feedbacks and investigation information, it will
carry out the statistic induction and represent the information with percent style (Table 3.3).
29
Table 3.2 Investigating table of importance order

Importance Air pollution Surface water pollution Groundwater pollution Noise pollution
order
First
Second
Third
Last
Table 3.3 Information table of statistic induction (%)

Factor First importance Second importance Third importance Last importance
Air pollution 71% 22% 7% 0
Surface water pollution 3% 78% 19% 0
Groundwater pollution 13% 20% 19% 48%
Noise pollution 2% 31% 45% 22%
Table 3.2 is an example for one expert to compare the importance the importance among the evalua-
tion factors. The table shows that in this expert’s opinion, air pollution is the most important for
evaluating the environmental quality, followed by noise pollution, surface water pollution and
groundwater pollution. Other experts will also be asked to show their opinions by filling in the same
kind of investigating table. All the experts’ opinions about the factors importance order have been
showed in Table 3.3 by statistic induction. For example, as to the factor of air pollution, 71% of the
experts acknowledge it is the most important, second importance with 22%, third importance with
7%, and none of the experts think air pollution is the least important.
Table 3.3 is a statistic induction matrix and is represented by matrix D . Then we can create a weight-
ing order vector. This vector is represented by natural number order, namely C = ( 4,3,2,1) . Actually
it uses numerical number to identify the difference among the first importance, second importance,
third importance and last importance. Finally matrix D and vector C are integrated and each factor
weight is calculated according to Formula 3.5:
T
0.71 0.22 0.07 0 4
0.03 0.78 0.19 0 3
W = [ D • C T ]T = • = (3.64,2.84,1.98,2.13)
0.13 0.20 0.19 0.48 2
0.02 0.31 0.45 0.22 1
Formula 3.5 Statistic integration
After the normalization, the original weighting vector, W = (3.64,2.84,1.98,2.13) , is converted into
the final weighting vector, W = (0.344,0.268,0.187,0.201) . As such, each factor weight is showed
in Table 3.4.
30
Table 3.4 Factors weights by statistic integration

Factor Air pollution Surface water pollution Groundwater pollution Noise pollution
Weight 0.344 0.268 0.187 0.201
There are some advantages in applying the method of statistic integration. First this method is based
on inquiring many experts with different contexts. Each weight is calculated by statistic induction and
integration, and the calculated values are relatively precise. It is suitable for the evaluation that needs
to allow for large number of factors. The most disadvantage of statistic integration is that it has some
subjectivity to create the weighting order vector (vector C ). This weighting order vector has an im-
portant effect on the relationship among the final values, namely different weighting order vectors
will generate different weights according to Formula 3.5, and the ratio among the weights will be dif-
ferent. To some extent, the final weights calculated by the method of statistic integration are relatively
precise and scientific, but it can’t completely get away from the man-made interference after all.
3.5.2. Hierarchic analysis of nine-degree

Nowadays it is prevalent to apply the method of hierarchic analysis of nine-degree in the multicriteria
weighting system (Xu 1999; R.Ramanathan 2001; Sharifi and Herwijnen 2003). First this method
needs to compare evaluation factors with each other (pairwise comparison), including comparing the
factor with itself. Secondly it will use a nine-degree scale to transform the comparison judgements
into numerical values. The nine-degree scale is explained in Table 3.5. Then a structural judgment
matrix (Table 3.6) is generated by every factor’s numerical value. After the normalization, the eigen-
vector of the largest eigenvalue derived from the structural judgment matrix is the corresponding fac-
tors’ weighting vector.
Table 3.5 Importance comparison of nine-degree
Degree Description of pairwise comparison judgements
1 Factor i is equally important as factor j, or comparing i with i, or j with j.
3 Factor i is moderately more important than factor j.
5 Factor i is strongly more important than factor j.
7 Factor i is very strongly more important than factor j.
9 Factor i is extremely more important than factor j.
2,4,6,8 The importance is between the upper degree and lower degree.
Reciprocals It is contrary to the above importance, namely factor j is more important than factor i.
Table 3.6 Structural judgment matrix of nine-degree

Factor Air Surface water Groundwater Noise
pollution pollution pollution pollution
Air pollution 1 3 8 4
Surface water pollution 1/3 1 5 2
Groundwater pollution 1/8 1/5 1 1/4
Noise pollution 1/4 1/2 4 1
Here it also uses the example of environmental quality evaluation to explain the method of hierarchic
analysis of nine-degree. The example includes four factors: air pollution, surface water pollution,
31
groundwater pollution, and noise pollution. First it needs to make the pairwise comparison based on
the stakeholder analysis. For example, when air pollution (factor i) is compared with surface water
pollution (factor j), if four experts acknowledge that factor i is moderately more important than factor
j in the urban environment (the value should be ‘3’ according to the nine-degree scale), while three
experts accept that factor i is strongly more important than factor j (the value should be ‘5’). So the
final numerical value will be ‘3’ in the structural judgment matrix. As such, the structural judgment
matrix (Table 3.6) will be generated by the numerical values based on Table 3.5 and the stakeholder
analysis.
It should be noted that in Table 3.6, all the values are ‘1’ in the main diagonal of the structural judge-
ment matrix (self comparison). The corresponding values between the upper triangle and the lower
triangle are reciprocals (the upper triangle represents comparing factor i with factor j while the lower
triangle represents comparing factor j with factor i). After the normalization, the eigenvector of the
largest eigenvalue is W = (0.56,0.24,0.05,0.15) , and those four factors weights are determined by
this eigenvector (Table 3.7).
Table 3.7 Factors weights by hierarchic analysis of nine-degree

Weight 0.56 0.24 0.05 0.15
The method of hierarchic analysis of nine-degree can be potentially useful for determining factors
weighs in many ways. The powerful mathematics is the main reason why this method is so prevalent.
It helps to elicit the complex judgements of different experts in a common platform. It also ensures
accuracy in the sense that it has an inbuilt method to check the inconsistency of judgements (see:
Saaty and Vargas, 1984). This ensures that the judgements are provided only with sufficient care and
the error due to negligence is thus minimised. However, due to the fast growing number of pairwise
comparisons, it is not sensible to use the method of hierarchic analysis of nine-degree for a large set of
factors. In other words, it is difficult to determine the numerical value from the nine-degree scale
when too many factors need to be considered. For example, air pollution is more important than sur-
face water pollution for evaluating the environmental quality. But it is not easy to determine the value
of ‘3’, ‘4’, or ‘5’ to identify the importance scale. Hence, the final eigenvector of the largest eigen-
value derived from the structural judgement matrix also has some subjectivity.
3.5.3. Hierarchic analysis of three-degree

The analysis above shows one of the disadvantages of hierarchic analysis of nine-degree, which is
relatively difficult to convert the comparison judgements into numerical values according to the nine-
degree scale. It is easier to only judge the more important, same important and less important these
three classes in the pairwise comparison. Thus the method of hierarchic analysis of three-degree is
developed. The first step of hierarchic analysis of three-degree is to make a comparison matrix based
on the three-degree pairwise comparison. Then a structural judgment matrix is generated by the above
comparison matrix and mathematical formula. After the normalization, all the factors weights will be
represented by the eigenvector of the largest eigenvalue derived from the structural judgment matrix.
The example of this method is explained below:
32
(1) Comparing all the factors with each other and quantitating by Formula 3.6, and using the quantita-
tive values to generate the comparison matrix (Table 3.8).
0 (factor i is less important than factor j )

K ij = 1 (factor i is as important as factor j , or the factor is compared with itself)
2 (factor i is more important than factor j )
Formula 3.6 Quantitative comparison of three-degree
In the above formula, K ij is the comparing quantitative value between factor i and factor j . It is the
corresponding element in the comparison matrix. If i = j , namely K 11 , K 22 , K 33 , K 44 , it means fac-
tors are compared with themselves, the value will be ‘1’ in the main diagonal of the comparison ma-
trix (Table 3.8). Just like the structural judgment matrix of nine-degree (Table 3.6), the values in this
comparison matrix of three-degree are determined based on Formula 3.6 and the stakeholder analysis.
Table 3.8 Comparison matrix of three-degree

4
Factor K1 K2 K3 K4
Ki
i =1
Air pollution K1 1 2 2 2 7
Surface water pollution K2 0 1 2 2 5
Groundwater pollution K3 0 0 1 0 1
Noise pollution K4 0 0 2 1 3
Table 3.9 Structural judgment matrix of three-degree

ri rj
r1 r2 r3 r4
Air pollution r1 1 10/3 8 17/3
Surface water pollution r2 3/10 1 17/3 10/3
Groundwater pollution r3 1/8 3/17 1 3/10
Noise pollution r4 3/17 3/10 10/3 1
4
(2) Using the cumulant ( K i ) from the comparison matrix to make the structural judgment matrix
i =1
by Formula 3.7.
Ki − K j
• (bm − 1) + 1( K i ≥ K j )
K max − K min
rij =
Ki − K j
1 /[ • (1 − bm ) + 1]( K i < K j )
K max − K min
Formula 3.7 Hierarchic analysis of three-degree
33
In the above formula, rij represents the elements in the structural judgment matrix. K i , K j respec-
4
tively represents the rows cumulant ( K i ) of the corresponding factor in the comparison matrix.
i =1
For example, K 1 = 7, K 2 = 5, K 3 = 1, K 4 = 3 in Table 3.8. K max , K min are the maximum value and
4
minimum value in K i . bm is equal to K max + K min , which is the sum of the corresponding values
i =1
between the most important factor and the lest important factor. Here bm = 8 . Thus the structural
judgment matrix (Table 3.9) is made.
(3) Just like the hierarchic analysis of nine-degree, after the normalization the eigenvector of the larg-
est eigenvalue derived from the structural judgment matrix is W = (0.588,0.256,0.048,0.108) . Note
that the eigenvector is showed as a vector, and the values in this vector represent the corresponding
factors weights. Unlike the integration matrix in the method of statistic integration, hierarchic analysis
of three-degree needn’t create a weighting order vector to generate the structural judgement matrix.
These two different matrixes contain two different kinds of information. The factors weights are
showed in Table 3.10.
Table 3.10 Factors weights by hierarchic analysis of three-degree

Weight 0.588 0.256 0.048 0.108
Comparing with hierarchic analysis of nine-degree, the method of hierarchic analysis of three-degree
is easier for decision makers to understand and take into application. It is more explicit and has better
certainty. Particularly when a large number of evaluation factors need to be considered, it is easier to
make the pairwise comparison and transform the comparison judgements into numerical values, only
allowing for the three-degree scale. Xu (1999) has proved that this advantage will be showed com-
pletely when it needs to evaluate more than ten factors. Here the example only allow for four factors,
so this advantage cannot be showed obviously. However, the discernment in hierarchic analysis of
three-degree is not very good. For example, one expert acknowledges that air pollution is very
strongly more important than groundwater pollution and it gets the value of ‘2’ in the three-degree
scale. Another expert thinks air pollution is weakly more important than groundwater pollution, but it
also gets the same value of ‘2’ according the three-degree scale. As such, two different opinions will
generate the same weighting system, which can be showed that the discernment in hierarchic analysis
of three-degree is not very powerful.
3.6. Evaluation methods for ranking

The aim of evaluation method for ranking is to reject “bad” alternatives or identify good ones, to bet-
ter understand the potentials and limitations of different alternatives, to identify the best alternative
and make a ordering of alternatives. In the opinion of Janssen (1994),“An evaluation method is any
procedure that supports the ranking of alternatives using one or more decision rules. An evaluation
method can generate:
34
(1) a complete ranking : A>B>C>D

(2) the best alternative : A>(B,C,D)
(3) a set of acceptable alternatives : (A,B,C)>D
(4) an incomplete ranking of alternatives : A>(B,C,D) or (A,B)>(C,D)
(5) a presentation of alternatives”
He thinks any criterion (factor) can be converted to a measurable quantity. The style is a kind of nu-
meric value that reflects the degree to which a particular objective is achieved. And “an objective is a
statement about the desired state of the system.”
Evaluation methods differ in the type of decision rule applied, the characteristics of the set of alterna-
tives they can handle and the type of rules used to value the attributes (Jankowski and Richard 1994).
As to the set of alternatives, this research is a kind of discrete decision problems comprising a finite
set of alternatives (six suitability scenarios in this research), while continuous decision problems are
characterized by an infinite number of feasible alternatives. The problem in discrete evaluation is to
identify the attractiveness of alternatives, based on two elements: 1) the consequences of the alterna-
tives in terms of the decision criteria; 2) the weights assigned to the criteria.
There are many methods to evaluate the ranking of alternatives till now. Most of them are designed to
process quantitative information on attributes (e.g. weighted summation, ideal point method). In order
to process qualitative or mixed information, some evaluation methods such as regime method, evamix
method or permutation method have been developed. According to the aggregation method, the DM-
analyst relationship and to the way of elucidation of the preferences, evaluation methods can be
grouped in three classes: compensatory approach, outranking approach and non-compensatory ap-
proach. Compensatory approach assumes that there is an absolute compensation in the evaluation,
which means high performance in one criterion of one alternative can compensate its weak perform-
ance in another criterion. Outranking approach is partially compensatory. It is based on the pairwise
comparisons of all alternatives and their outranking relations. Non-compensatory approach accepts the
hypothesis that there is no compensation among the criteria al all. (Table 3.11)
Table 3.11 Overview of evaluation methods for ranking

Classes Methods Processing information
Weighted summation Quantitative
Ideal point method Quantitative
Compensatory Multiattribute utility model Quantitative
Regime method Qualitative or mixed
Permutation method Qualitative or mixed
Evamix method Qualitative or mixed
Outranking Electre method Quantitative
Non-compensatory Dominance method Qualitative or mixed
All methods mentioned in the above table will be evaluated according to their information require-
ments, the efficiency of using information and the effectiveness of the method in terms of the results.
35
Allowing for the quantitative information in this research, it will use two methods including weighted
summation and electre method to make the ranking of alternatives (six suitability scenarios).
3.6.1. Weighted summation

Weighted summation is a simple and often used evaluation method. First all criteria scores need to be
standardized. Then an appraisal score is calculated for each alternative by multiplying each standard-
ized score by its corresponding weight, followed by summing of the weighted scores for all criteria.
The final ranking of alternatives can be formulated as follow:
i
S= ( xij wi ) (i = 1,......i; j = 1,...... j )
i =1
Formula 3.8 Weighted summation
Where S represents the final score of alternatives. The higher the S, the better the ranking of alterna-
tive. xij is the score of alternative j according to criterion i. wi is the weight assigned to the evalua-
tion criterion. Thus it can be showed that the final scores and ranking depend on the standardization
method and weights assigned. Weighted summation can only be applied if the attributes are additive
(Sharifi and Herwijnen 2003). It requires quantitative information on scores and weights. Only the
relative values of this information are used in the evaluation. Weighted summation can provide a
complete ranking and information on the relative differences between alternatives.
3.6.2. Electre method

