Applying SLEUTH for simulating urban expansion of Hangzhou

Yong Liu∗, Xiuhua Liu

College of Resources and Environment, Southwest University, Chongqing, 400716, China

ABSTRACT Urbanization is found to be closely associated with land use/land cover change which has an important influence in our environment and ecosystems, such as urban heat island effect, biodiversity loss, soil erosion, and pollutions. Studies on accurately simulating urban expansion have been inspired by increasing concerns of the sustainability of urban development. This paper reports our research aiming to simulate the expansion of Hangzhou city using SLEUTH (slope, landuse, exclusion, urban extent, transportation and hillshade) urban growth model. In this research, we investigates the urban spatial growth patterns based on Landsat Thematic Mapper (TM)/Enhanced Thematic Mapper Plus (ETM+) images and creates four land cover change scenarios in 2020 with different socio-economic conditions. The results show that the SLEUTH model is less effective for depicting wave-like urban growth. From the four projected scenarios, urban area in this city will increase linearly and the shape of the city continues to be multi-nuclei in 2020. The hotspot area featured by intensive urban growth would, however, shift from urban center to sub-centers.

Keywords: land use/land cover change, urban expansion, SLEUTH, cellular automata

1. INTRODUCTION Land Use/Cover Change (LUCC) has gained increasing attention in past three decades [1, 2]. Urban land cover occupies less than 2% of the earth’s land surface, but it plays an important role in our ecosystem functions, such as global and local climate, hydrological cycle, and biodiversity. Estimated by the United Nations, the percentage of urban population will increase from 50% at present to 61% in 2030 [3]. The increase of urban population will result in a rapid land transformation from rural to urban areas in the near future. Both qualitative description model and quantitative prediction model are essential to effectively simulate and forecast urban growth and landscape changes. Among existing urban growth models, Cellular Automata (CA) models are robust in investigating the dynamics of urban development through simple local transitional rules [4, 5, 6]. They are useful for simulating urban growth, particularly in a complex environment. Most studies on CA model focus on the design and implementation of the model. However, less attention has been paid to the use of CA model for the assessment of land use policy [7]. Specifically, little research has been done to investigate the impacts of institutional and cultural factors on urban growth and landscape changes by comparing scenarios under different socio-economic conditions in developing countries [8].

SLEUTH is one of the most popular CA models, which consists of an urban growth modeling module and a land-cover change transition model. SLEUTH model was originally developed by Keith Clarke, and named after its six input layers, including Slope, Land use, Exclusion, Urban extent, Transportation and Hillshade [9, 10, 11]. This model has been successfully used in urban growth simulation, land use policy assessment and land cover change scenario optimization in developed countries [11, 12, 13]. Unfortunately, this model has rarely been used to model the development of cities in China [14, 15]. This paper employs SLEUTH model to project four urban growth scenarios under different socio-economic conditions by using Hangzhou as a case study. The results are useful for local agencies and planners making wise land use policy [16, 17, 18].

2. STUDY AREA Hangzhou is located in the southern part of Yangtze River Delta in Zhejiang Province. The geography of this city is characterized by the junction of plains and mountains. Hangzhou is well known as a tourist city. The climate is warm and humid with four distinct seasons. The Hangzhou metropolitan area contains eight districts with 4,015,900 population. The total area is 3,068 sq km. The study area covers a main center and three sub-centers, namely Xiasha, Jiangnan and Linping, with an area of 1,067 sq km (Fig.1). ∗ ly6505@163.com; phone +86 150 8698 3180; fax +86 23 6825 0844

Second International Conference on Earth Observation for Global Changes, edited by Xianfeng Zhang, Jonathan Li, Guoxiang Liu, Xiaojun Yang,

Proc. of SPIE Vol. 7471, 747104 · © 2009 SPIE · CCC code: 0277-786X/09/$18

Proc. of SPIE Vol. 7471 747104-1

Fig. 1. Location of the study area

3. METHODS 3.1 Data processing

The image data used in this research were four Landsat TM/ETM+ images acquired in 1991, 1996, 2000, and 2005. Three main procedures were involved in the image processing. They were geometric correction, image classification through the integration of supervised automated classification and manual interpretation, and accuracy assessment. The land cover classification scheme adopted here included four major classes, namely urban land, cultivated land, forest, and water. As a result, temporal land cover maps were derived from these Landsat images. Also, road maps in 1991, 1996, and 2000 were delineated from the image data. The road map in 2005 was provided by the transportation agency.

