Modelling urban growth evolution and land-use

Environ Earth Sci (2013) 70:425–437 DOI 10.1007/s12665-012-2137-6

ORIGINAL ARTICLE

Modelling urban growth evolution and land-use changes using GIS based cellular automata and SLEUTH models: the case of Sana’a metropolitan city, Yemen Mohamed Al-shalabi • Lawal Billa • Biswajeet Pradhan • Shattri Mansor Abubakr A. A. Al-Sharif

•

Received: 16 December 2011 / Accepted: 18 November 2012 / Published online: 1 December 2012 Ó Springer-Verlag Berlin Heidelberg 2012

Abstract An effective and efficient planning of an urban growth and land use changes and its impact on the environment requires information about growth trends and patterns amongst other important information. Over the years, many urban growth models have been developed and used in the developed countries for forecasting growth patterns. In the developing countries however, there exist a very few studies showing the application of these models and their performances. In this study two models such as cellular automata (CA) and the SLEUTH models are applied in a geographical information system (GIS) to simulate and predict the urban growth and land use change for the City of Sana’a (Yemen) for the period 2004–2020. GIS based maps were generated for the urban growth pattern of the city which was further analyzed using geostatistical techniques. During the models calibration process, a total of 35 years of time series dataset such as historical topographical maps, aerial photographs and satellite imageries was used to identify the parameters that influenced the urban growth. The validation result showed an overall accuracy of 99.6 %; with the producer’s accuracy of 83.3 % and the user’s accuracy 83.6 %. The SLEUTH model used the best fit growth rule parameters

M. Al-shalabi Department of Geography, University of Sana’a, Sana’a, Yemen L. Billa School of Geography, University of Nottingham, Malaysia Campus, Semenyih, Malaysia B. Pradhan (&) S. Mansor A. A. A. Al-Sharif Department of Civil Engineering, Faculty of Engineering, Geospatial Information Science Research Center (GISRC), University Putra Malaysia, Serdang Selangor, Malaysia e-mail: [email protected]; [email protected]

during the calibration to forecasting future urban growth pattern and generated various probability maps in which the individual grid cells are urbanized assuming unique ‘‘urban growth signatures’’. The models generated future urban growth pattern and land use changes from the period 2004–2020. Both models proved effective in forecasting growth pattern that will be useful in planning and decision making. In comparison, the CA model growth pattern showed high density development, in which growth edges were filled and clusters were merged together to form a compact built-up area wherein less agricultural lands were included. On the contrary, the SLEUTH model growth pattern showed more urban sprawl and low-density development that included substantial areas of agricultural lands. Keywords Urban growth Land-use change Remote sensing GIS Cellular automata SLEUTH Sana’a

Introduction Most of damages and harmful effects in environment are caused by anthropogenic activities. The changes of land use and land cover occur due to urbanization caused by unplanned and uncontrolled urban sprawl which leads to change nature, destroy green cover and pollute the water resources. The present needs are analysis, understanding, modelling urban growth evolution and land use changes to save and provide suitable and safe environment for the mankind. Urban growth modelling is essential for analytical and the prediction of the dynamics of urban growth. GIS technique has been used to study, analyze and correlate urban activities and land use changes and its effects on ground water and environment. Most of these studies

123

426

revealed that the urbanization and change in land use are important factors affecting the water resources and environment (Bathrellos et al. 2008). Urban models have been used to forecast future changes or trends of development, to describe and assess impacts of future development, and to explore the potential impacts of different policies. The geographical information system (GIS) offers a powerful tool for the spatial analysis of a multi-dimensional phenomenon (Youssef et al. 2011). The susceptibility map is a practical tool in natural and urban planning. It reflects the availability of lands for urbanization and other requirements. Many factors such as slope, aspect, road network and land use are generally used during the model building (Bathrellos et al. 2009; Rozos et al. 2011; Al-shalabi et al. 2012). In the literature, it is common to encounter methodologies using spatial multiple criteria analysis in GIS environment for urban growth management. The Analytic Hierarchy Process (AHP) and GIS has been popularly used to predict urban growth using many factors that affects future growth and water demand (Panagopoulos et al. 2012). In the last two decades, a lot of research has been done on urban spatial modelling due to increased computing power, improved availability of spatial data, and the need for innovative planning tools for decision support (Geertman and Stillwell 2004; Brail and Klosterman 2001; Bathrellos et al. 2012; Biro et al. 2011; De Abreu and Filho 2011; Jabbar and Zhou 2011; Lv et al. 2011; Peng et al. 2011; Pradhan 2011). Urban system models have shown potential in simulating the complexity of dynamic urban processes and can provide an additional level of knowledge and understanding of spatial and temporal change (Wu and Silva 2009; Kaya and Curran 2006; Pettit et al. 2002). These aforementioned models include both macro and micro integrated cellular automata (CA) models (RIKS model) (Engelen et al. 1997), fuzzy CA model (Wu 1996), ANN CA model, (Li and Yeh 2002), multi-CA model (Cecchini and Rinaldi 1999) and the SLEUTH urban growth model that has been widely applied in the United States and many other parts of the world (Jantz et al. 2004; Leao et al. 2004; Yang and Lo 2003; Esnard and Yang 2002; Silva and Clarke 2002; Clarke et al. 1997; Zeug et al. 2006). These models have been useful in planning and decision support for urban environmental management. They are applied together with other models, such as; the logistic regression model to identify the significant variables and rules that differentiate urban or city from rural and forest environments; the relative probability model which uses spatial interactions of neighbourhood, distance, patch size (parcels), and site-specific characteristics; and the focus group involvement to create a human input layer, set the growth scenarios, evaluate predictions and disseminate the information (Allen and Lu 2003).