Electre method, also known as concordance analysis, which is the most common technique based on
pairwise comparison, determines the ranking of alternatives by means of pairwise comparison. The
comparison is based on calculations of the concordance measure, which represents the degree of
dominance of alternative i over alternative k, and the discordance measure, which represents the de-
gree of dominance of alternative k over alternative i (Voogd 1983). According to these measures, the
differences between alternatives are quantified and a final score is calculated for every alternative.
From this score the alternatives can be ranked from best to worst.
Thus it can be showed the basic idea of this method is to measure the degree to which scores and their
associated weights confirm or contradict the dominant pairwise relationships among alternatives. This
dominance relationship for every pair of alternatives can be derived by a concordance index and a dis-
cordance index. The concordance index for the alternatives j and k, conc (j,k), which represents the
degree to which alternative j is better than alternative k, can be defined by summing the weights asso-
ciated with criteria contained in the concordance set (Formula 3.9). The discordance index, disc (j,k),
representing the degree to which alternative j is worse than alternative k, can be defined as the largest
of those differences between the scores of both alternatives from the discordance set (Formula 3.10).
This reflects that beyond a certain level, bad performance on one criterion cannot be compensated for
by good performance on the other criterion (Janssen 1994).
36
n
conc( j, k ) = wi ( wi = 1 , xij and xik are the original scores)
i: xij ≥ xik i =1
Formula 3.9 Concordance index
− − − −
disc ( j , k ) = max ( xik − xij ) ( xik and xij are the standardised scores)
i:xij ≥ xik
Formula 3.10 Discordance index
Then two concordance threshold values and two discordance threshold values have to be defined by
decision maker. These threshold values, in combination with the concordance and discordance indices
are used to establish a weak and a strong outranking relationship between each pair of alternatives.
Finally an iterative procedure of step-by-step elimination is used to convert the weak and the strong
graph representing these outranking relationships into an overall ranking of the alternatives.
3.6.3. Summary
The assumptions behind weighted summation are that weights are quantitative, the criteria scores are
determined on a ratio scale, and value is aggregated by addition. The most frequently violated
assumption is the second one. It is sometimes impossible to measure all the relevant criteria on a ratio
scale. Another potential disadvantage is that if the criteria weights are not linear, it will make the lin-
ear aggregation of value not very precise. The advantage of this method is its simplicity. However,
this simplicity not only limits the ranking of alternatives but also limits sensitivity analysis.
There are also some advantages and disadvantages in electre method. First, the pairwise comparison
between alternatives can provide a much more elegant evaluation of the relative performance of a
given alternative. Second, this method is relatively thorough because every alternative is compared
with every other alternative instead of the performance of the alternatives being stated by an average
achievement. Third, the threshold values give the user some control over what minimum standards are
to be enforced. However, this method is somewhat complex that only users familiar with this tech-
nique can understand the mechanics, which will make the control of threshold values meaningless. In
some sense, this analysis is a ‘black box’ approach that miraculously decides which alternative is the
best. Therefore, electre method makes better use of available information than the weighted summa-
tion, but it may not necessarily be better for decision support.
3.7. Sensitivity analysis

Sensitivity analysis is a technique that enables decision makers to test the validity of scores, weights
used and the ranking of alternatives. It can measure the impact on evaluation results of changing one
or more key input values about which there is uncertainty. In problems where the criterion scores and
weights can be estimated with complete certainty and where all evaluation methods yield the same
ranking of alternatives, this ranking is certain. In the majority of problems, however, scores and
weights are uncertain and evaluation methods involve different assumptions. According to Colson
37
(1989), uncertainty arises from lack of information or knowledge of some aspects of a decision prob-
lem. Sometimes uncertainty can be avoided by collecting more information, and sometimes it must be
accepted because it is due to unforeseen occurrences. The main factors of uncertainty are the role of
chance, the ambiguity and insufficient understanding of a problem and in making an abstraction of the
actual problem.
Multicriteria evaluation (MCE) consists of four main components: criteria, criteria scores, weights
(priorities) and the choice of an evaluation method. Generally there are some ambiguities existing in
one or all of these components, so the MCE results will always be associated with some uncertainties.
Since the aim of evaluation is to provide the decision maker with the best alternative or with a ranking
of alternatives, these uncertainties are only relevant to their impacts on the ranking. In this research,
therefore, procedures are introduced to analyse the sensitivity of the ranking to criteria scores and
weights.
3.7.1. Uncertainty on scores

Sensitivity of the ranking to the criteria scores can be tested from two aspects. One is to look at the
overall uncertainty of the scores. The other is to look at the influence on the ranking when only chang-
ing one score.
1. Overall uncertainty of the scores

Sensitivity of the ranking of alternatives to overall uncertainty in scores is analysed by using a Monte
Carlo approach (Janssen 1994). The decision maker has to estimate the maximum percentage that the
actual criteria values may differ from the estimated values included in the scores table. A random
generator is used to convert this information into a large number of scores tables around the original
scores table. Rankings are then determined for all scores tables and aggregated into a final ranking.
This method can be applied for every evaluation method resulting in a complete ranking. But it should
be noted that the user must pay more attention to the interdependences among the criteria when apply-
ing the random generator. Otherwise, they may greatly distort the results of sensitivity analysis on the
rankings of alternatives.
2. Uncertainty of one score

The second type of uncertainty is to look at the influence of changing one criterion score on the rank-
ing of two selected alternatives. This method generates a certain interval in which the ranking of the
two selected alternatives is not sensitive to changes in value of the selected criterion. If much debate
exists on the value of a certain criterion score and the certainty graph of this score shows that small
changes in the score result in a different ranking of the alternatives, then more research will be needed
to remove or decrease the uncertainty (Sharifi and Herwijnen 2003).
3.7.2. Sensitivity on weights

Weights or priorities, reflecting the relative importance among the criteria, also have a great effect on
the final ranking of alternatives. This is because there are some uncertainties existing in determining
38
the weights, particularly when the weights need to be quantitatively assessed. According to Sharifi
and Herwijnen (2003), sensitivity of the ranking to the weights can be tested by: (1) looking at the
overall uncertainty of the weights; (2) looking at the influence on the ranking when changing one
criterion weight; (3) looking at changes in all weights; (4) using different sets of weights. The first
two methods can be analysed in a similar manner as the above methods concerning the sensitivity of
the ranking to the criteria scores. Here it only emphasizes on the latter two methods.
1. Changes in all weights

This method determines which combination, most similar to the original combination of weights,
changes the ranking order of a selected alternative. The following three ways can carry out this opera-
tion (Sharifi and Herwijnen 2003). (1) Rank reversal of two alternatives: This procedure determines a
most similar weight combination for which the ranking of two selected alternatives changes. (2) First
position of an alternative: This procedure determines a most similar weight combination for which
one selected alternative becomes the best. (3) Loss of first position of an alternative: This procedure
determines a most similar weight combination for which the (selected) best alternative will be dis-
placed by another alternative.
2. Different sets of weights

This method aims to test the sensitivity of the ranking of alternatives to the weights by using different
sets of weights, which acknowledges the existence of more than one viewpoint and represents differ-
ent strategic viewpoints. For example, as to an environmental problem, different relevant departments
may have different viewpoints from their own purposes. So they will pay different attentions to the
same criteria of such an environmental problem. In other words, there will be different sets of weights
corresponding to the same criteria. Too many viewpoints may make things unclear and make it diffi-
cult to select a desired alternative. Therefore, it is important to have a few clear viewpoints and pay
more attention to the valuable one.
3.7.3. Summary
There are several advantages of using sensitivity analysis on the ranking of alternatives. First, sensi-
tivity analysis recognizes the uncertainty associated with the input and gives information about the
range of output variability. Second, it identifies evaluation criteria that may have the greatest potential
impact on the ranking. This can allow analyst to focus on those criteria that are most important. Third,
the process provides more information upon which to base a decision. In particular, it provides a no-
tion of where the impacts of uncertainty are important for the analysis and where they are not. This
will cause the analyst to gather additional information. Fourth, it is relatively easy to compute the
necessary information required for either approach. In fact, the analyst can assume a range of values
around the most likely case, without undertaking a great deal of work. Fifth, because the process re-
quires a careful examination of the criteria and weights most likely to influence ranking of alterna-
tives, the analyst is better informed as to what the results of the analysis truly represent. Finally sensi-
tivity analysis can be used on any measure of alternatives and can be used when there are little infor-
mation, resources, and time for more sophisticated techniques.
39
Several disadvantages are also prevalent. First, there is no explicit probabilistic measure of risk expo-
sure. That is, although the analyst might be sure that one of several outcomes might happen, the
analysis contains no explicit measure of their respective likelihood. So it does not reflect the effects of
diversification and does not incorporate any information about the possible magnitudes of the forecast
errors. Second, while the method fails to account for criteria interaction, the scenario approach usually
only includes a small number of potential scenarios. Finally, the lack of a systematic method for de-
termining the appropriate combination of criteria and weights used to define given scenarios limits the
reliability of sensitivity analysis.
3.8. Methodology flow chart

In the research methodology, first it pays attention to the urban planning context, such as location,
physical characteristics and social-economic characteristics in the study area. Secondly it needs to
propose the desirable environmental quality from two aspects: existing situation and greening indices.
And after the stakeholder analysis for suitability, seven suitability factors including air quality, land-
scape quality, surface water quality, historic culture value, water system influence, noise influence, and
existing land use, will be selected to carry out the GIS-based suitability analysis of the urban green
space system, by taking advantage of the method of weighted score.
As such, six suitability scenarios will come out according to different sets of scores and weights, and
certainty factors. Here, it uses statistic integration to calculate the weights for ‘suitability scenario 1’
and ‘suitability scenario 4’. Hierarchic analysis of nine-degree is applied in generating ‘suitability
scenario 2’ and ‘suitability scenario 5’. Weights calculated by hierarchic analysis of three-degree are
used to generate ‘suitability scenario 3’and ‘suitability scenario 6’. The first three suitability scenarios
are based on the traditional suitability analysis model (TSAM). Certainty factor is another important
element applied in generating ‘suitability scenario 4’, ‘scenario 5’ and ‘scenario 6’. These three suit-
ability scenarios are based on the improved traditional suitability analysis model (ITSAM). After-
wards, it wants to use MCA (multicriteria analysis) to carry out the ranking among the above six suit-
ability scenarios (alternatives). Sensitivity analysis is then used to test the validity of scores, weights
and the ranking of alternatives. Thus, we can make a conclusion based on all the above analyses. If the
conclusion is satisfied, we can take the final best suitability scenario into application. Otherwise, we
have to turn back to select another set of suitability factors, repeat all the above steps and try to get a
satisfied conclusion. The methodology flow chart is presented below:
40
Note: Calculating weights by statistic integration. Calculating weights by hierarchic analysis of

nine-degree. Calculating weights by hierarchic analysis of three-degree.
Figure 3.3 Methodology flow chart
41
4. Case study in Dongguan
4.1. Study area: Dongguan municipality
4.1.1. Location
China Administrative region of Dongguan
Guangdong
Wanjiang
Urban
centre
Dongcheng
Huangchun
Dongguan municipality
Map 4.1 The location of Dongguan municipality (study area)
Administrative region of Dongguan lies in the central-south of Guangdong province, and by the lower
passage of Dong River in the Pearl River Delta centre. To its north is Guangzhou and Shenzhen lies
south; to its east is Huizhou and Bolo while Panyu is parted by Pearl River on its west. It is situated in
the Guangzhou-Shenzhen-Hongkong economic corridor, which is renowned as a prosperous, high-
valued region.
42
This research only focuses on studying Dongguan municipality, which is located in the west north of
administrative region of Dongguan. Dongguan municipality is 50 kilometres far away from Guang-
zhou, 84 kilometres far away from Guangzhou and 80 kilometres far away from Guangzhou. It in-
cludes four districts: Urban centre, Huangchun district, Dongcheng district and Wanjiang district. Ac-
tually this study area is also a planning area validated from 2000 to 2015. It includes the existing ur-
ban area and some surrounding reserved areas for the urban future development. The location of this
study area (Gongguan municipality) is showed in Map 4.1.
4.1.2. Physical characteristics

The east south and centre of Dongguan municipality are mountains. Its north is a plain of Dong drain-
age area, while west and west north are close to Pear River port. So the terrain inclines from east
south to west north. Wanjiang River, Dongguan Channel and Dongguan Canal go through Dongguan
municipality from east north to west south. There are 3 big reservoirs in the south and one of them has
the water reserves of over 500 thousand. All of which form an abundant water network composed of
rivers, lakes and reservoirs. The average ground evaluation is only 2~3 metres in some areas (e.g. the
areas near Guangzhou-Shenzhen expressway in Huangchun district) while a century flood line is over
6 metres. So these areas are not suitable for urban construction and it is necessary to make green belts
to prevent the flood.
The climate of Dongguan municipality is characterized by very hot in summer and warm in winter. It
is very sunny and rainy. The annual average temperature is 22 centigrade and the absolute highest
temperature reaches 41.2 centigrade. Particularly in July and in August, the temperature usually
ranges from 26 centigrade to 40 centigrade. The annual average rainfall is 1788.6 millimetres, the
highest rainfall reaches 367.8 millimetres, and the annual average relative humidity is 79%. Thus it
can be showed that the climate in Dongguan municipality is not very suitable for a good human set-
tlement. With the aim of making a better adjustment and improvement to the urban climate, one of the
most efficient methods is to make a suitable urban green space system.
4.1.3. Social-economic characteristics

The total area of Dongguan municipality is 227.58 square kilometres, with a population of 826.7 thou-
sand including 316.4 thousand Hukou (Household registration) residents and 510.3 thousand tempo-
rary residents (1999). The built up area in Dongguan municipality is 54.12 square kilometres and its
total population is 668 thousand, which includes 240 thousand Hukou residents and 428 thousand
temporary residents. There are 158 thousand non-agriculture residents in the Hukou residents and the
landuse per capita in the built up area is 81.02 square metres. Thus we can see that the built up area is
relatively small and it is only 23.78% of the total area of Gongguan municipality, but the population
in the built up area is 80.8% of the total area. This is because the environment, public facilities, econ-
omy, etc. in the built up are much better than that in other areas. It is an arduous task for decision
makes to improve the urban environment by making a suitable green space system, in order to attract
those numerous population in the built up area to other developing areas, and finally to lead the urban
development into the right direction.
43
The economy of Dongguan is mainly export-oriented, most capitals coming from abroad; meanwhile,
raw materials also coming from overseas, and the product distribution are carried out by international
market. Dongguan’s foreign capitals originated from ‘San Lai Yi Bu’ enterprises. But these enter-
prises have been developed into the joint, cooperative and single-foreign-invested enterprises. The
gross domestic product (GDP) of Dongguan was 41.284 billion RMB in 1999, which was 17.9% more
than that in 1998. The average GDP has been improved from 3961 RMB in 1989 to 27561 RMB in
1999. Using overseas capital and the total imports-exports are respectively 1.457 billion dollars and
28.463 billion dollars in 1999, which are 8.91% and 22.3% more than that in 1998. Dongguan has be-
come an international processing and manufacturing base, and the industry of electronic communica-
tion and electric machine develop very quickly. As such, much pollution has been generated with the
fast economy development, particularly the industry development. Then how to eliminate this pollu-
tion and improve the urban environment, to keep the existing economy development and attract more
investors, it is a burning issue for the economists, urban planners and local decision makers.
4.2. Current green space system analysis in Dongguan
4.2.1. Existing situation of the green space system

The built up area in Dongguan municipality is 54.12 square kilometres (1999). There are nine public
green spaces existing in the study area, which include People’s Park, Qifeng Park, Keyuan Garden,
Wanjiang Park, Cultural Square, Huying Suburban Park, Fengjing Golf Course and Dongcheng Cul-
tural Square. The total areas of these public green spaces are 363.70 hectare (not including those natu-
ral reservations and green belts such as Qifeng Mountain, etc.). Public green space per capita (see sec-
tion 4.2.3) is 5.44 square metres, and greening coverage rate (see section 4.2.3) in the built up area is
35.1%. Compared with the national greening indices that green space per capita must be more than 9.0
square meters, and public green space per capita must be more than 7.0 square meters, the current
greening indices in Dongguan municipality are obviously not enough. So it is an important task for the
government to improve these greening indices and then to improve the environmental quality.
4.2.2. Problems existing in the green space system
1. Public green space and suburban forestry

(1) Public green spaces are not distributed uniformly. Public green space per capita is 5.44 square me-
tres, but most public green spaces are located in the east south. There are some green spaces in the
urban centre such as People’s Park, Keyuan Garden and Cultural Square. However, their areas are
only 25.7 hectare. Greening rate (see section 4.2.3) is relatively low in contrast to the numerous popu-
lation in the urban centre. The area of Qifeng Park is 90 hectare and it is more than 80% of the total
parks areas. It is situated in the urban fringe; the area is large enough but the using rate is very low.
There is only one park of 2.3 hectare in Wanjiang district and even no parks in Huangchun district.
Middle-scale parks and small-scale parks are extremely not enough in Dongguan municipality, so it
can’t satisfy the residents’ needs and play a good ecological role.
(2) It can’t use the predominant conditions of abundant mountains and water to construct a horticul-
tural city. Originally there were many mountains, lakes and rivers in the study area, and the natural
44
mountain-water skeleton was very good. But in the previous urban construction, some mountains were
destroyed and some lakes were used for other landuses. Moreover, the buildings were very close to
the rivers and lakes. Thus it can be showed that the favourable natural conditions can make Dongguan
be an elegant mountain-water horticultural city, but it cannot be well realized in the urban planning
and construction. Now the current suburban forestry can be regarded as the urban pulse, however,
only several types of trees existing in the forestry result in the bad ecological function.
(3) It cannot form a uniform-distributed park system integrated with large-scale parks, middle-scale
parks, and small-scale parks. The ecological diversity is poor and it lacks science and rationality to
distribute to the parks. The worst is that more and more parks are continuously replaced by other
commercial landuses.
2. Residential green space and departmental (work unit) affiliated green space
In Dongguan, one of the urban construction emphasises is to improve the environmental quality in the
residential areas and the work units. However, now it only focuses on the housing construction, re-
gardless of the residential green spaces. Particularly residential greening (afforestation) is the weak-
ness in the urban centre. In the residential districts, building density and floor area rate are high while
the road is narrow and there is a lack of green trees. The planning fieldwork has showed that there are
no formal parks in the residential districts except Bu Bu Gao residential quarter. So greening rate of
residential districts and work units is relatively low in the study area. Moreover, the quality and man-
agement of residential green spaces are not very good. Some high-class residential quarters can suc-
ceed in managing their green spaces, but most residential quarters are the exception, which has results
in the low environmental quality and poor ecological benefits (see section 2.3.1). The government
doesn’t attach enough importance to the balcony and roof greening. The problems existing in the
green spaces of work units are similar to that of residential districts. Namely greening rate is relatively
low and green spaces can’t play a good ecological role to the work units.
3. Road green space

The areas of road green spaces are 148.5 hectare in Dongguan municipality. Most roads are relatively
narrow with one board in the urban centre. Greening rate of roads is low and only one row of tree is
situated in the roadside. The new major roads are composed of three green belts: car-separated green
belt, street tree and roadside green belt. Most of them have been planted with certain amount of trees.
The current road greening has been finished quite well in the study area. Generally the trees can be
planted where the roads are constructed. The problems existing in the road green spaces are: (1) Ex-
cept that some roads have a good landscape, street trees in most roads are so young that they cannot
play a good function of beautifying roads, supplying passengers with umbrage and improving the en-
vironmental quality in the roadside. (2) As to the new three-board roads, the width of car-separated
green belt in some sections is less than 1.5 metres, which will influence on the arbors growth and the
landscape effects. (3) The breeds of street trees are not enough to form a kind of Dongguan-style road
green spaces.
45
4. Productive and defensive green space

There are productive green spaces of 78.5 hectare (nursery, flower garden, grass garden) in the study
area. Productive green space is a productive base to supply the city with seedlings. The deficiency of
productive green spaces will have a bad effect on the annual greening quality and quantity. Depending
on buying seedlings outside is difficult to ensure the breeds of required seedlings and their standards.
It will also influence on the surviving rate of new-planted trees. Thus the government has to plant
what it can buy. It is impossible to form a Dongguan-style urban landscape. In the suburb, farmlands
and forestry network have suffered certain destructions. So far, defensive green spaces are only 1.67
hectare.
5. Landscape forestry land

It is incompletely reported that the areas of landscape forestry lands are 6.1 hectare. They are mainly
located in Dongcheng district and Wanjiang district. The forestry lands inside the study area are very
different from those outside the study area. The inside section lacks in forestry while the outside sec-
tion is abundant of forestry.
4.2.3. Greening indices
1. Definitions of three greening indices

From the view of ecological balance, environmental protection and environmental improvement, there
must be some green spaces and a certain greening coverage rate in the city, which can play an active
role of ecological benefits (see section 2.3.1). Greening index is one of the important measures used to
plan the green space system and assess the environmental quality. It is a development expectation for
a certain period, which should be determined by the practical requirements, national situation and ur-
ban situation. So far, there are three greening indices used to assess the environmental quality in
China:
(1) Public green space per capita. It refers to the number of public green spaces that every resident
can own in the city. This index shows urban residents can directly share certain amount of green
spaces. To some extent, it also shows the condition with which green spaces can supply people’s rec-
reation. (2) Greening coverage rate. It refers to the percentage that the vertical projective areas of all
vegetation (arbour, shrub, grass, etc) share the total urban areas. (3) Greening rate. It refers to the
percentage that all kinds of green spaces share the total urban areas.
The detailed calculations of the above three greening indices are:
Total urban green spaces (m2) = Park areas + Public green space areas + Productive green space areas
+ Defensive green space areas + Ecological landscape green space areas + Specific green space areas
+ Road green space areas
Public green space per capita (m2/Person) = (Park areas + Public green space areas) / (Total urban
population)
46
Greening coverage rate (%) = (Park areas + Coverage areas of street tree + Coverage
areas of other green spaces) / (Total urban areas) *100%
Greening rate (%) = Total urban green spaces / (Total urban areas) *100%
2. Functions of three greening indices

Different countries have different urban greening indices according to their different situations, and
they use many types of indices to assess the environmental quality. In china, it primarily concerns
about the greening coverage rate and public green space per capita before 1990. These two indices
have played certain roles in history, however, they cannot exactly reflect the overall situation of the
environmental quality.
Greening coverage rate can reflect the coverage level within the city. It is one of the efficient stan-
dards used to assess the environmental quality. Except those massive grass lands, most greening cov-
erage rates are figured out by the stem size and by assuming different coverage areas. In China, urban
population is large and landuse per capita is relatively low. So in the urban afforestation, it should try
to develop the road greening and to plant trees in those vacant areas as much as possible. Some trees
‘occupying sky not occupying earth’ can efficiently improve the environmental quality. Thus it can be
showed that calculating greening coverage rate can accelerate the urban afforestation.
Now greening rate has been integrated with the greening coverage rate to act as the assessing index in
a district or a work unit. Thus it can be more exactly to reflect the environmental quality. But after
eliciting the greening rate, two problems should be analysed more critically. One is that it should not
use the unit of tree quantity in the planning and assessment of urban green space system, but use the
unit of green space area. The other is there should be a quantitative rule to identify to which extent
vegetation can be converted into green spaces, in order to ensure the green space system quality and
function as ecological benefits (see section 2.3.1). These two problems should be integrated with the
urban affroestation characteristics to make a further rule.
Public green space per capita is mainly to reflect the exoteric quality, namely it is an index to assess
the green space system for supplying people with recreation. From the view of ecological horticulture,
those public green spaces for recreation is one part of the urban green space system, and the function
of supplying people with recreation is only one aspect in the diverse social benefits (see section
2.3.2). According to Min (1999), recreation is the most important in the social benefits and it is feasi-
ble to regard public green space per capita as the most important index. However, only reflecting the
function of recreation is not enough. It should also reflect the total of all kinds of green spaces. The
index of greening rate that has been used since 1991 can make up this disadvantage. In summary,
greening rate integrated with public green space per capita and greening coverage rate can be used to
reflect the urban afforestation level and to assess the environmental quality in the round.
47
4.2.4. Desirable environmental quality in Dongguan

Some national indices have been proposed to assess the environmental quality in China (e.g. green
space per capita must be more than ‘9.0’ square meters and public green space per capita must be
more than ‘7.0’ square meters). Actually these indices can be regarded as the basic requirements that
the country exerts on the cities. Different city has different urban property, geographic environment,
historic condition and economic development level. So it is impossible to apply the same indices in
each city. It is not difficult to realize these indices and even have realized them in some cities. While
in other cities, it is difficult to do that and cannot realize these requirements. Guided by the national
policies, therefore, every city should take the existing situation into account to make some local-style
greening indices.
Based on the urban planning context, the existing situation of green space system and the national
greening indices, Dongguan planning bureau has proposed the desirable environmental quality:
Greening coverage rate in the planning area should be more than ‘50%’ until 2015, greening rate
should range from ‘30%~40%’, and public green space per capita in the built up area should be ‘10’
square metres (not including the Qingfeng mountain reservation and other landscape green belts). As
such, the government wants to import the forestry into the city and make the city approach the nature.
It will carry out the overall afforestation in the planning areas to reduce the loss of water and soil.
Taking advantage of the favourable climate condition and the good soil condition, Dongguan govern-
ment tries to construct an ecological city, realize sustainable development and create an optimal hu-
man settlement.
4.3. Suitability analysis of green space system based on GIS
4.3.1. Stakeholder analysis for suitability

Stakeholders are those whose interests are affected by the issue or those whose activities strongly af-
fect the issue. Stakeholder Analysis is a vital tool for identifying those people, groups and organisa-
tions who have significant and legitimate interests in specific urban issues. Clear understanding of the
potential roles and contributions of the many different stakeholders is a fundamental prerequisite for a
successful participatory urban governance process, and stakeholder analysis is a basic tool for achiev-
ing this understanding (Hemmati 2002). As such, four group stakeholders are involved in the analysis.
1. Urban planners
Six urban planners were invited to the group discussion. This group provided the most vivid discus-
sion, mostly about the potential for the development of tourism and the problems existing in the green
space system. Four urban planners acknowledged that a suitable green space system could not only
play an ecological role, but also attract more tourists just because of its rational layout and high land-
scape quality. But two planners only concerned about the ecological benefits of urban green spaces.
They felt the most important was that green spaces can effectively clean air, improve urban climate
and eliminate noise, and social benefits such as improving landscape quality, historic culture value
and economic benefits were not very important. Overall, the group accepted the comment that the
problems existing in the green space system were because it was not systematically planned and the
48
number of green spaces was not enough. So the urban planners had responsibility to solve these prob-
lems by planning a suitable green space system. They later acknowledged that in previous years they
had seen those land uses unsuitable for building because high water levels may threaten the buildings,
but now beginning to see the potential of those lands for attracting tourists, as they can be converted
into recreation green spaces.
2. Environmentalists
The group of environmentalists comprised five individuals. They were mostly concerned about the
pollution being generated by development and degrading the urban environment, and the existing
green spaces were not enough with the fast increase of urban population. Hence, a new suitable green
space system should be able to effectively deal with the pollution (e.g. toxic gas, dust, polluted water,
noise). On the other hand, the suitable sites for green spaces development should approach the water
source, and be located with appropriate slope and good surface water quality. If possible green spaces
can be expanded at the expense of other land uses, as long as they can play an ecological role to the
urban environment. Concern was expressed about the conflicts between ‘eco-tourism’ and mass tour-
ists activities. The group was against the building of large hotels and favoured the development of
small units, and a mixture of activities for tourists, such as building of a landscape corridor. Some en-
vironmentalists thought it was a good idea to build a new big park in the centre of built up area, as
tourists would arrive in greater numbers and gain easier assess to the area. However, other environ-
mentalists pointed out this new park could degrade the landscape quality and historic culture value,
and occupy many important land uses, having negative consequences for tourism in the longer term.
Thus it can be showed that the group was in favour of further development. They were concerned that
both economic benefits and ecological benefits should be considered in future plans, and there were
trade-offs to be made between the two.
3. Local residents
The group of local residents was composed of ten families, five with children and five with no chil-
dren. They felt that green space system played an important role to the human settlement but the exist-
ing green spaces were not enough with the development of living standard, and there was a responsi-
bility to preserve the existing green spaces rather than to destroy them. They were mostly concerned
about the problem of pollution resulted from the industry and traffic. The local residents acknowl-
edged that the fast growing industry and traffic had led to the degradation of air quality and the in-
crease of noise. However, they felt it was not their responsibility to talk to the industry employers, but
that of local government officials. So the government should pass laws to make the environment sus-
tainable. In addition, some families with children stated that there should be more green spaces for
their children’s recreation near their residential areas, while other families with no children hoped
there would be more parking grounds for them instead of the recreation green spaces.
4. Local government officials

This group comprised three individuals. Two officials provided a number of interests and ideas con-
cerning about the future development of Dongguan municipality, particularly insisted that the new
green space system should be first to consider the existing land use. In other words, new green spaces
49
should be restricted to certain areas such as those vacant areas and government reserved areas instead
of built up areas. Green spaces could be enjoyed but not interfere with future building developments.
Hence, the green spaces were perceived as one of a number of important land uses, which should have
defined boundaries, and should certainly not be expanded at the expense of other important land uses.
However, one official only focused on the local function that green space system can play. He stated
that as long as green spaces can play an active role in improving air quality, preventing flood and
eliminating noise, it was possible to convert some existing land uses of built up area into green spaces.
Overall, the group favoured the development of green space system to move some urban population
out of the built up area. They all regarded green spaces as an important local resource and accepted
that they had a responsibility for preserving the green spaces as a natural asset, and framed this in
terms of future ecological development.
4.3.2. Selecting suitability factors