123

Environ Earth Sci (2013) 70:425–437

In a data deficient country like Yemen, urban environment planning and management have many challenges and most particularly the lack of a mechanism to forecast and predict urban growth trends and patterns. In Yemen, development in uncontrolled and planning is haphazard, lacking any clear vision about the future. For that reason, cities have many socio-economic, infrastructural and management problems. In this study, two urban growth models, i.e. the CA and the SLEUTH are applied in modelling the urban growth trend of Sana’a the biggest city in Yemen and to predict and forecast the growth pattern from the year 2004–2020. These advanced urban growth models have been developed using the factors by considering certain assumptions which have been popularly applied in many cities around the world (ESCWA 2007; Pedro and Zamyatin 2006). This study tests the applicability of these models in a city where growth condition may be different; it also compares the output growth pattern of the two models.

Urban growth modelling using cellular automata (CA) and the SLEUTH models Cellular automata (CA) models were developed for many modelling purposes but have been popularly applied in the area of modelling urban studies and growth processes depending upon their transition rules, and calibration methods (Couclelis 1985, 1997). Many forms of urban CA have been used to simulate the spatial pattern of an actual city or a synthetic city (Batty 1997; Yeh and Li 2001). Wagner (1997) explored the potential use of CA in urban planning and discussed theoretical obstacles in incorporating CA models in geographical context. These models have been used to study the evolution of urban land use, modelling urban forms; urban models and many other urban studies (White and Engelen 1993, 1994; Clarke et al. 1997, Clarke and Gaydos 1998a, b; Batty and Xie 1997; Batty 2000). In recent years, new CA models, such as URBANSIM, UPLAN, SLEUTH have been used to forecast future changes, trends of development, describe and assess impacts of future development, and to explore the potential impacts of different policies (Herold et al. 2001). CA models have many advantages in the modelling of urban processes that include the abil-ity to perform spatial dynamics, and time explicitly. After successfully analysing the similarities and capabilities of CA, Wagner (1997) proposed that CA can be considered as analytical engine of GIS. Raster GIS with map algebra can be integrated with enhanced capabilities as discussed by Takeyama and Couclelis (1997). CA are considered to have a ‘‘natural affinity’’ with raster data. It has similarities with GIS, such as the representation of attribute information in a layered


fashion, and the manipulation of information with operators (Overlay in GIS, Transitional rules in CA). The focal sum or focal mean functions of GIS has direct analogous with neighbourhood functions. Because of its natural affinity with GIS, it was obviously adopted by geographers as a tool for modelling spatial dynamics (Singh 2003). Clarke et al. (1997) introduced the Clarke urban growth model, a precursor to the SLEUTH model for simulating historic change. The model aids the illustration and explanation of growth processes at a regional scale and predicts future urban growth trends. The model was successful in simulating urban change between 1900 and 1990 for the San Francisco area, and was later applied to the Baltimore Washington corridor (Clarke and Gaydos 1998a, b), where calibrations and long term predictions for both San Francisco and Baltimore Washington were presented, allowing for an effective comparison to be made between the growth patterns and processes of the two urban systems. The SLEUTH model is now in a public domain C-language source code, available for download with documentation online from USGS website at http://www.ncgia.ucsb.edu/projects/gig. The model simulates land use change as a consequence of urban growth by working on a grid space of pixels, with a neighbourhood of eight cells of two cell states (urban/nonurban), and five transition rules that act in a sequential time steps. The states are acted upon by behaviour rules, and these rules can self-modify to adapt a place and simulate change according to what have been historically the most important characteristics. SLEUTH requires five GIS-based inputs: urbanization, land use, transportation, areas excluded from urbanization, slopes, and hill-shading as a background. The input layers must have the same number of rows and columns, and should be correctly geo-referenced. For statistical calibration of the model, urban extent must be available for at least four time periods. Urbanization results from a ‘‘seed’’ urban file with the first urban year used, with at least two road maps that interact with a slope layer is used in order to allow the generation of new nuclei for outward growth. Besides the topographic slope, a constraint map represents water bodies, natural and agricultural reserves. After reading the input layers, initializing random numbers and controlling parameters, a predefined number of interactions take place that correspond to the passage of time. A model outer loop executes each growth history and retains statistical data, while an inner loop executes the growth rules for a single year.

Study area The city of Sana’a, Yemen is located within the latitudes 15° 100 0000 and 15° 300 0000 North and longitudes 44° 050

427

0000 and 44° 200 0000 East (Fig. 1). The total area was 15,284.84 km2 and the built up area was around 138.66 km2 in 2003 (Al Shaibi et al. 2006). The city has a moderate climate all year round, due to its location at about 2,200 m above sea level. Sana’a city is located to the Northern-central part of Yemen in a high valley extending from south to north. The old city has a unique urban and architectural design that dates back to over 2,500 years. In 2004 the total population of Sana’a was estimated at 1,747,834 (Al Shaibi et al. 2006; Alderwish and Almatary 2011).

Data and methodology Various data used in this study includes Quick bird satellite imagery (0.60 m) acquired in 2003, aerial photography (4 m) acquired in 1994, digital contour line map (10 m interval), hydrological map 1:50,000, road network and the city master plans maps from the Ministry of Public Works and urban planning and the statistical information for different years from Central Statistical Organization in Yemen. In developing countries the availability of data is a major concern for this kind of studies. In this study, the available data were collected between the years 1978–2003, this made it possible for the calibration process to register the progression of urban growth signature over the time period. Data for growth after 2003 were not available so could not be included in the calibration process. Data generally comprised maps of different types, dates, scales and time, those were pre-processed into a uniform geo-reference to create a profile of urban extent of Sana’a city over space and time and organized into spatiotemporal GIS database. Other types of data such as statistical information were collected from various reports published by different ministries and departments in Yemen in 1999. These data was further classified and clipped to the map extent (Hurskainen and Pellikka 2004) and transformed to the same cell size raster grids at 45 meter resolution. This resolution was used for all data layers. The grid dimensions were 527 columns by 811 rows. These datasets were also used as an input for the CA and SLEUTH model calibration process. Prediction of urban growth and land use change using CA transition rules A suitability map was used to estimate the annual demand of land for urban development depending on the historical growth of Sana’a city. The evolution of a cell was determined by the suitability value and the number of cell that were developed. The cells were then calculated to establish how many cells had attained a particular state of a time

123

428


Fig. 1 Location map of the Sana’a city and its urban extent

transition rule (Xi et al. 2009; Watkiss 2008). Cells are allocated to a particular state by selecting them from the set of available (the cells which can undergo transition) cells that are spread over the city. Subsequently, the cell candidatures were evaluated using multi criteria analysis (MCA) technique to determine their possible inclusion in the model. Equations 1 and 2 ere used for formulation of the model (Samat 2005). t t t utþ1 u ¼ f ; X ; ð1Þ S i;j i;j i;j i;j where, utþ1 i;j = the state of the cell at row i and column j at time t ? 1; uti;j = the state of the cell at row i and column j at time t; Xti;j = the development of cells within the neighborhood of the cell at row i and column j; and Sti;j = the suitability score for the cell at row i and column j for urban development. In the above equation, the function f is formulated using IF, THEN and ELSE statements as shown in Eq. 1. The repeated application of this rule produced a complex spatial pattern. IF ðSti;j Xti;j threshold value); tþ1 Then utþ1 i;j = urban; Else ui;j = non-urban.

123

ð2Þ

The urban growth rules involve the selection of a location by investigating the spatial properties of the neighboring cells, and urbanizing the cell under consideration based on a set of weighted probabilities using landuse and suitability map. High priority and suitable areas such as the master plan areas and proximity to existing developed areas are given a score factor of 5. The unsuitable and development restricted lands are given a score of 0. The details of these score factors are developed for transition rule processes (Table 1). After cells have been assigned score, the top scoring cells (cell available for transition) are allocated to particular states (land use type) using Eqs. 1 and 2 and afterwards each cell is assigned to an integer value based on land use classes. Prediction of urban growth using the SLEUTH model SLEUTH is a moniker for the input data required to use the model: slope, land use, exclusion, urban, transportation and hill shade. It employs a raster data (8-bit GIF) as input format but it does not use GRIDs directly. For the modelling and prediction of urban growth, the model supports three different modes: test, calibration, and prediction. The model is computation-intensive especially in the


429

Table 1 The factors and scores effect of the transition rule Factors

Score

Proximity to existing developed areas

5

Proximity to prioritized land (master plan proposed areas)

5

Table 2 The parameters of the coarse, fine, final calibration (45 meter pixel size) Dispersion

Breed

Spread

Slope

Road gravity

Parameters of coarse calibration

Suitability factors

Start

1

1

1

1

1

High suitable

5

Suitable

4

Step Stop

25 100

25 100

25 100

25 100

25 100

Moderate suitable

3

Possible units

5

5

5

5

5

Less suitable

1

4

Unsuitable

Restricted

Monte Carlo iterations Possible combinations

3,125

Elapsed time

4 Days, 8 h, 7 min and 3 s.

Land use factors Agricultural land

5

Industry land

Restricted

Mountains

1

Green areas

Restricted

Other land use

Restricted

Parameters of fine calibration Start Step

80 5

70 5

20 5

45 5

45 7

Stop

100

85

35

65

100

Possible units

5

4

4

5

12

Monte Carlo iterations

6

Possible combinations

4,800

Elapsed time

7 days, 3ours, 49 min and 13 s

Parameters of final calibration Start 88

78

17

40

94

Step

1

1

1

2

1

Stop

93

84

23

56

98

Possible units

6

7

7

9

5

Monte carlo iterations

8

Possible combinations

1,3,230

Elapsed time

9 days, 4 h, 7 min and 33 s

animation that illustrates the urban growth pattern and change over time. Figure 2 shows the structure of the SLEUTH model as illustrated by Yang and Lo (2003). Model calibration