One of the main concerns of physical planning is the proper designation of suitable sites for appropri-
ate land uses. The selection of suitable sites for specific land uses must be based upon a set of local
criteria to ensure that the maximum cost-benefit ratio for a community is attained (Hofstee and Brus-
sel 1999). In suitability analysis, the most important is the determination of how relative values, or
weights, are to be given to two or more combined factors. Factors are simple characteristics of land
that are grouped as attributes (Pease, Coughlin et al. 1996). Slope, soil type, surface water quality,
vegetative cover, existing land use, etc. are examples of such factors. Actually the meaning of ‘factor’
is the same as that of ‘criterion’. This research decides to use the name of ‘factor’ instead of ‘crite-
rion’ because ‘factor’ is more often used in suitability analysis.
In this case study, several sources of information are used to identify the ecological benefits of green
space system, factors and suitability classes are needed to perform the suitability analysis. The pri-
mary source for the identification of ecological benefits is the desirable environmental quality (see
section 4.2.4) set forth in the Dongguan Master Plan. The desirable environmental quality and its as-
sociated elements have been further supplemented by advice from community officials and the gen-
eral public. Factors within the ecological benefits are selected by using the stakeholder analysis for
suitability in the above section and the literature review (section 2.3.1). Information on the attributes
associated with each factor is derived from a combination of published literatures, expert opinions and
fieldwork. Here the attribute is defined as the categories, or classes, used to describe a factor, such as
air quality classes, landscape quality classes, etc. In this case study, the following seven factors in se-
lecting suitable sites for green space system development are considered: air quality, landscape
quality, surface water quality, historic culture value, water system influence, noise influence and
existing land use.
1. Available data
Suitability analysis of green space system requires many different types of data. Generally it should
allow for ecological, social and economic these three types of data. Due to the limited time for the
fieldwork and the difficulty to collect the spatial data, this research primarily emphasizes on the eco-
logical benefits. The nature of the spatial data is collected to represent this type of ecological benefits.
Factor and suitability are dependent on the nature of the component. Based on the above opinions,
50
what it should be first to do is to set up a GIS to deal with the site selection problem of green space
system. The following data set become available for analysis:
Air a CAD map of air quality distribution, derived from the air evaluation map.
Lscape a CAD map of landscape quality distribution, derived from the landscape planning
map and fieldwork.
Swater a CAD map of surface water quality, derived from the water evaluation map.
History a CAD map of historic culture value, derived from the planning map of historic
culture protection and fieldwork.
Wsystem a CAD map containing the major rivers, reservoirs and lakes in the study area,
derived from the water system distribution map.
Noise a CAD map of noise pollution distribution, derived from the noise evaluation map.
Luse a CAD map of urban land uses in 1999, derived from the existing land use map
and fieldwork.
2. Data pre-processing
As it has been mentioned above that, all the available data are AutoCAD maps (Dwgfiles). Some of
the data are in a format compatible for importing to the GIS. However, others have to need additional
processing before they can be incorporated. Since suitability analysis involves determination of the
suitability of an area, and the identification of those areas meeting specific factors, the primary entity-
type for this analysis (especially in the overlay operation) is polygon data. Where the spatial data are
not represented by polygons (vector or line data), they have to be converted into polygons by using a
buffering technique. Buffer widths should be assigned according to the suitability identifications (e.g.
distance to water system at 250m and 750m intervals). The basic procedure of data pre-processing has
been shown in Figure 4.1. First, it needs to create a new personal geodatabase by ArcCatalog in the
ArcGIS software. Secondly, it imports the original CAD map to geodatabase to generate polygons.
After that, the shapefile is coming out by exporting the geodatabase to shapfile. This study wants to
take advantage of ILWIS software to carry out the suitability. So the next step is to import the shape-
file to generate the polygon format compatible for the ILWIS analysis. Finally all the polygon maps
should be transferred into the raster maps through the operation of polygon to raster. As such, the data
pre-processing has been finished before it can use the slicing, distance calculation, overlay and other
GIS techniques.
Geodatabase (in ArcGIS)

Import CAD to geodatabase Export Geodatabase to shapefile
Polygon
CAD map (Dwgfile)
Polygon to raster Import by ILWIS
Raster Polygon Shapefile
Figure 4.1 Data pre-processing
51
4.3.3. Scoring
1. Scores of suitability factors

Scoring factors used in suitability analysis can be measured using interval or ratio values (scores) or
ordinal values (score classes). The choice of either scores or score classes can markedly influence the
efficiency of scoring factors. There is little consistency in the use of terminology in the suitability
analysis literature, and frequently the term ‘score’ is used to refer to both interval /ratio and ordinal
values. Interval or ratio values are referred to this research as ‘scores’ while ordinal values are termed
‘score classes’.
Ordinal values, as the name suggests, implies an order or ranking among factors. The order may be
either ascending or descending according to the application. And a ranking implies on an ordering
among factors but nothing more. Ordinal value is a kind of qualitative measurement scales that are
often used, especially in environmental decision problems. However, it should be noted that ordinal
values couldn’t be added or multiplied. Errors often arise from the addition of ordinal values. Interval
values have ordinal properties and they have the meaning about the intervals between objects. Corre-
sponding intervals on different parts of an interval scale have the same meaning (Sharifi and Herwi-
jnen 2003). For example, the interval between two objects with values of ‘20’ and ‘10’ (an interval of
10) is equal to the interval between two objects with values of ‘40’ and ‘30’. Interval values can also
be used in arithmetic operations such as addition and multiplication. But after adding or multiplying
interval values, it cannot be identified that a total of ‘60’ is twice as good as a total of ‘30’.
Ratio values are the highest level, which have ordinal and interval properties, as well as the property
of ratios. Just like the interval values, corresponding ratios on different parts of a ratio scale have the
same meaning. For example, the ratio between two objects with values of ‘10’ and ‘5’ is equivalent to
the ratio between two objects with values of ‘200’ and ‘100’. In this research, first it wants to give
scores and weights to the suitability factors, and then it will use multiplication and addition to get the
composite scores. Finally the composite suitability can be classified by mathematical rules based on
the composite scores. From the above opinions it can be showed that ratio values are more compatible
for scoring suitability factors in this suitability analysis. The description of suitability classes within
factors and corresponding scores are listed in Table 4.1. The scores of ‘3, 2,1’ are used to identify the
differences among high suitability, moderate suitability and no suitability. The maps of suitability
classes and scores in every factor (single factor map) are presented from Map 4.2 to Map 4.8. They
have been converted into raster format compatible for the operation in ILWIS.
52
Table 4.1 Suitability classes and scores

Factor Suitability Class description Score
class
Air quality
High Areas with low air quality 3
Moderate Areas with moderate air quality 2
No Areas with high air quality 1
Landscape quality
High Landscape area, garden scene area, reservoir scene area 3
Moderate Cuneal ecological green space, main road avenue, riverside green 2
space, urban square, landscape corridor, urban landscape axis
No Other areas with potential of good biodiversity 1
Surface water quality
High Some areas showing indications of bad erosion, little or no vegeta- 3
tion present, presence of toxic chemical products
Moderate Low levels of human activity, some bank stabilization needed, few 2
erosion factors present
No Banks are well vegetated, capable of trapping sediment and slowing 1
erosion, with potential for active nutrient cycling by aquatic plants
Historic culture value
High Tour area, reservoir conservation, old city relic, cultural relic and 3
historic site conservation
Moderate Landscape corridor with mountains and water, historic site, normal 2
green space
No Other areas with potential of high historic culture value 1
Water system influence
High Distance to main rivers, reservoirs, or lakes is < 250m 3
Moderate Distance to main rivers, reservoirs, or lakes is ≥ 250m, but < 750m 2
No Distance to main rivers, reservoirs, or lakes is ≥ 750m 1
Noise influence
High Areas with much noise pollution 3
Moderate Areas with not much noise pollution 2
No Areas with little noise pollution 1
Existing land use
High Existing green spaces 3
Moderate Vacant areas and government reserved areas 2
No Other important land uses 1
From the above table we can see that this suitability analysis of green space system is primarily based
on the point of ecological benefits, because the goal of this green space system is to improve the air
quality and surface water quality, prevent flood (water system influence), and eliminate noise (noise
influence). As such, green spaces should be best located in the areas with low air quality, low surface
water quality and with much noise pollution, and the distance to water system should be best within
250 metres to prevent the flood. This analysis also considers social factors including landscape qual-
ity, historic culture value and existing land use. In order to improve the urban landscape quality, green
spaces can be used to plan some areas of high landscape quality in the urban environment, such as
Landscape area, garden scene area or reservoir scene area. Likewise, tour area, reservoir conservation,
53
old city relic, cultural relic and historic site conservation will be planned to improve the historic cul-
ture value by using the green spaces. Note that green spaces cannot be expanded at the expense of
other important land uses (e.g. commercial areas, municipal utilities areas, or public facilities areas).
Besides the existing green spaces, the new green spaces should be best located in those vacant areas
and government reserved areas.
3 3
2 2
1 1
Map 4.2 Air (air quality) Map 4.3 Lscape (landscape quality)
3 3
2 2
1 1
Map 4.4 Swater (surface water quality) Map 4.5 History (historic culture value)
54
3 3
2 2
1 1
Map 4.6 Noise (noise influence) Map 4.7 Luse (existing land use)
Map 4.8 Wsystem (water system influence)
2. Certainty factor
As mentioned in section 3.4.2, certainty factor is used to improve the GIS-based traditional suitability
analysis model. In order to make the following explanation more clear, here we want to give a name to
each single factor map: Air (air quality), Lscape (landscape quality), Swater (surface water quality),
History (historic culture value), Wsystem (water system influence), Noise (noise influence), Luse (ex-
isting land use). Air, Swater and Noise these three factors are raw data (measured in practice), which
won’t be changed with any man-made elements. So it is not necessary to give them the certainty fac-
tors. Lscape, History and Luse are qualitative data, which can get the certainty factor arranging from
‘0.2’ to ‘0.8’. For example, as to the landscape quality, cuneal ecological green space and urban
square belong to moderate suitability class and get the same score of ‘2’. But the cuneal ecological
green space is more suitable for the green space development than the urban square, then we can give
a certainty factor of ‘0.8’ to the cuneal ecological green space and ‘0.6’ to the urban square, in order
to identify the difference between them. Thus the certainty factors for Lscape are coming out (Map
55
4.9). The certainty factors for History and Luse are determined in the same way (Map 4.10 and Map
4.11). Wsystem is quantitative data and its certainty factors (Map 4.12) can be determined by the fol-
lowing certainty factor function (like Formula 3.3):
x−0
CF ( x) = 1 − (0m ≤ x < 250m)
250
x − 250
CF ( x) = 1 − (250m ≤ x < 750m)
500
x − 750
CF ( x) = 1 − (750m < x ≤ 6110m)
5360
( x is the distance to main rivers, reservoirs, or lakes)
0.8 0.8
0.7 0.7
0.6 0.6
0.4 0.4
0.3 0.3
0.2 0.2
Map 4.9 Cerlscape Map 4.10 Cerhistory

(certainty factors for landscape quality) (certainty factors for historic culture value)
1.00
0.80
0.60
0.40
0.8
0.20
0.7
0.00
0.6
0.4
0.3
0.2
Map 4.11 Cerluse Map 4.12 Cerwsystem

(certainty factors for existing land use) (certainty factors for water system influence)
56
Now we can get the composite certainty factors (Map 4.13 to Map 4.16) for the above four factors
according to Formula 3.4. The detailed formulas used are presented below:
Clscape = Lscape − 1 + Cerlscape

Chistory = History − 1 + Cerhistory
Cluse = Luse − 1 + Cerluse
Cwsystem = Wsystem − 1 + Cerwsystem
2.8 2.8
2.3 2.3
1.8 1.8
1.2 1.2
0.7 0.7
0.2 0.2
Map 4.13 Clscape (composite certainty Map 4.14 Chistory (composite certainty
factors for Landscape quality) factors for historic culture value)
3.00
2.46
1.91
1.37
2.8 0.82
2.3 0.28
1.8
1.2
0.7
0.2
Map 4.15 Cluse (composite certainty Map 4.16 Cwsystem (composite certainty
factors for existing land use factors for water system influence)
57
As such, Lscape, History, Luse and Wsystem have got their respective certainty factors and have been
transferred into Clscape, Chistory, Cluse and Cwsystem. Integrated with other three suitability factors
including Air, Swater and Noise, the above four factors will be used to generate ‘final suitability sce-
nario 4’, ‘final suitability scenario 5’ and ‘final suitability scenario 6’ (see section 4.3.5), based on the
improved traditional suitability analysis model (ITSAM) (see section 3.4.2).
4.3.4. Weighting
Weighting in suitability analysis refers to assigning a weight to each factor in order to recognize its
relative importance (Pease, Coughlin et al. 1996). Just like the scores, weights also have a major im-
pact effect on the final results (composite scores). As it has mentioned in section 3.8, this research
wants to make use of three methods to calculate the weights for suitability analysis. Weights for gen-
erating ‘suitability scenario 1’ and ‘suitability scenario 2’ are calculated by the method of statistic in-
tegration (see section 3.5.1). Hierarchic analysis of nine-degree (see section 3.5.2) is used to calculate
the weights for ‘suitability scenario 2’and ‘suitability scenario 5’. And the weights for ‘suitability
scenario 3’ and ‘suitability scenario 6’ are calculated by hierarchic analysis of three-degree (see sec-
tion 3.5.3).
1. Calculating weights by statistic integration

As mentioned in section 3.5.1, the method of statistic integration is a kind of special ‘expert-assess’
method. Here the ‘expert-assess’ is based on the stakeholder analysis. Stakeholder analysis is an im-
portant element to select the suitability factors (see section 4.3.1). Also, it is the foundation for calcu-
lating the weights of the suitability factors. In order to use the method of statistic integration to calcu-
late the factors weights, twenty-four stakeholders including six urban planners, five environmentalists,
ten local residents and three local government officials were invited to determine the importance order
among the seven suitability factors. As such, twenty-four investigating tables (like Table 3.2) were
sent to these stakeholders to show their opinions about the importance order. The investigating infor-
mation of the stakeholders has been aggregated by statistic induction in Table 4.2.
Table 4.2 Investigating information by statistic induction