Fig. 2 General structure of the SLEUTH model (after Yang and Lo 2003)

calibration mode (Xi et al. 2009; Candau 2002). The computation time varies from hours to days depending on the spatial resolution and the image size. According to Jantz et al. (2004), it requires a several days to calibrate the urban growth for a medium-size city. The outputs include a series of GIF images, representing the yearly urban scenarios, which can be compiled into time series for

Urban areas throughout the world grow at different rates and in different ways due to varying economic conditions and environmental constraints (Kaya and Curran 2006; Daniels 1999). Urban growth modelling (UGM) of Sana’a city will thus require predictions that are consistent with the factors and constraint with the city. Sana’a city model’s coefficient values were derived from historical dataset that was ‘‘fit’’ to area through the UGM’s calibration phase using the brute force method (Goldstein et al. 2004). The model was first run in a calibration phase to obtain a suitable set of parameters which was further refined in the sequential calibration steps. The calibration allows the

123

430


Table 3 Prediction of spatial urban growth in Sana’a City 2004–2020

Table 4 Land-use change for the year 2010 Land use

2003

2010

Years

Area (pixel)

Area (km2)

Annually developed cells

2004

71,052

143.88

2,578

Agricultural land

2005

73,891

149.63

2,839

Mountains

2006

76,912

155.75

3,021

Green area

2007

80,130

162.26

3,218

Industry

2008

83,529

169.15

3,399

Other land use

11.30

2009

87,061

176.30

3,532

2010 2011

90,676 94,407

183.62 191.17

3,615 3,731

2012

98,342

199.14

3,935

2013

102,372

207.30

4,030

2014

106,550

215.76

4,178

2015

110,917

224.60

4,367

2016

115,408

233.70

4,491

2017

119,986

242.97

4,578

2018

124,647

252.41

4,661

2019

129,461

262.16

4,814

2020

134,436

272.23

4,975

Built up area

Fig. 3 Predicted growth areas for 2010 using the CA model

model to simulate urban growth from the past into the present with a very high degree of fit between the simulated years and the control years. The coarse calibration, took 25

123

Value

138.66

184.75

45.74

444,742.36

399.04

-43.43

229.48

227.17

-2.31

6.62

24.90

18.28

10.12

10.12

None

11.30

None

unit steps for each entire coefficient space, for all coefficients. The fine calibration, took 5 unit steps and the final calibration took either 1–2 unit steps through the coefficient space. Table 2 shows a summary of the parameters and processing outputs obtained in the calibration.

Results and discussion The prediction of urban growth change was performed by using GIS-based CA transition rule. The annual demand of land was estimated based on the models calculation of rate of change and urban growth for the calibration period/ historical data (1978–2003). The results presented in Table 3 shows that urban growth/extent of 138.6 km2 in 2003 will increase to 149.63 km2 in 2005, then to 183.62 km2 in 2010 and 272.23 km2 in 2020 respectively, with an average annual growth rate of 4.05 %. The prediction of the urban growth extent was paced with a difference of five years (2005, 2010, 2015 and 2020) so as to allow enough significant growth during the period to register a change in the prediction. The graphical presentation of predicted urban growth is shown in the sampled results for the year 2010 (Fig. 3). The figure shows the growth pattern using a GIS-based CA model are compact in the suitability lands, i.e. the growth has filled spaces and has merged many clusters together around existing built-up areas and occupies the prioritized land in the master plan. In this study, land-use was defined as state of cells at time ‘t’, represented by five major categories: built-up area, agricultural land, industry land, green areas, and mountains. The suitability land was represented by five categories: high suitable, suitable, moderate suitable, less suitable, and unsuitable. The sampled results obtained from the prediction modelling for the year 2010 are presented here for better discussion and explanation of the growth and change of the land-use. Table 4 and Fig. 4 show the spatial growth and land-use changes respectively. Agricultural land is reduced from 444,742 km2 in 2003 to 339 km2 in the year 2010, while the green areas was increased from 6.62 km2 in 2003 to 24.9 km2 in 2010. The industrial land does not show any changes because; it has


431

Fig. 4 The change in land-use from 2003 to 2010 using GIS-based CA model

Table 5 Validation result of the GIS-based CA prediction for the year 2003 Actual pixels

Predicted pixels

Correct pixels

Overall accuracy %

Procedure’s accuracy %

User’s accuracy %

68,451

68,688

57,204

68,451/ 68,688

57,204/ 68,688

57,204/ 68,451

99.6

83.3

83.6

already been included in the western part of the city as future industrial area as documented in the master plan 1999. Validation of the CA model The validation of the modelling results of CA was carried out by predicting the spatial urban growth from 1994 to 2003 and comparing it with the actual growth pattern of the city in 2003. The validation result (Table 5) shows an overall accuracy of 99.6 %. This high level of accuracy was achieved because very high spatial data was used in the processing. The producer accuracy however was 83.3 % while the user’s accuracy was 83.6 %. Considering the high level of accuracy achieved by the prediction, the dataset (1994–2003) was further used in the application of

Fig. 5 Actual and predicted urban extent for the year 2003

the model. The graphical maps of the actual urban area of 2003 and predicted were overlaid for visual comparison (Fig. 5).

123

432


Results of the SLEUTH urban growth model The five coefficient values that were obtained in the calibration process were used to predict the urban growth from 2003 to 2020 for the city of Sana’a. Table 6 shows the results of the predicted urban growth. The result indicated that urban area will increase dramatically from 138.6 km2 in 2003 to 143.75 km2 in 2005,159.18 km2 in 2010 and 199.71 km2 in 2020. The growth patterns indicate a very high value of diffusion and breed and also showed very high road gravity coefficient and low slope coefficients. Growth clearly occurred at the urban fringe resulting organic growth and increase in suburbs. Very high value of diffusion can be attributed to the fact that the city has no zoning plan which has allowed unrestricted outward growth in all directions. Table 7 shows the effect of the urban growth on land-use change. The annual probabilities

transition from agricultural land to urban land is about 11.92 % and the annual transition for the mountain area is 0.03 % with 0.065 km2 per year. The graphical results of the SLEUTH model prediction of urban growth for the year 2010 is shown in Fig. 6. In this map it can be seen that the outward growth is scattered all over the map. The prediction of land-use change for the same year is shown in Fig. 7. Comparison between the CA and SLEUTH model A comparison was made between the two sampled modelling results for the year 2010 for both CA model and SLEUTH model to understand the dynamics of transitional probability change and also to estimate the variance between the two modelling approaches. Figure 8 shows an

Table 6 Model prediction of growth extent of Sana’a urban area (2004–2020) Years

Area (pixel)

Area (km2)

Annually developed cells

2004

69,730

141.20

1,281

2,005

70,990

143.75

1,260

2006

72,413

146.65

1,423

2007

73,858

149.56

1,445

2008

75,415

152.72

1,557

2009

76,993

155.91

1,578

2010

78,609

159.18

1,616

2011

80,312

162.63

1,703

2012

82,028

166.11

1,716

2013

83,744

169.58

1,716

2014

85,722

173.59

1,978

2015

87,667

177.53

1,945

2016

89,789

181.82

2,122

2017

91,885

186.17

2,096

2018

93,988

190.33

2,103

2019 2020

96,257 98,622

194.92 199.71

2,269 2,365

Table 7 Annual land-use transition probabilities

123

Fig. 6 Predicted growth areas for 2010 using the SLEUTH model

Unclass

Urban

Agricultural

Mountains

Park

Industry

Airport

Others

Unclass

99.57

0.02

0.31

0.11

0

0

0

0

Urban

0

98.89

1.08

0

0

0

0.02

0

Agricultural

0

11.92

86.87

0.69

0.01

0

0.5

0

Mountains

0.03

0.03

0.08

99.73

0

0

0.13

0

Park

0

0

0

0

100

0

0

0

Industry

0

0

0

0

0

100

0

0

Airport

0

0.05

0

0

0

0

99.95

0

Others

0

0

0

0

0

0

0

0


433

Fig. 7 Predicted land-use change for 2010 using the SLEUTH model

450

424.45 399.04

SLEUTH Model

GIS-based CA Model

400 350

Km2

300 250 200

228.33 228.17 184.75 159.18

150 100 50

24.9 24.9

11.3 11.3

10.12 10.12

0 Built up area

Agricultural Mountains Industry land

Green land

others

Fig. 9 Statistical comparison of the CA and SLEUTH models in the land-use change prediction

Fig. 8 Comparison of prediction results of 2010 for the CA and SLEUTH models

overlay comparison of the two modelling results. The SLEUTH model results showed a wider spread of the growth patterns and smaller urban clusters, while the GISbased CA model showed a compact growth pattern where small clusters are merged together. The CA grow pattern reflected the proposal made in the Sana’a master plans where it is suggested to incorporate adjacent lands into the city’s residential development. In Fig. 9 the comparison results are presented based on the five land-use classes. It showed that the predicted growth for 2010 using the CA model is about 184.75 km2 when compared to 159.18 km2

123

434


Fig. 10 Validation of Urban growth predicted using CA-model for the year 2010

using the SLEUTH model. In general the results from the SLUETH model showed extensive urban sprawl and spatial distribution when compared with the CA model. The verification of the predicted urban growth for Sana’s city by field measurement was not possible due to logistical constraints. Moreover, due to scarcity of high spatial resolution satellite images for prediction time zone, the verification of the models application was conducted using the Google images (Digital Globe image 0.6 m) in Google Earth free software, that provide the capability of time tuning through 2003–2010. The predicted results for the year 2010 was possible by overlaying the images on the google images. For the validation purpose, we chose the Google image of 2010. A comparison was thus made by visual validation process that showed significant similarity between the existence and predicted urban growth for the Sana’a. By logical inference and visual inspection, the