First Second Third Fourth Fifth Sixth Last
Factor impor- impor- impor- impor- impor- impor- impor-
tance tance tance tance tance tance tance
Air quality 25% 25% 0 8.3% 16.7% 16.7% 8.3%
Landscape quality 33.3% 8.3% 25% 8.3% 8.3% 16.7% 0
Surface water quality 8.3% 25% 25% 16.7% 8.3% 16.7% 0
Historic culture value 0 0 8.3% 16.7% 16.7% 25% 33.3%
Water system influence 8.3% 8.3% 33.3% 16.7% 16.7% 8.3% 8.3%
Noise influence 0 16.7% 0 0 33.3% 8.3% 41.7%
Existing land use 25% 8.3% 16.7% 33.3% 0 8.3% 8.3%
From Table 4.2 we can see that, as to the factor of air quality, six stakeholders acknowledged that it
was the first importance for the suitability analysis of the urban green space system, namely 25% of
58
the twenty-four stakeholders showed this opinion. Second importance with 25%, fourth importance
with 8.3%, fifth importance with 16.7%, sixth importance also with 16.7%, last importance with
8.3%, while none of the stakeholders thought air quality was the third importance for this suitability
analysis. Likewise, other factors importance has been aggregated in Table 4.2 in the same way.
Then according to the following formula (like formula 3.5), the preliminary weights of the above
seven suitability factors are presented below:
T
0.250 0250 0 0.083 0.167 0.167 0.083 7
0.333 0.083 0.250 0.083 0.083 0.167 0 6
0.083 0.250 0.250 0.167 0.083 0.167 0 5
W = [ D • CT ]T = 0 0 0.083 0.167 0.167 0.250 0.333 • 4 = (4.500,4.994,4.582,2.417,2.830,2.584,4.664)
0.083 0.083 0.333 0.167 0.167 0.083 0.083 3
0 0.167 0 0 0.333 0.083 0.417 2
0.250 0.083 0.167 0.333 0 0.083 0.083 1
After the normalization, the final weights W = (0.169,0.188,0.172,0.091,0.107,0.097,0.176) calcu-

lated by statistic integration will be presented in Table 4.6.
2. Calculating weights by hierarchic analysis of nine-degree

The method of hierarchic analysis of nine-degree is a kind of analytic hierarchy process (AHP). It is
based on factors that are measured on a ratio scale. By using this hierarchic analysis of nine-degree,
the user first has to make a comparison for every pair of factors: first qualitative and then quantitative
on a degree from ‘1’ to ‘9’ to make the method operational (see Table 3.5). As such, the original four-
teen stakeholders including six urban planners, five environmentalists and three local government of-
ficials were invited to make the pairwise comparison according to the nine-degree scale. Those ten
local residents were excluded in this invitation just because of their weak sense, which would make it
difficult for them to use the nine-degree scale to make the pairwise comparison among the seven suit-
ability factors. The method then creates a structural judgment matrix based on the pairwise compari-
son judgments of these fourteen stakeholders (Table 4.3).
Table 4.3 Structural judgment matrix of suitability factors by nine-degree

Factor K1 K2 K3 K4 K5 K6 K7
Air quality K1 1 1/4 1/2 1/3 2 1/5 1/6
Landscape quality K2 4 1 3 2 5 1/2 1/3
Surface water quality K3 2 1/3 1 1/2 3 1/4 1/5
Historic culture value K4 3 1/2 2 1 4 1/3 1/4
Water system influence K5 1/2 1/5 1/3 1/4 1 1/6 1/7
Noise influence K6 5 2 4 3 6 1 1/2
Existing land use K7 6 3 5 4 7 2 1
The values in the above matrix were determined by using the rule of ‘the minority should obey the
majority’. For example, there were three different opinions when comparing the factor of historic cul-
ture value with the factor of air quality in this suitability analysis. Six urban planners stated that his-
toric culture value was moderately more important than air quality (the value should be ‘3’ in the
59
judgment matrix according to the nine-degree scale), one environmentalist and three local government
officials acknowledged that historic culture value was strongly more important than air quality (the
value should be ‘5’), and four environmentalists accepted that air quality was moderately more impor-
tant historic culture value (the value should be ‘1/3’). Therefore, the value comparing historic culture
value with air quality was ‘3’ in the judgment matrix according to the above rule, because most of the
stakeholders (six urban planners) supported the first opinion. Likewise, other values in the judgment
matrix were determined in the same way.
Based on this judgment matrix, we can use MATLAB (a mathematical software) to calculate its ei-
genvector (W ) of the largest eigenvalue. Here, the original eigenvector
W = (0.094,0.333,0.1420,0.2177,0.0655,0.5041,0.7444) . After the normalization, this original
eigenvector is converted to the final vector, W = (0.045,0.159,0.068,0.104,0.031,0.240,0.354) .
The values in this final vector represent the weights of those seven suitability factors, which will be
presented in Table 4.6.
3. Calculating weights by hierarchic analysis of three-degree

Hierarchic analysis of nine-degree is particularly designed to calculate weights within a hierarchic
structure of the factors. Due to the fast growing number of pairwise comparisons, however, it is not
sensible to use this method for a large number of factors. Within hierarchic analysis of nine-degree,
the number of comparisons increases with the number of effects (n*(n-1)/2). For example, in order to
calculate the weights of those seven suitability factors, it needs to make twenty-one comparisons in
the structural judgement matrix. Moreover, it needs to make a judgement among those nine degrees
(see Table 3.5), which will often make the user rather puzzled.
Hierarchic analysis of three-degree is also a kind of analytic hierarchy process (AHP). But in this
method, it involves no more than three degrees when making a pairwise comparison: less important,
same important and more important (see Formula 3.6). Thus it can be showed that three-degree is eas-
ier for the user to make a judgement than nine-degree. In order to determine the final values in the
comparison matrix (Table 4.4) based on the three-degree scale, only the original six urban planners
were invited to make the pairwise comparisons among the seven suitability factors, from their profes-
sion perspective of urban planning. Just like hierarchic analysis of nine-degree, this method also
adopts the rule of ‘the minority should obey the majority’. As such, when comparing the factor of air
quality with the factor of landscape quality, four urban planners stated that air quality is less important
than landscape quality for this suitability analysis (the value should be ‘0’ in the comparison matrix).
While other two planners thought it was the other way around (the value should be ‘2’). According to
the above rule, the value between air quality and landscape quality was ‘0’ in the comparison matrix.
Likewise, other comparisons’ values were determined in the same way (Table 4.4).
60
Table 4.4 Comparison matrix of suitability factors by three-degree

7
Factor K1 K2 K3 K4 K5 K6 K7
Ki
i =1
Air quality K1 1 0 1 0 2 0 0 4
Landscape quality K2 2 1 2 2 2 0 0 9
Surface water quality K3 1 0 1 0 2 0 0 4
Historic culture value K4 2 0 2 1 2 0 0 7
Water system influence K5 0 0 0 0 1 0 0 1
Noise influence K6 2 2 2 2 2 1 0 11
Existing land use K7 2 2 2 2 2 2 1 13
Based on the above comparison matrix, the structural judgment matrix is generated below according
to Formula 3.7 (see section 3.5.3).
Table 4.5 Structural judgment matrix of suitability factors by three-degree

ri rj
r1 r2 r3 r4 r5 r6 r7
Air quality r1 1 12/77 1 4/17 17/4 12/103 12/259
Landscape quality r2 77/12 1 77/12 19/6 29/3 6/19 3/16
Surface water quality r3 1 12/77 1 4/17 17/4 12/103 4/43
Historic culture value r4 17/4 6/19 17/4 1 15/2 3/16 2/15
Water system influence r5 4/17 3/29 4/17 2/15 1 6/71 1/14
Noise influence r6 103/12 19/6 103/12 16/3 71/6 1 6/19
Existing land use r7 259/12 16/3 43/4 15/2 14 19/6 1
According to the above structural judgment matrix, this method also uses MATLAB to calculate its
eigenvector (W ) of the largest eigenvalue. Here,
W = (0.0480,0.2429,0.0532,0.1340,0.0248,0.4393,0.8511) . After the normalization, this eigen-
vector can represent the weights of the corresponding suitability factors, namely
W = (0.027,0.135,0.030,0.075,0.014,0.245,0.474) , which will be presented in Table 4.6.
4. Weighting results for suitability scenarios

As mentioned in section 3.8, weights calculated by statistic integration are used to generate ‘suitabil-
ity scenario 1’ and ‘suitability scenario 4’. Hierarchic analysis of nine-degree is used to calculate the
weights for ‘suitability scenario 2’and ‘suitability scenario 5’. And the weights for ‘suitability sce-
nario 3’and ‘suitability scenario 6’ are calculated by hierarchic analysis of three-degree. Hu (1994)
and Xu (1999) suggested that the ranking weights as calculated should be normalized for each func-
tion under analysis. The normalized weighting results for the suitability scenarios of the urban green
space system are presented in Table 4.6. These results indicate that landscape quality has the highest
priority among the seven suitability factors according to statistic integration, followed by existing land
use. This result also shows that historic culture value, water system influence and noise influence are
61
much less important factors. But that is not the case according to the other two weighting methods:
hierarchic analysis of nine-degree and three-degree. Therefore, weight is a very important element in
the suitability analysis. Different scores and different weights will generate different suitability sce-
narios.
Table 4.6 Weighting results for suitability scenarios

Air Landscape Surface Historic Water Noise Existing
Factor quality quality water culture system influence land
quality value influence use
Scenario 1,
Scenario 4 0.169 0.188 0.172 0.091 0.107 0.097 0.176
Scenario 2,
Weight Scenario 5 0.045 0.159 0.068 0.104 0.031 0.240 0.354
Scenario 3,
Scenario 6 0.027 0.135 0.030 0.075 0.014 0.245 0.474
Note: Weights calculated by statistic integration. Weights calculated by hierarchic analysis of

nine-degree. Weights calculated by hierarchic analysis of three-degree.
4.3.5. Suitability scenario

According to the ratio values (scores) for all suitability classes, a total of seven GIS data coverages
are created, one for each of the suitability factors identified above (Map 4.2 to Map 4.8). These cover-
ages are based on the individual suitability values for each factor. In addition, weights for the seven
factors and certainty factors have been determined, so these seven single factor maps can be combined
into a composite suitability map by addition of the suitability scores with a weighting system. That
means we can use a GIS overlay procedure to get a suitability evaluation of all the seven factors ac-
cording to Formula 3.2.
As mentioned in section 3.8, six suitability scenarios will be generated according to different scores,
weights and certainty factors. Here we give the names of ‘Scenario 1’, ‘Scenario 2’, ‘Scenario 3’,
‘Scenario 4’, ‘Scenario 5’ and ‘Scenario 6’ to these six suitability scenarios. The frontal three scenar-
ios are generated based on the traditional suitability analysis model (TSAM), while the latter three
scenarios are based on the improved traditional suitability analysis model (ITSAM) that they need to
take advantage of the certainty factors from section 4.3.3. The detailed calculation formulas are pre-
sented below (Wai represents the weight calculated by statistic integration; Wbi represents the weight
calculated by hierarchic analysis of nine-degree; Wci represents the weight calculated by hierarchic
analysis of three- degree):
Scenario 1=Air*Wa1+Lsapce*Wa2+History*Wa3+Swater*Wa4+Wsystem*Wa5+Noise*Wa6+Luse*Wa7
Scenario 2=Air *Wb1+Lsapce*Wb2+History*Wb3+Swater*Wb4+Wsystem*Wb5+Noise*Wb6+Luse*Wb7
Scenario 3=Air *Wc1+Lsapce*Wc2+History*Wc3+Swater*Wc4+Wsystem*Wc5+Noise*Wc6+Luse*Wc7

62
Scenario 4=Air *Wa1+Clsapce*Wa2+Chistory*Wa3+Swater*Wa4+Cwsystem*Wa5+Noise*Wa6+Cluse*Wa7
Scenario 5=Air *Wb1+Clsapce*Wb2+Chistory*Wb3+Swater*Wb4+Cwsystem*Wb5+Noise*Wb6+Cluse*Wb7
Scenario 6=Air *Wc1+Clsapce*Wc2+Chistory*Wc3+Swater*Wc4+Cwsystem*Wc5+Noise*Wc6+Cluse*Wc7
As such, the composite suitability score for each scenario can be derived based on the above formulas.
Ideally, this composite suitability score ranges from ‘1’ to ‘3’. But that is not the case in these six suit-
ability scenarios (e.g. the composite suitability score of ‘Draft suitability scenario 1’ ranges from
‘1.169’ to ‘2.818’). The distribution of these composite scores seems not very regular in the draft suit-
ability scenarios (e.g. Map 4.17). It is recommended the composite suitability score should be grouped
in classes to help clarify the data and to make the resulting scenarios more decipherable and easier to
interpret. Thus the final suitability scenarios (Map 4.18 to Map 4.23) can be grouped in three levels
according to the following suitability intervals (Vs represents the composite suitability score), by us-
ing the histogram of the composite suitability score in ILWIS:
High − suitability : 2 ≤ Vs ≤ 2.818 High − suitability : 2 ≤ Vs ≤ 2.795

Scenario1 Moderate − suitability : 1.812 ≤ Vs < 2 Scenario 2 Moderate − suitability : 1.646 ≤ Vs < 2
No − suitability : 1.169 ≤ Vs < 1.812 No − suitability : 1.046 ≤ Vs < 1.646

Scenario3 Moderate − suitability : 1.526 ≤ Vs < 2 Scenario 4 Moderate − suitability : 1.612 ≤ Vs < 2

Scenario5 Moderate − suitability : 1.464 ≤ Vs < 2 Scenario6 Moderate − suitability : 1.326 ≤ Vs < 2
2.818
2.488
2.158
1.828
1.499
1.169
Map 4.17 Draft suitability scenario 1
63
no suitability no suitability
moderate suitability moderate suitability
high suitability high suitability
Map 4.18 Final suitability scenario 1 Map 4.19 Final suitability scenario 2
64
4.4. Multi-criteria analysis for ranking

In order to help in the development of more accurate and reliable resource management plans, many
software systems have been developed. This will allow an easier planning of the resources at hand by
using automated tools. It will also give some flexibility to the plan in a way that it may be easily
modified and improved. Meanwhile it will look for the optimal solution for the use of a conflicting
resource. The flexibility of the programs is a must in today’s world because of the changing character-
istics of the conflicts.
In this research, DEFINITE, a decision support system (DSS) software, will be used to make a ranking
of the above six suitability scenarios and sensitivity analysis. DEFINITE is designed to support deci-
sion making on a finite set of alternatives. Actually it is a whole toolbox of methods that can be used
on a wide variety of problems. If you have a problem to solve, and you can identify alternative solu-
tions, then DEFINITE can weigh up the alternatives for you and assess the most reasonable. The pro-
gram contains a number of methods for supporting problem definition as well as graphical methods to
support representation. To be able to deal with all types of information, DEFINITE includes multicri-
teria methods, cost-benefit analysis and graphical evaluation methods. Related procedures such as
weight assessment, standardization, discounting and a large variety of methods for sensitivity analysis
are also available. A unique feature of DEFINITE is a procedure that systematically leads an expert
through a number of rounds of an interactive assessment session and uses an optimisation approach to
integrate all information provided by the experts to a full set of value functions (Janssen, Herwijnen et
al. 2001).
4.4.1. Effects table