123

study concluded that where the predicted urban growth shows a good correlation over the Google images. Figures 10 and 11 show the validation results of 2003 and 2010 that inevitably support the predicted spatial urban growth in Sana’a. The actual urban growth and land use changes in Sana’a city that occurred from 2003 to 2010 are shown in Figs. 10 and 11 respectively. These figures give a clear assessment and comparison of the results of urban sprawl and land use changes that we have predicted using CA and SLUETH models used in this study. By comparing the results of both models with the real urban growth from 2010 (Figs. 10 and 11), it is evident that that models employed in this study is sufficient enough for prediction of urban growth. The advantage of GIS-based CA is that it generated results with geographical reference and also handles large amount spatial data which allows for the creation of


435

Fig. 11 Validation of Urban growth predicted using SLEUTH model for the year 2010

constraints and other criteria by assigning the weight that helped in determining whether sites were suitable or unsuitable to the growth modelling. The prediction by GISbased CA model reflected the projections of the master plan; thus can be used to develop master plan projection to support planning and decision making in the future. The prediction also shows a growth trend towards increased urban expansion due to increasing population growth. The pattern of urban growth would be greatly influenced by the controlled and protected lands and the physical characteristic of the terrain such as mountains ranges in northeast part of the city. The limitations of the model are highly dependent on the data quality, i.e. data of higher quality results in more accurate outputs. Other weaknesses affecting the models are growth in multiple directions and

also their inability to recognize new urban development that occurred away from existing urban areas. The SLEUTH model was found to be useful in visualizing and quantifying spatial growth extent. It also showed saving of about 25.41 km2 of growth encroachment on agricultural land-use when compared with the CA model because its growth trend has more sprawl and low-density development patterns that leads to substantially less consumption of agricultural lands. The limitation of the SLEUTH model is that, it is very sensitive to the spatial resolution of data. Consequently, a higher resolution data improves the quality of data and the model output. The SLEUTH model also require high end computational capabilities such as the use of parallel computing environment, making this model extremely difficult to apply in

123

436

a developing country with limited data inputs. It also observed that the composite results of the optimum values for the diffusion, spread, slope and road gravity parameters show successive improvement in the parameters that control the behavior of the system.

Conclusions Urban modelling is an important technique for forecasting and studying urban dynamics to understand the potential impact of growth and future development. This study found that the major urban growth occurred in agriculture areas and around unplanned (low price) areas outside township boundaries where no infrastructure facilities (such as drainage systems and disposal management system) exist. This reflects the lack of clear policy that could control and guides urban sprawl in the city. This situation will results in serious socioeconomic and environmental problems in the near future. In this study the growth is mainly affected by topography and road networks. It is necessary to accommodate the fast population growth by following clear plan that take into consideration of the socioeconomic conditions, and exploit the growth factors to put clear policy and regulation by providing facilities and infrastructures which encourage the urban growth in controlled planned areas. The CA and SLEUTH are some of the models that have been extensively used in growth modelling. Although these techniques are useful for forecasting and prediction for future growth which support planning and decision making, they have seen little used in developing countries such as Sana’a, Yemen where their impact could be greatly appreciated. In this study the results of their application for prediction and simulation of the urban growth and land use change in Sana’a city showed a high overall accuracy of 99.6 %, producer accuracy of 83.3 % and user accuracy of 83.6 %. The growth pattern of CA model presented a compact and high density development, while the SLEUTH model presented an urban spread pattern with very low-density development. The study concluded that both models are very useful urban modelling tools and enable the prediction and generation of different urban growth scenarios in support of planning and decision making. By coupling the models GIS, output results showed geo-referenced map that helped to identify and demarcate specific locations for the implementation of planning policies. Implementation of the SLEUTH model however required many datasets and high end computing power for processing; this will be a major limitation in developing countries that have deficient resources. The study demonstrated an efficient implementation of the CA and SLEUTH models coupled with GIS in urban growth and land-use modelling and

123


allowed the testing of different policy alternatives on growth scenarios. Acknowledgments The study benefited from the academic scholarship provided by the Ministry of Higher Education Yemen. Authors are also very thankful for the support given by various ministries and departments in Yemen with the provision of data and other information used in the study. Thanks to two anonymous reviewers for their useful comments on the earlier version of the manuscript.