In section 4.3.5, each suitability scenario has been grouped in three classes: ‘high suitability’, ‘moder-
ate suitability’ and ‘no suitability’. Actually these three classes can be regarded as three effects (in
DEFINITE, effects are equivalent to criteria) to make a ranking among the six suitability scenarios.
And the areas of these three classes can be regarded as the effects scores (criteria scores). Thus an
effects table is generated in DEFINITE (Figure 4.2).
Figure 4.2 Effects table
65
The effects table above includes six alternatives (from ‘suitability scenario 1’ to ‘suitability scenario
6’), three effects (‘high suitability’, ‘moderate suitability’, ‘no suitability’) and the effects scores
(suitability areas). These three effects are measured by the same unit, ‘hectare’. The areas of ‘high
suitability’ and ‘moderate suitability’ are what we need in the green space system, which mean the
larger the areas, the higher the suitability. So the effects of ‘high suitability’ and ‘moderate suitability’
belong to the benefit effect. The areas of ‘no suitability’ are the least important in the green space sys-
tem, so it is a cost effect that means the larger the areas, the lower the suitability.
4.4.2. Standardization
Making the criteria scores comparable is often called standardization or normalization. Through a
standardization procedure the measurement units can be made uniform, and the scores lose their di-
mension along with their measurement units (Sharifi and Herwijnen 2003). There are many methods
available to standardize the criteria scores till now. The method to use should depend on the problem
character and the attributes character. In this research, three methods are used to standardize the three
suitability criteria (effects).
1. Goal standardization for ‘high suitability’

Goal standardization scales the scores for each criterion according to the relative distance between a
goal value and a minimum value (Formula 4.1). These two values should be determined by decision
makers based on the practical requirements. The total area of Dongguan municipality is ‘22758’ hec-
tare. In order to make a good environmental quality, the greening rate should range from ‘30%~40%’
(see section 4.2.4). So the goal value of green spaces areas for ‘high suitability’ is
‘22758*40%=9103.2’ hectare, and the minimum value is ‘22758*30%=6827.4’ hectare. Figure 4.3
shows the original scores and the matching standardized values. The result of this standardization is
positive scores for the benefits effects. The advantage is that the standardized values have a clear, real
meaning independent of the evaluation alternatives.
1
0.9
0.8 − xij − v min
0.7
xij =
v g − v min
Standardized
0.6
Formula 4.1 Goal standardization
0.5
−
0.4 (Where xij is the standardized score, xij is
0.3
the original score, v min is the minimal value,
0.2
v g is the goal value.)
0.1 scenario
scenario
scenario
6 352 scenario
scenario41
0
4,000 6,000 8,000 10,000
high suitability [hectare]
Figure 4.3 Standardization for ‘high suitability’
66
2. Interval standardization for ‘moderate suitability’

Interval standardization is somewhat similar to the goal standardization, yet instead of the goal and
minimum values the highest score and the lowest score are specified (Formula 4.2). It scales the crite-
rion scores according to their relative position on the interval between the lowest and highest scores.
Figure 4.4 shows the graph representing the conversion of the original scores to the standardized
scores. In this benefit effect, the absolute highest score of ‘scenario 3’ is indicated with a ‘1’, and the
lowest one of ‘scenario 4’ with a ‘0’. The scores of other scenarios in between are scaled proportion-
ally. For this benefit effect the graph goes from low to high, in other words a low score gets a low-
standardized value.
scenario 3
1
0.9 scenario 2
0.8 − xij − min xij

j
xij =
0.7 max xij − min xij
j j
Standardized
0.6
scenario 6 Formula 4.2 Interval standardization
0.5
−
0.4 scenario 5 (Where xij is the standardized score, xij is
0.3 scenario 1 the original score, min xij is the lowest
j
0.2
score, max xij is the highest score.)
0.1scenario 4 j
0
4,000 5,000 6,000 7,000 8,000
moderate suitability [hectare]
Figure 4.4 Standardization for ‘moderate suitability’
3. Maximum standardization for ‘no suitability’

Maximum standardization scales the scores for each criterion according to the relative distance be-
tween the original and the highest score (Formula 4.3). Figure 4.5 shows the scores are standardized
with a linear function. ‘no suitability’ is a cost effect, so the absolute highest score is indicated with a
‘0’. Other standardized values are determined proportionally. On the X-axis are the original scores,
from the lowest score of ‘scenario 3’ to the highest score of ‘scenario 5’. On the Y-axis are the stan-
dardized scores. For this cost effect the graph grows from high to low, namely a low score gets a high-
standardized value.
67
1
0.9
0.8
− xij
0.7 xij = 1 −
max xij
Standardized
j
0.6
0.5 Formula 4.3 Maximum standardization
−
0.4
(Where xij is the standardized score, xij is
0.3 scenario 3
scenario
scenario 12 the original score, max xij is the highest
0.2 scenario 4 j
0.1 scenario 65 score.)

0
0 5,000 10,000
no suitability [hectare]
Figure 4.5 Standardization for ‘no suitability’
4.4.3. Weight
Assigning weights to effects (criteria) is often a complicate task (Janssen, Herwijnen et al. 2001).
DEFINITE includes various methods to support this task: Direct assessment, Pairwise comparison,
Expected value method, Random weights, Extreme weights. From the effects table (Figure 4.2) we
can easily rank the effects according to their relative importance: ‘high suitability’ > ‘moderate suit-
ability’ > ‘no suitability’. So Expected value method is used to determine the ranking of the effects.
Expected value method assumes each set of weights that fits the ranking order of effects has equal
probability. The weight vector is calculated as the expected value of the feasible set. Therefore, it can
simulate the average idea of a group (Sharifi and Herwijnen 2003). Expected value method calculates
the weight, wi , for effect i according to Formula 4.4. Where n is the number of effects. The weights
fit the ranking order of effects, meaning that w1 ≥ w2 ≥ ≥ wn ≥ 0 .
n +1− i 1
wi = k =1
n(n + 1 − k )
Formula 4.4 Expected value method
The final weights for those three effects including ‘high suitability’, ‘moderate suitability’ and ‘no
suitability’ are presented in Figure 4.6.
68
Figure 4.6 Standardizations and weights
4.4.4. Ranking
1. Weighted summation
The standardized scores and weights have been determined, now we can use weighted summation to
make a ranking of alternatives (from scenarios 1 to scenario 6) according to Formula 3.8.
Figure 4.7 Ranking results I by weighted summation
69
Above we can see the ranking results represented graphically. Note that ‘scenario 3’ is the best alter-
native, followed by ‘scenario 2’, ‘scenario 6’, ‘scenario 5’, ‘scenario 1’, and ‘scenario 4’. In order to
explain the results with more details, we can display the weight as a pie graph next to this bar graph.
And a stacked bar graph can be used to view the contributions of the effects scores to the total scores.
The stacked bar divides the bar of every alternative into the relative contributions of the effects. Those
contributions depend on the size of the effect and its weight (Figure 4.8).
Figure 4.8 Ranking results II by weighted summation
The stacked bar above clearly shows that the effect of ‘moderate suitability’ is the main cause for
‘scenario 3’ to be the best alternative. Over half of the final score of ‘scenario 3’ derives from the con-
tribution of ‘moderate suitability’. The next contribution originates from the effect of ‘no suitability’,
while the effect of ‘high suitability’ only makes a few contributions. Furthermore, the pie graph shows
that the influence of ‘moderate suitability’ on the ranking of alternatives is limited because of the low
weight assigned to this effect. After that, a scatter diagram can be used to have a closer look at the two
effects that contribute most to the total scores (Figure 4.9).
70
Figure 4.9 Ranking results III by weighted summation (Scatter diagram)
The diagram above shows the standardized scores of the alternatives for the effects ‘no suitability’
and ‘moderate suitability’. The best ideal alternative would have a score of ‘1’ for both effects and
would be found in the top right corner of the diagram. The alternative located closest to this ideal is
the best. And we can find the best alternative by moving the green line from the upper right corner to
the lower left. The first alternative crossing this line is the best. Thus it can be showed that ‘scenario
3’ is the best alternative, followed by ‘scenario 2’ and ‘scenario 6’. The assumption made here is that
the ranking of the alternative with equal weights for ‘no suitability’ and ‘moderate suitability’. Chang-
ing the weights of the two effects may move a different alternative to the first or other positions. For
example, if the weight of ‘no suitability’ is increased to ‘0.8’, ‘scenario3’ and ‘scenario 2’ are also the
two better alternatives, but they are followed by ‘scenario 1’ instead of ‘scenario 6’. This proves that
‘scenario 3’ is still the best one although the weights of effects have been changed.
2. Electre method
Based on the effects table (Figure 4.2) and the standardized effects table (Table 4.7), electre method
can generate a ranking of alternatives through a large number of intermediate results, which include
concordance table, discordance table, the strong and the weak results.
71
Table 4.7 Standardized effects table

scenario 1 scenario 2 scenario 3 scenario 4 scenario 5 scenario 6 weight
high 0 0 0 0 0 0 0.611
suitability
moderate 0.21 0.81 1 0 0.32 0.47 0.278
suitability
no 0.17 0.16 0.20 0.09 0 0.01 0.111
suitability
(1) Concordance table

The first intermediate result is the concordance table showed in Figure 4.10. According to Formula
3.9, the value ‘0.72’ in the concordance table for the alternatives ‘scenario 1’ and ‘scenario 2’ is cal-
culated as the sum of the weights of the effects for which ‘scenario 1’ is not worse than ‘scenario 2’.
Here it means that, the original score of ‘6528.36’ of ‘high suitability’ in ‘scenario 1’ is higher than
the score of ‘3766.80’ in ‘scenario 2’, and original score of ‘11482.16’ of ‘no suitability’ in ‘scenario
1’ is higher than the score of ‘11620.18’ in ‘scenario 2’, so this concordance index
conc( j , k ) = 0.611 + 0.111 = 0.722 . Other concordance indices in the concordance table are gener-
ated in the same way. In this case the effect ‘high suitability’ with weight ‘0.611’, the effect ‘moderate
suitability’ with weight ‘0.278’ and the effect ‘no suitability’ with weight ‘0.111’ (Table 4.7).
Figure 4.10 Concordance table
Based on the above concordance table, the decision maker has to determine a strong threshold Ps and
a weak threshold Pw for this table. Here the strong concordance threshold value is specified as
‘0.886’ and the weak concordance threshold value is specified as ‘0.725’.
72
(2) Discordance table

The second intermediate result is the discordance table showed in Figure 4.11. According to Formula
3.10, the value of ‘0.61’ for both alternatives, ‘scenario 1’ and ‘scenario 2’, is defined as the largest
deference of all effects in which ‘scenario 1’ performs worse than ‘scenario 2’. In this example it
means that, only the original score of ‘4747.48’ of ‘moderate suitability’ in ‘scenario 1’ is lower than
the score of ‘7371.08’ in ‘scenario 2’, so the discordance index disc ( j , k ) = 0.81 − 0.21 = 0.60 .
Other discordance indices are generated in the discordance table in the same way.
Figure 4.11 Discordance table
Based on the above discordance table, the decision maker also has to specify a strong threshold Qs
and a weak threshold Qw for this discordance table. Here the strong discordance threshold value is set
at ‘0.225’ and the weak discordance threshold value is set at ‘0.275’.
(3) Strong and weak graph

Two concordance thresholds ( Ps , Pw ) and two discordance thresholds (Qs , Qw ) have been deter-
mined above. Using these thresholds, a strong graph and a weak graph can be made according to the
following strong and weak rules:
Strong graph:
j is strongly preferred to k ⇔ (con( j , k ) > Ps ) and (disc( j , k ) < Qs )
Weak graph:
j is weakly preferred to k ⇔ (con( j , k ) > Pw ) and (disc( j , k ) < Qw )
73
Here the strong threshold Ps for the concordance index is specified as ‘0.886’ and the strong thresh-
old Qs for the discordance index is specified as ‘0.225’. This means the alternative ‘scenario 1’ is
better than ‘scenario 4’, due to the value of ‘1.00’ comparing ‘scenario 1’ with ‘scenario 4’ in concor-
dance table is higher than the strong threshold Ps of ‘0.886’, and the value of ‘0.00’ in discordance
table is lower than the strong threshold Qs of ‘0.225’. This is shown with a ‘1’ in the strong graph
(Figure 4.12). The strong graph is generated composed of the value of ‘1’ (representing a ranking) and
‘0’ (representing no ranking). The weak graph showed in Figure 4.13 can be made in the same way
according to the above weak rule. Here the weak threshold values for the concordance index and dis-
cordance index ( Pw and Q w ) are respectively set at ‘0.725’ and ‘0.275’.
Figure 4.12 Strong graph (0: no ranking, 1: a ranking)
Figure 4.13 Weak graph (0: no ranking, 1: a ranking)
74
(4) Ranking results

In the final step, the circuits of the strong and weak graph are reduced. Then those alternatives of the
strong graph that are strongly preferred to other alternatives are selected. From these selected alterna-
tives a second selection is carried out based on the weak graph: the alternatives that are weakly pre-
ferred to other alternatives. These alternatives are given rank number one. The alternatives already
ranked are then deleted from the two graphs and the selection of alternatives starts again until the two
graphs are examined totally. Finally a complete ranking showed in Figure 4.14 is generated by com-
bining two different complete pre-orders: one obtained by first selecting the alternatives that are not
outranked by any other alternative and then going down the graph; the other obtained by first selecting
the alternatives that are outranked by other alternatives and then going up the graph.
Figure 4.14 Ranking results by electre method
The ranking results by electre method above show the same order of alternatives as that by weighted
summation (Figure 4.7). This means ‘scenario 3’ is the best alternative, followed by ‘scenario 2’,
‘scenario 6’, ‘scenario 5’, ‘scenario 1’, and ‘scenario 4’. Thus the alternatives have been ranked. Of
course not all information is completely certain. The sensitivity of the ranking for uncertainties in
scores and weights can be analysed in the next section ‘Sensitivity analysis’.
4.5. Sensitivity analysis

Multicriteria analysis (MCA) is mainly composed of four uncertainty components: criteria, criteria
scores, weights and the choice of an evaluation method (see section 3.7). This section systematically
focuses on the sensitivity of the ranking to uncertainties in scores and weights.
75
4.5.1. Uncertainty of one score