References Alderwish AM, Almatary HA (2011) Hydrochemistry and thermal activity of Damt region, Yemen. Environ Earth Sci (Article online first available). doi:10.1007/s12665-011-1192-8 Allen J, Lu K (2003) Modeling and prediction of future urban growth in the Charleston Region of South Carolina: a GIS-based integrated approach. Conserv Ecol 8(2):2. (Online) URL: http://www.consecol.org/vol8/iss2/art2 Al-shalabi M, Pradhan B, Billa L, Mansor S, Althuwaynee OF (2012) Manifestation of remote sensing data in modeling urban sprawl using the SLEUTH model and Brute Force calibration: a case study of Sana’s city, Yemen. J Indian Soc Remote Sens. doi: 10.1007/s12524-012-0215-6 Bathrellos GD, Skilodimou HD, Kelepertsis A, Alexakis D, Chrisanthaki I, Archonti D (2008) Environmental research of groundwater in the urban and suburban areas of Attica region, Greece. Environ Geol 56(1):11–18 Bathrellos GD, Kalivas DP, Skilodimou HD (2009) GIS-based landslide susceptibility mapping models applied to natural and urban planning in Trikala, Central Greece. Estud Geol 65(1):49–65 Bathrellos GD, Gaki-Papanastassiou K, Skilodimou HD, Papanastassiou D, Chousianitis KG (2012) Potential suitability for urban planning and industry development using natural hazard maps and geological–geomorphological parameters. Environ Earth Sci 66(2):537–548 Batty M (1997) Cellular automata and urban form: a primer. JAPA 63(2):266–274 Batty M (2000) Editorial: less is more, more is different: complexity, morphology, cities, and emergence. Environ Plann B27:167–168 Batty M, Xie Y (1997) Possible urban automata. Environ Plann B 24:175–192 Biro K, Pradhan B, Buchroithner MF, Makeschin F (2011) An assessment of land use/land-cover change impacts on soil properties in the northern part of Gadarif region, Sudan. Land Degrad Dev (article on-line first available). doi:10.1002/ldr.1116 Brail RK, Klosterman RE (eds) (2001) Planning support systems: integrating geographic information systems, models and visualization tools. ESRI Press, Redlands Candau J (2002) Temporal calibration sensitivity of the SLEUTH urban growth model, master’s thesis. Department of Geography, University of California, Santa Barbara Cecchini A, Rinaldi E (1999) The multi-cellular automaton: a tool to build more sophisticated models. A theoretical foundation and a practical implementation. In: Rizzi P (ed) Computer in urban planning and urban management 6th international conference. Milano, Franco Angeli Clarke KC, Gaydos L (1998a) Loose coupling a cellular automaton model and GIS: long-term growth prediction for San Francisco and Washington/Baltimore. Int J Geogr Inf Sci 12(7):699–714 Clarke K, Gaydos L (1998b) Loose-coupling a cellular automaton model and GIS: long-term urban growth prediction for San Francisco and Washington/Baltimore. Int J Geogr Inf Sci 12(7):699–714

Environ Earth Sci (2013) 70:425–437 Clarke KC, Hoppen S, Gaydos L (1997) A self-modifying cellular automaton model of historical urbanization in the San Francisco Bay area. Environ Plann B24:247–261 Couclelis H (1985) Cellular worlds: a framework for modelling micro-macro dynamics. Environ Plann B1:585–596 Couclelis H (1997) From cellular automata to urban models: new principles for model development and implementation. Environ Plann B 24(2):165–174 Daniels T (1999) When city and country collide: managing growth in the metropolitan fringe. Island Press, Washington De Abreu ASS, Filho OA (2011) Engineering geological data in support of municipal land use planning—a case study in Analaˆndia, southeast Brazil. Envrion Earth Sci (Article on-line first available). doi:10.1007/s12665-011-1089-6 Engelen G, White R, Uljee I (1997) Integrating constrained cellular automata models, GIS and decision support tools for urban planning and policy making. In: Timmermans HPJ, Spon EFN (eds) Decision support systems in urban planning. London, pp 125–155 ESCWA (2007) Regional campaign on secure housing and land tenure and good urban governance, 2003–2006. http://www. escwa.org.lb/rcshltgug/doc/editoradmin/frame3.pdf Esnard AM, Yang Y (2002) Descriptive and comparative studies of the 1990 urban extent data for the New York metropolitan region. URISA J 14:57–62 Geertman S, Stillwell J (2004) Planning support systems: an inventory of current practice. Comput Environ Urban 28:291–310 Goldstein NC, Candau JT, Clarke KC (2004) Approaches to simulating the March of bricks and mortar. Comput Environ Urban 28:125–147 Herold MG, Menz G, Clarke KC (2001) Remote sensing and urban growth models. Demands and perspectives. In: Juergens C (ed) Proceedings of the symposium on remote sensing of urban areas. Regensburg Hurskainen P, Pellikka P (2004) Change detection of informal settlements using multi-temporal aerial photographs—the case of Voi, Se-Kenya. In: Proceedings of the 5th African association of remote sensing of environment conference, Nairobi Jabbar MT, Zhou X (2011) Eco-environmental change detection by using remote sensing and GIS techniques: a case study Basrah province, south part of Iraq. Environ Earth Sci 64(5):1397–1407 Jantz CA, Goetz SJ, Shelley MK (2004) Using the SLEUTH urban growth model to simulate the impacts of future policy scenarios on urban land use in the Baltimore -Washington metropolitan area. Environ Plann 31(2):251–271 Kaya S, Curran PJ (2006) Monitoring urban growth on the European side of the Istanbul metropolitan area: a case study. Int J Appl Earth Obs 8:18–25 Leao S, Bishop I, Evans D (2004) Spatial–temporal model for demand allocation of waste landfills in growing urban regions. Comput Environ Urban 28:353–385 Li X, Yeh AGO (2002) Neural-network-based cellular automata for simulating multiple land use changes using GIS. Int J Geogr Inf Sci 16(4):323–343 Lv Z, Wu Z, Wei J, Sun C, Zhou Q, Zhang J (2011) Monitoring of the urban sprawl using geoprocessing tools in the Shenzhen Municipality, China. Environ Earth Sci 62(6):1131–1141 Panagopoulos GP, Bathrellos GD, Skilodimou HD, Martsouka FA (2012) Mapping urban water demands using Multi-Criteria Analysis and GIS. Water Resour Manag 26(5):1347–1363 Pedro C, Zamyatin A (2006) Three land change models for urban dynamics analysis in Sintra-Cascais area. In: Proceedings of the 1st EARSeL workshop of the SIG urban remote sensing. Humboldt-Universita¨t zu, Berlin Peng J, Xu Y, Cai Y, Xiao H (2011) Climatic and anthropogenic drivers of land use/cover change in fragile karst areas of