In this section a procedure is described to estimate certainty intervals for criterion scores. Within a
certainty interval the ranking of two alternatives is not sensitive to changes in criterion scores. Figure
4.15 shows the sensitivity of the ranking for ‘moderate suitability’ of ‘scenario 1’. The ‘moderate
suitability’ areas of ‘scenario 1’ are estimated at ‘4747.48’ hectares (see Figure 4.2). If this estimate is
correct ‘scenario 3’ is the best alternative. The question whether ‘scenario 3’ is still the best if ‘mod-
erate suitability’ areas of ‘scenario 1’ prove to be higher can be answered using this figure. The origi-
nal estimate, a ‘moderate suitability’ of ‘4747.48’ hectares, is marked with a vertical line. At
‘4747.48’ hectares the total score of ‘scenario 3’ is above the total score of ‘scenario 2’, ‘scenario 6’
and other scenarios. It is clear that the difference between ‘scenario 3’ and ‘scenario 1’ is somewhat
large but that a certain increase in the ‘moderate suitability’ areas of ‘scenario 1’ will make ‘scenario
1’ the better alternative. Reversal occurs at a ‘moderate suitability’ by ‘scenario 1’ of about ‘8200’
hectares. This ranking is shown to be not very sensitive to changes in this score. The sensitivity of the
ranking to changes in one weight can be analysed in the same way. But in calculating certainty inter-
vals for one weight it is assumed that all weights add up to ‘1’ and that the ratios between all other
weights remain unaltered.
Figure 4.15 Sensitivity of the ranking for changes in one score
4.5.2. Overall uncertainty of the weights

As mentioned in section 3.7.1, the sensitivities of rankings of alternatives to overall uncertainty in
scores and weights are analysed by using a Monte Carlo approach. The decision maker is asked to
estimate the maximum percentage that actual values may differ from the values included in the ele-
ments of the effects table or set of weights. Figure 4.16 shows a result of sensitivity analysis on the
ranking of alternatives by looking at the overall uncertainty of the weights. It is the probability that an
alternative is ranked first, second, etc. Here it assumes that the weights can be ‘50%’ higher or lower
76
than the assigned weights and that this deviation is normally distributed. The figure shows that the
alternatives ‘scenario 3’, ‘scenario 2’ and ‘scenario 6’ are always ranked on the first three positions,
while ‘scenario 5’, ‘scenario 1’ and ‘scenario 4’ almost rank the last three positions. Numbers associ-
ated with this figure can also be derived. For example, ‘scenario 3’ and ‘scenario 2’ are found ‘100%’
on first and second positions, while ‘scenario 4’ ‘100%’ on last position. It can be seen that ‘scenario
6’ has the highest probability of being ranked third, but that ‘scenario 5’ even ‘scenario 1’ can rank
third. The large-sized circles on the main diagonal indicate that the ranking of the alternatives under
‘50%’ weight uncertainty is relatively stable. Therefore it is quite certain that ‘scenario 3’ is the best
alternative, followed by ‘scenario 2’ and other scenarios.
Figure 4.16 Uncertainty analysis on the weights (50%)
4.5.3. Changes in all weights (rank reversal of two alternatives)

As mentioned in section 3.7.2, three methods can be used to test the sensitivity of ranking to changes
in all weights. The method of rank reversal of two alternatives is described in this section. Figure 4.17
shows the result of a rank reversal of two alternatives ‘scenario 1’ and ‘scenario 2’, the second posi-
tion. In the upper part of this window, the column on the left shows the original weights. The column
on the right shows the weight combination most similar to the original weights for which the alterna-
tives ‘scenario 1’ and ‘scenario 2’ change the rank order. The ranking of the alternatives belonging to
this weight combination is found in the lower part of this window in the column on the right, while
the original ranking results using weighted summation are presented in the column on the left. It is
obvious that the largest changes in weights decreasing the weight of ‘moderate suitability’ and in-
creasing the weight of ‘no suitability’ may result in a better result of alternative ‘scenario 3’ (Note
that the order of importance of the criteria has been changed). From all above sensitivity analyses of
ranking for uncertainties in scores and weights, it can be concluded that ‘scenario 3’ is the best alter-
native although there are some uncertainties existing in the multicriteria analysis.
77
Figure 4.17 Weight combination by rank reversal between ‘scenario 1’ and ‘scenario 2’
4.6. Comparison
The best suitability scenario has come out through the ranking and sensitivity analysis, now we can
compare it with the urban master plan to find the commons and differences between them, and to find
the reasons behind them. Map 4.24 is a master plan of Dongguan municipality (2000-2015) made by
Chinese Academy of Sciences and Dongguan Institute of Urban Planning and Design. Map 4.25 is the
comparison between the best suitability scenario and the master plan. In order to make a better expla-
nation for the comparison, here we give a name of ‘proposed plan’ to the best suitability scenario.
78
Map 4.24
Residential land
Industrial land
Public facilities
Municipal utilities
Green space
Forestry land and farmland
Rural land
Water area
Roads and squares
Planning boundary
Map 4.24 Master plan of Dongguan municipality (2000-2015)
The best suitability scenario (proposed plan) Master plan of Dongguan municipality (2000-2015)
Green space
no suitability Water area
moderate suitability Roads and squares
high suitability Other landuses
Planning boundary
Note: Qifeng mountain area

Tongsha reservoir area

Baima area

Rivers intersection area

Industrial area
Map 4.25 Comparison map

79
4.6.1. Commons
(1) Seen from the holistic distribution of the green space system, green spaces are mainly located in
the east and the south in both plans (Map 4.25). There are some famous historic relics existing in
these areas, but so little importance was attached to them that they were not protected well in the past.
The government wants to plan a cultural relic and a historic site conservation to improve the historic
culture value. Moreover, the landscape quality and surface water quality are rather low in the south,
which needs to use green spaces to improve these two kinds of quality. The most important is that the
built up area is primarily situated in the middle part, while most of the vacant areas and government
reserved areas lie in the east and the south. Hence, the green spaces won’t interfere with the future
buildings development and certainly not be expanded at the expense of other important landuses.
(2) Idea of applying large-scale and small-scale green spaces in the green space system is the same in
both plans. The proposed plan and the master plan consider the suitable sites for green space system
should be not only in the built up area but also in the rural area. Large-scale green spaces in the rural
area are the main parts of the green space system. They can play a good ecological role in cleaning air,
preventing and reducing hazard. Also, both plans create more places with high landscape quality and
supply better places for the residents’ recreation to move the large number of urban population out of
the built up area, with the aim of solving this kind of over-urbanization problem. Some small-scale
green spaces decentralize in the built up are, which to some extent can adjust and improve urban cli-
mate, adjust the residents’ psychology, and eliminate the noise caused by the fast growing industry
and traffic. The green space system composed of large-scale and small-scale green spaces is consistent
with the existing urban structure. Many buildings are located in the built up area that make the urban
landuses be less potential. Thus the urban structure of Dongguan municipality has developed into a
radiated style, and the direction for urban development is primarily showed on the development se-
quence.
(3) The green space system in both plans is developed centring around the mountains and water areas.
In the built up area, Wanjiang River, Dongguan Channel and Dongguan Canal are the centres for the
green spaces development, while Qifeng Mountain is the development centre of green spaces in the
east and the three reservoirs are the other centres in the south (Map 4.25). The location of Qifeng
Mountain area (see in Map 4.25) is almost the same. In the proposed plan, this area is almost equal
to the existing green space area. Master plan states that Qifeng Mountain area is not enough for the
fast increasing population any more. It should be expanded at the expense of its surrounding forestry
lands and farmlands. Both plans acknowledge that the slide in this area is potentially dangerous and
the soil is very sensitive to the neighbouring residents’ activities. Some natural resources and historic
relics exist in the Qifeng Mountain area but have been partially destroyed. All of which need to be
improved by planning a suitable green space system. On the other hand, green spaces can play a good
education function (see section 2.3.2) to the residents that they can enjoy in this area but have a re-
sponsibility for preserving the local resources. In summary, the green space system in both plans is
characterised by that urban landuses are tightly integrated with the elegant mountains and water area,
which will finally play an ecological role to the urban environment. The whole city seems to be en-
closed by the green spaces, mountains and water areas, and the urban structure has a development
tendency of cross-axis.
80
4.6.2. Differences
(1) In the proposed plan, the green space system is planned primarily allowing for the ecological fac-
tor (air quality, surface water quality, water system influence, noise influence), while the master plan
mainly takes the social factor into account, particularly the factor of existing land use and historic cul-
ture value. As the research has mentioned the definition of urban green space system in section 2.1, in
the proposed plan, those water areas enabling people to contact the nature and those greenways that
can play a good ecological function are included in this urban green space system. For example, in
Tongsha reservoir area (see in Map 4.25), Tongsha reservoir showing good ecological benefits is
one of the important green spaces in the south. In the east south, Baima area (see in Map 4.25) is
the other ecological green space including some forestry lands and water areas. This is very different
from the opinions in the master plan, which states that the existing water areas, forestry lands and
other important natural resources should be strictly preserved. They cannot be occupied by other lan-
duses but green spaces can be planned around them to play the ecological function. These opinions
are obviously showed in Tongsha reservoir area and Baima area in both plans. As such, the green
space system in the proposed plan is more systematic and decentralized than that in the master plan.
(2) In the rivers intersection area (see in Map 4.25), green spaces in the master plan are mainly
close to Wanjiang River, Dongguan Channel and Dongguan canal, with the aim of improving the sur-
face water quality and acting as a green buffer to prevent the flood. In the proposed plan, only some
small-scale green spaces decentralise around the rivers. The factor of surface water quality and the
factor of water system influence for identifying the suitability are less important than other ecological
and social factors (see the weighting results in Table 4.6). The proposed plan accepts that in the past
few years, many industries of electronic communication and electric machine near the rivers before
have been moved to the areas surrounding the built up area. Thus the water quality in Wanjiang River,
Dongguan Channel anf Dongguan canal has been relatively improved. In addition, the water line of
these three rivers is much lower than the 10-year flood line, so more green spaces should be planned
in other areas that have higher flood potential, show more indications of advance erosion and high
level of human activities (e.g. the three reservoir areas in the south). In summary, both plans acknowl-
edge that the flood should be prevented by green buffers in the future. In the proposed plan, this can
be showed from the areas around the rivers in the north of the built up area not with ‘high suitability’,
but with ‘moderate suitability’.
(3) In the south of the built up area, green spaces in the master plan are mainly on the margin of the
industrial area (see in Map 4.25), as a green belt of absorbing toxic gas, trapping dust and eliminat-
ing noise. In the proposed plan, some green spaces are on the margin of the industrial area, others are
planned at the expense of some industrial areas, in order to play a better ecological role, improve the
landscape quality and adjust the workers psychology. Residents working in the industrial area not
only need to eliminate the industrial pollution as much as possible, but also need an elegant environ-
ment with high landscape quality to adjust their oppressive psychology. This can be realized by plan-
ning the green spaces situated in the industrial area and on its margin. However, the master plan states
that industry is the base of the urban economy. The existing industrial lands are so important that they
cannot be occupied by other landuses. As long as the green spaces are suitably located on the margin
of the industrial area, they can also play a good role in improving the working environment. In addi-
tion, with the fast growing of the urban traffic, noise has become a serious pollution that has a bad
81
effect on the residents’ health. The proposed plan accepts that planning a suitable green space system
is the most effective method to eliminate the noise. This can be showed from the areas with ‘moderate
suitability’ beside the main roads in the proposed plan, while it is neglected in the master plan.
82
5. Conclusion and recommendation
5.1. Conclusion
This research involves two themes, which are urban green space system and suitability analysis. Ur-
ban green space system in this research means that there are some good green areas in the urban spa-
tial environment, which are mainly covered with natural or man-made vegetation and can function as
ecological balance, playing an active role to urban environment, landscape, and residents recreation.
Suitability analysis is the process to determine whether the land resource is suitable for some specific
uses and to determine the suitability level. Combing these two themes to one is the research purpose
to develop an approach of GIS-based suitability analysis to identify suitable sites for urban green
space system development. The conclusions can be addressed as follow:
Suitability analysis can be regarded as a relatively difficult task partially due to large number of fac-
tors and large volume of data that may be required for the determination. Seven suitability factors are
selected in this research, but they are not the optimal ecological factors combination and not enough
to carry out the GIS-based suitability analysis. The research approach for such suitability analysis de-
pends heavily on the data available. Comprehensive information needs to be available for each of the
suitability factors applied in the analysis. This information includes accurate suitability mapping, de-
lineated city boundary, urban planning context, existing situation assessments, greening indices and
future development restrictions. Stakeholder analysis becomes important as a means of selecting the
suitability factors and evaluating the relative importance among them. Qualitative and quantitative
factors information for the certainty factor is also necessary.
Integrating the certainty factor with GIS is an efficient method to improve the traditional suitability
analysis model (TSAM). The improved traditional suitability analysis model (ITSAM) using the cer-
tainty factor can efficiently identify the difference among the different subclasses within the same
suitability class, and can easily reduce the difference between the two different suitability classes
around the division of scores. So ITSAM encompasses more enriched and precise information than
TSAM. Six suitability scenarios based on TSAM and ITSAM are generated in this research. ‘Final
suitability scenario 3’ (Map 4.20) based on TSAM is found to be the best alternative after the ranking.
This is because only four suitability factors (landscape quality, historic culture value, water system
influence, existing land use) can take advantage of the certainty factor. The certainty factor is not suit-
able for the other three factors (air quality, surface water quality, noise influence) due to their defi-
cient numerical values (observation data). Also, it has some subjectivity to determine the certainty
factor when making the subclasses in the suitability class. Further, it is somewhat subjective to deter-
mine the final suitability levels (‘no suitability’, ‘moderate suitability’, ‘high suitability’) by using the
histogram of the composite suitability score in ILWIS.
Suitability analysis can be used to help direct the future growth of the green spaces and protect the
other important landuses at the same time. The result of implementing the proposed approach is a map
that categorises and illustrates the different levels of green space suitability throughout the study area.
As can be seen on Map 4.25 of the best suitability scenario (proposed plan), large portions of the
83
study area are found not to be suitable for the green space system development. This is a direct result
of the growing conflict between the ecological environment and development restriction. Areas that
receive high and moderate suitability levels are revealed to be the predominant areas of the existing
green spaces, vacant areas and government reserved areas. This is suitable for the future growth of the
green spaces. After comparing the proposed plan with the urban master plan (2000-2015), it can be
showed that their ideas and part of the results are similar. The result in the proposed plan is reason-
able and consistent with the local master plan, but it still needs to improve.
In order to advance the art of land suitability analysis, it is important that not only the result is replic-
able within a study area, but also the approach is transferable, or at least adaptable in other places. By
applying the proposed approach in Dongguan municipality, this research provides an example of such
transferability. The suitability results of the urban green space system are mathematically achieved, so
they can be applied as valid factors into the generation of other approaches to achieve alternative lan-
duse or design strategies. Revaluation of initial conceptions or updated information also can easily
regenerate landuse or design strategies. The evaluation process is so explicit that alternative scenarios
can be modelled without reconstructing the entire procedure. From the view of the technical level, this
GIS-based suitability analysis approach is quite able to integrate ecological and social geographic data
with human knowledge in an objective and manageable nature. It allows for all kinds of information
from experts and general public to be used in the weighting process. From the view of the organiza-
tional level, the use of GIS technology brings people together, including the urban planners, environ-
mentalists, local residents and local government officials. The high level of cooperation and involve-
ment generates a broad-based approach to multiobjective suitability analysis for the green space sys-
tem. From this work we are able to suggest that, the approach applied in this study area can be
adapted in other areas of Guangdong province.
In summary, the suitability results can provide helpful knowledge on the factors interactions and their
relation to the urban environment within the study area. It can be used to help people make better,
more informed decisions, thus providing a more healthy quality of life for the community and the sur-
rounding ecological environment. Suitability analysis is a powerful tool for green space system plan-
ning. Continued development and refinement of suitability analysis, particularly with GIS technology,
can enable urban planners to help local government officials and local residents to create a suitable
green space system in the urban environment.
5.2. Recommendation
Nowadays many methods can be used to make the suitability analysis, such as sieve mapping, land-
scape unit method, grey tone method (map overlay) and computer method (GIS). In this research, GIS
is used to carry out the suitability analysis just because it can handle large number of data, powerfully
visualize current and old information, produce new maps, etc. By applying GIS in the suitability
analysis, what we should do first is to establish a set of suitability factors and weighting system. This
is the most important and difficult step in the suitability analysis.
This search selects seven factors to make the suitability analysis, including air quality, landscape qual-
ity, surface water quality, historic culture value, water system influence, noise influence and existing
84
land use. It is a relatively new attempt to select these factors to analyse the suitability in the urban
planning. In some sense these factors may be not enough for showing the influence on the urban envi-
ronment by the green space system. However, there are no many useful data (maps) available for this
case study. As to the green space system, we should also consider the other ecological factors such as
soil fertility, groundwater recharge, vegetation cover, erosion control, potential damage and some
other social-economic factors (e.g. development pressure, extending public parcels), which are only
presented in the books rather than in the maps. Hence, it is difficult to integrate these data with GIS,
which will finally result in the relative shortage of urban foundational data. It shows that the scientists
can’ t have a good cooperation with the urban planners. So it is recommended that the scientists and
the planners should keep in more touch with each other, work together to carry out some researches of
the relevant speciality and make up the urban foundational data.
This research is a preliminary attempt applying GIS-based suitability analysis in the urban green space
system development. The GIS software (ILWIS) used in the case study is of a raster-based design.
The major advantages of raster-based design are their capability of individual pixel analysis and map
algebra inherent in the raster systems, which make this suitability analysis much easier. However, de-
pending on the scale of the individual cell values, the analysis may have been less precise than the
vector-based design. Allowing for the limited time and the deficient data, the pixel precision may be
not very appropriate for this suitability analysis. Further, the original maps’ scales are different that
sometimes there are some errors in transforming the data. But this suitability result is enough for the
regional control, which can be showed from comparing the proposed plan with the master plan (Map
4.25). Therefore, if a detailed planning needs to be made in a region, it is better to use vector-based
and raster-based GIS software (e.g. ArcView, MapGIS) to carry out the suitability analysis, not only
making the suitability analysis free from the manual way, but also making the suitability results more
precise and suitable for the planning.
It is widely accepted that GIS is a toolbox capable of providing support for spatial problem-solving
and decision-making. However, current GIS analysis is based on simple spatial geometric processing
operations such as overlay comparison, slicing, and distance calculation. It does not provide optimisa-
tion, iterative equation solving, and simulation capabilities necessary in the suitability analysis. Also,
GIS does not integrate nonspatial data or decision makers’ preference into the decision analysis.
Given the current set of suitability factors, GIS land suitability mapping generates a set of suitability
scenarios, rather than presenting the decision maker with an optimum scenario. Six suitability scenar-
ios are generated in this research by using GIS software (ILWIS). After that, a decision support sys-
tem (DSS) software (DEFINITE) is used to make the ranking and sensitivity analysis to find the best
suitability scenario. These two softwares cannot easily exchange the data and only can allow for one
set of weighting system at one time. Therefore, GIS should be integrated with the DSS software to
make the suitability analysis in a systematic way, taking into account spatial and nonspatial informa-
tion and allowing the decision maker to analyse the information based on one or more sets of weight-
ing systems.
GIS-DSS is an advancement in decision-making for the suitability analysis, but there is still a clear
need for further research in this field. One is about the extension of links between GIS and DSS soft-
ware, in order to increase the exchange of data in the reverse direction from DSS software to GIS.
Another topic for further research is the development of new approaches of generating alternatives
85
within the GIS. Development of such approaches should be able to increase the usefulness of GIS-
DSS, because a major disadvantage of any decision analysis is the generation of a feasible, complete
set of alternatives.
86
References
Bedward, M. and R. L. Pressey (1991). "Scores and score classes for evaluation criteria: A compari-
son based on the cost of reserving all natural features." Biological Conservation 56(3): 281-294.
Bedward, M., R. L. Pressey, et al. (1992). "A new approach for selecting fully representative reserve
networks: addressing efficiency, reserve design and land suitability with an iterative analysis." Bio-
logical Conservation 62(2): 115-125.
Bruijn, C. A. d. (1991). "Spatial factors in urban growth: towards GIS methods for cities in develop-
ing countries." ITC Journal 1991(4): 221-231.
By, R. A. d., R. A. Knippers, et al. (2000). Principles of Geographic Information Systems. Enschede,
ITC.
Chen, S. (2000). Urban Expansion and Landuse Increment. Beijing, Chinese Architectural and Indus-
trial Press. (in Chinese)
Colson, G. and C. d. Bruyn (1989). "Models and methods in multiple objectives decision making."
Mathematical Computer Modelling 12(1989): 1201-1211.
Costanza, R., L. Wainger, et al. (1993). "Modeling complex ecological economic systems." BioSci-
ence 43(8): 545-555.
Gaoyuan, R., Z. Yang, et al. (1983). Urban Green Space Planning. Beijing, Chinese Architecture and
Industry Press. (in Chinese)
Hemmati, M. (2002). Multi-stakeholder processes for governance and sustainability: beyond deadlock
and conflict. London, Earthscan.
Hofstee, P. and M. Brussel (1999). Analysis of suitability for urban expansion. Enschede, ITC.
Hopkins, L. D. (1977). "Methods for generating land suitability maps: a comparative evaluation."
Journal for American Institute of Planners 43(10): 386-399.
Hu, P. (1994). "Suitability analysis for system planning of city open space." Journal of Chongqing
Jianzhu University 18(3): 8-14. (in Chinese)
Huang, B. (1997). "Certainty factor function and fuzzy mathematics." Remote Sensing Technology
and Application 12(1): 14-18. (in Chinese)
Huang, G. and Y. Chen (2002). Ecocity: Theory and Design Approach. Bejing, Science Press. (in
Chinese)
87
Jankowski, P. and L. Richard (1994). "Integration of GIS-based suitability analysis and multicriteria
evaluation in a spatial decision support system for route selection." Environment and Planning 21:
323-340.
Janssen, R. (1994). Multiobjective Decision Support for Environmental Management. Amsterdam,

Kluwer Academic Publishers.
Janssen, R., M. v. Herwijnen, et al. (2001). DEFINITE: Getting started manual. Amsterdam, Institute
for Environmental Studies: 39.
Jia, J. (2001). Planning and Design of Green Space System. Beijing, Chinese Forestry Press. (in Chi-
nese)
Kalogirou, S. (2002). "Expert systems and GIS: an application of land suitability evaluation." Com-
puters, Environment and Urban Systems 26(2-3): 89-112.
Kuang, P. (1996). "Suitability analysis for system planning of city open space." Journal of Chongqing
Jianzhu University 18(3): 8-14. (in Chinese)
Li, M. (1999). Urban Green Space System and Human Settlements Planning. Beijing, Chinese Archi-
tectural and Industrial Press. (in Chinese)
Liu, L. (2003). Distributed behaviour: a Case Study on Recycling Behaviour in Chongqing, China.
Urban Planning and Land Administration. Enschede, ITC: 82.
Lynch, k. (1961). The Image of City. London.
M.J.Cohen (1993). "Integrated ecology: the process of counselling with nature." The Humanistic Psy-
chologist 21(3).
McDonald, G. T. and A. L. Brown (1984). "The land suitability approach to strategic land-use plan-
ning in urban fringe areas." Landscape and Planning 11(2): 125-150.
McHarg, I. (1969). Design with Nature. Garden City, New York, The Natural History Press.
Miller, W., M. G. Collins, et al. (1998). "An approach for greenway suitability analysis." Landscape
and Urban Planning 42(1998): 91-105.
Mugenyi, A. (2002). Use of Green Space in Zaqmalek, Cairo. Urban Planning and Land Administra-
tion. Enschede, ITC: 30.
Mumford, L. (1979). Interpretations and forecasts: 1922-1972: studies in literature, history, biogra-
phy, technics, and contemporary society. New York etc, Harcourt Brace Jovanovich.
88
Nijkamp, P. and A. V. Delft (1977). Multi-criteria Analysis and Regional Decision-making. Leiden,
Amsterdam.
Nisar Ahamed, T. R., K. Gopal Rao, et al. (2000). "GIS-based fuzzy membership model for crop-land
suitability analysis." Agricultural Systems 63(2): 75-95.
Pease, Coughlin, et al. (1996). Land Evaluation and Site Assessment. A Guidebook for Rating Agri-
cultural Lands. Ankeney,IO, Soil and Water Conservation Society.
Prasad, S. (1999). Suitability Analysis and Decision Making Using GIS. Spatial Modelling.
R.Ramanathan (2001). "A note on the use of the analytic hierarchy process for environmental impact
assessment." Journal of Environmental Management 63 (2001): 27-35.
Saaty and Vargas (1984). "Comparison of eigenvalue and logarithmic least squares and least methods
in estimating ratios." Mathematical Modelling 5 (5): 309-324.
Sharifi, A. and M. v. Herwijnen (2003). Spatial Decision Support Systems. Enschede, ITC.
Shi, S. (2002). Research on the Urban Spatial Structure Guided by the Urban Green Space System.
Physical Geography. Wuhan, Central China Normal University: 112. (in Chinese)
Shuaib, L. (1998). The Application of GIS in Land Suitability Analysis for Residential Development
in Kampala Uganda. Geo-Information for Urban Management Applications. Enschede, ITC: 32.
Steiner, F. (1983). "Resource suitability: Methods for analyses." Environment Manage 7(5): 401-420.
Steiner, F., L. McSherry, et al. (2000). "Land suitability analysis for the upper Gila River watershed."
Landscape and Urban Planning 50(4): 199-214.
Talento, M. F. (2002). Green Accounting: Improving the Business Criteria for Sustainable Forest
Management. Natural Resources Management. Enschede, ITC: 82.
Voogd, H. (1983). Multicriteria Evaluation for Urban and Regional Planning. Pion, London.
W.Miller, R. (1996). Uran Forestry: planning and managing urban greenspaces. New Jersy.
Wang, Z. (2003). A System Approach to Estimate Tree Resources Outside the Forest. Forestry for
Sustainable Development. Enschede, ITC: 71.
Westen, C. J. v. (1997). Tools for Map Analysis Applied to the Selection of a Waste Disposal Site.
Enschede, ITC.
89
Wu, C. (1998). Theory and Method of Contemporary Urban Recreation Design. Beijing, Chinese Ar-
chitectural and Industrial Press. (in Chinese)
Wu, R. (1998). "The development process of foreign urban green space system." Urban Planning
22(6): 50-58. (in Chinese)
Wu, R. (1999). "The classification of green space system." Chinese Horticulture 15(6): 26-32. (in
Chinese)
Xu, Z. (1999). Urban Environment Planning. Wuhan, Wuhan Technical University of Surveying and
Mapping Press. (in Chinese)
Xue, W. (2001). Site Selection for New Wastewater Treatment Plants. Urban Planning and Manage-
ment. Enschede, ITC: 64.
Yan, L. (2001). "The consideration of potential in the green space system." Chinese Horticulture
17(2): 66-72. (in Chinese)
Zaporozec, A. and F. D. Hole (1976). "Resource suitability analysis in regional planning--with special
reference to Wisconsin, USA." Geoforum 7(1): 13-22.
Zhao, H. (1999). Urban Planning and Urban Development. Nanjing, Dong Nan University Press. (in
Chinese)
Zhu, Z. (1989). Psychology Gradus. Beijiing, Science Press. (in Chinese)
90

Yang Manlun

Uploaded by

Copyright:

Available Formats

Yang Manlun

Uploaded by

Document Information

Original Title

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

Yang Manlun

Uploaded by

Copyright:

Available Formats

Suitability Analysis of

Urban Green Space

Thesis Assessment Board

Prof. Dr. Willem v.d. Toorn (Chairman)

INTERNATIONAL INSTITUTE FOR GEO-INFORMATION SCIENCE AND EARTH OBSERVATION

1.2. Problem statement

1.3. Research objective

1.3.1. Main objective

1.3.2. Specific objectives

1.4. Research questions

2. Analyse the strengths and weaknesses of current suitability analysis methods.

3. Evaluate the GIS-based traditional suitability analysis model.

Literature - Urban green space system

Selecting 7 factors for

Defining 3 sets of Using ILWIS to do the Getting alternatives

Using Definite to make

Figure 1.1 Research workflow

1.6. Structure of the thesis

2. Definition and conceptions of urban

2.1. Definition of urban green space system

Name Discipline Meaning

Ecological Urban Planning

Figure 2.1 Developmental skeleton of green space system

Table 2.1 Definitions of green open space

2.2. Classification of urban green space system

2.2.1 Classification in foreign countries

Table 2.2 Classification of parks in America

Table 2.3 Classification of urban green space system in Japan

Park for ordinary use

2.2.2. Classification in China

Table 2.4 Classification of urban green space system in China

2.3. Comprehensive benefits of urban green spaces

2.3.1. Ecological benefits

2. Adjust and improve urban climate

3. Prevent and reduce hazard

(2) Water and soil conservation

2.3.2. Social benefits

Requirements for ego reality: knowledge, idea, ambition, etc.

Requirements for respect: achievement, power, right, fame, etc.

Safety requirements: avoiding danger and horror, etc.

Physiological requirements: hunger, thirst, air, sex, etc.

Figure 2.2 Hierarchical requirement theory (Abraham H. Maslow)

Figure 2.3 Circular ring for education function of green spaces

2.3.3. Economic benefits

3.1. Definition of suitability analysis

3.2. Suitability analysis methods

3.2.1. Direct overlay

Formula 3.1 Equal-weight summation

3.2.2. Weighted score

Formula 3.2 Weighted score

3.2.3. Ecological factors combination

3.3. GIS application in suitability analysis

3.4. GIS-based traditional suitability analysis model (TSAM) and its

3.4.1. Traditional suitability analysis model (TSAM)

Table 3.1 Factors and weights in the traditional suitability analysis

3.0 3.0 2.0

Figure 3.1 TSAM

3.4.2. Improved traditional suitability analysis model (ITSAM)

Formula 3.4 Composite certainty factor