437 southwest China since the early 1970 s: a case study on the Maotiaohe watershed. Environ Earth Sci 64(8):2107–2118 Pettit C, Shyy TK, Stimson R (2002) An on-line planning support system to evaluate urban and regional planning scenarios. In: Geertman S, Stillwell J (eds) Planning support systems in practice. Springer, Heidelberg Pradhan B (2011) Use of GIS based fuzzy relations and its cross application to produce landslide susceptibility maps in three test areas in Malaysia. Environ Earth Sci 63(2):329–349 Rozos D, Bathrellos GD, Skillodimou HD (2011) Comparison of the implementation of rock engineering system and analytic hierarchy process methods, upon landslide susceptibility mapping, using GIS: a case study from the Eastern Achaia County of Peloponnesus, Greece. Environ Earth Sci 63(1):49–63 Samat N (2005) Validating the performance of GIS-based cellular automata spatial model: a case of Seberang Perai Penang, Malaysia. Jawatan Kuasa Pemetaan Dan data spatial Negara, Bil.1 Shaibi Al, Yahia M, Larbi H (2006) Sana’a city: medium to long-term city development strategy for sustainable development. Development Institute (AUDI), Arab Urban Silva EA, Clarke KC (2002) Calibration of the SLEUTH urban growth model for Lisbon and Porto, Portugal. Comput Environ Urban 26:525–552 Singh AK (2003) Modelling land use land cover changes using cellular automata in a geo-spatial environment. Master thesis, ITC Takeyama M, Couclelis H (1997) Map dynamics: integrating cellular automata and GIS through geo-algebra. Int J Geogr Inf Sci 11(1):73–91 Wagner D (1997) Cellular automata and geographic information systems. Environ Plann 24:219–234 Watkiss BM (2008) The Sleuth urban grow model as a forecasting and decision making tool. MSC Thesis, The University of Stellenbosch. Online at: http://www.ncgia.ucsb.edu/projects/gig/ Repository/references/SLEUTHPapers_Nov24/Watkiss,%20BM. pdf White R, Engelen G (1993) Cellular automata and fractal urban form: a cellular modelling approach to the evolution of urban land use pattern. Environ PlannA 25:1175–1199 White R, Engelen G (1994) Urban systems dynamics and cellular automata: fractal structures between order and chaos. Chaos Soliton Fract 4(4):563–583 Wu F (1996) A linguistic cellular automata simulation approach for sustainable land development in a fast growing region. Comput Environ Urban 20(6):367–387 Wu N, Silva EA (2009) Integration of genetic agents and cellular automata for dynamic urban growth modeling. Online at: http:// www.geocomputation.org/2009/PDF/Wu_and_Silva.pdf Xi F, Hu Y, Hung S, Wu X, Bu R, Chang Y, Liu M, Yu J (2009) Simulate urban growth based on RS, GIS, and SLEUTH model in Shenyang-Fushun metropolitan area northeastern China. 2009 Urban Remote Sensing Joint Event. Online at: http://www.ncgia. ucsb.edu/projects/gig/Repository/references/SLEUTH/shenyangfushun,%20china.pdf Yang X, Lo CP (2003) Modeling urban growth and landscape change in the Atlanta metropolitan area. Int J Geogr Inf Sci 17:463–488 Yeh AGO, Li X (2001) A constrained CA model for the simulation and planning of sustainable urban forms by using GIS. Environ Plann 28(5):733–753 Youssef AM, Pradhan B, Tarabees E (2011) Integrated evaluation of urban development suitability based on remote sensing and GIS techniques: contribution from the analytic hierarchy process. Arabian J Geosci 4(3–4):463–473 Zeug G, Eckert S, Steiner U, Kukuk T, Ehrlich D (2006) Monitoring urban growth and its impact on the environment: the case of Sana’a, Yemen. In: Ehlers et al. (eds) Proceedings digital earth summit on geoinformatics 2008, pp 206–211

123