In addition, a Digital Evaluation Model (DEM) was used to extract slope and hillshade information via ArcGIS 9.0. An exclusion layer, which contains of Qiantang River, West Lake, Xixi Wetland and Beijing-Hangzhou Grand Canal, was produced to exclude these areas from urban development in the SLEUTH model.

3.2 SLEUTH model and scenarios assumption

SLEUTH model was used to simulate four types of urban growth: spontaneous growth, new spreading center growth, edge growth, and road-influenced growth. These four growth types were determined by the interactions of five growth coefficients: diffusion, breed, spread, road gravity, and slope (Table 1). The five coefficients, coupled with the excluded layer, potentially determined the probability of any given location being developed [12]. The implementation of the model involved two general phases: 1) calibration. It used historical urban growth patterns to calibrate the SLEUTH urban growth model; (2) prediction. It employed the calibrated model to forecast the future urban development. 'Self-modification' function in SLEUTH model allowed changing the growth coefficients throughout the course of a model run. This function made the SLEUTH model more realistically simulate the different rates of growth [12].

The excluded layer served as the primary factor which represented different land use policies. If a cell in the excluded layer has the exclusion value of 50, it suggests that this cell has 50% probability of being excluded from being developed. In addition, future transportation networks were created and used in the SLEUTH model.

Through implementing the designed SLEUTH model, four urban growth scenarios in 2020 were simulated: Maintain Status Quo (MSQ), Expanded Roads (ER), Moderate Farmland Protection (MFP) and Compact City (CC). The MSQ scenario was produced with the assumptions of no change in current land use policy and no limitation of urban expansion except large water surface. The ER scenario considered the effect of road gravity on urban growth pattern. The construction and widening of major roads in urban planning were involved into the simulation of urban growth in 2020. The MFP scenario reflected the influence of farmland protection policy on urban expansion. In the excluded layer,

Proc. of SPIE Vol. 7471 747104-2

farmland has exclusion value of 60, which suggests a moderate level of protection. The last scenario of CC indicated a compact mode of urban development. In the excluded layer, buffer zones between 1 km and 2 km from the existing urban area have exclusion value of 50; zones beyond 2 km from the existing urban area have exclusion value of 70; areas around dispersed urban patches whose size is less than 10 ha have exclusion value of 100.

Table 1. Summary of growth types by the SLEUTH model (adapted from the reference [12]).

Growth cycle order

Growth type Controlling coefficients Summary description

1 spontaneous diffusion Randomly selects potential new growth cells.

2 new spreading center breed Growing urban centers from spontaneous growth.

3 edge spread Old or new urban centers spawn additional growth.

4 road-influenced road-gravity, diffusion, breed Newly urbanized cell spawns growth along road network.

Throughout slope resistance slope Effect of slope on reducing probability of urbanization.

Throughout excluded layer user-defined User specifies areas resistant or excluded to development.

3.3 Buffer analysis

To explore the impact of CBD (Central Business District) on urban development, eight buffer zones with 2 km interval were generated around Wuling Square which is the CBD of Hangzhou [19]. Also to investigate the spatial pattern of newly developed land, 15 buffer zones with 200 m interval were generated from four urban fringes to outside. The urban growth intensity per year (LTI) was calculated using the following formula [20]:

( ) %1001××

−=

TAAALTI ab (1)

where Aa and Ab are the urban areas in the starting year a and the ending year b, respectively; A is the total land area; T is the time span from the year a and year b.

4. RESULTS 4.1 Results of SLEUTH model

4.1.1 Calibration results

Compare statistic is a ratio of the amount of projected urban pixels by SLEUTH model and the amount of real urban pixels in the ending year. Lee-Sallee statistic is a ratio of the intersection and the union of the projected and real urban areas. During calibration, the above statistics showed the trend of ascending as expected. The Compare and Lee-Sallee statistics were up to 0.95 and 0.59 in the final calibration, respectively, which indicated a good fit of urban area and spatial pattern since the Lee-Sallee statistic seldom approached 0.6 in recent applications of SLEUTH [12]. The calibrated parameter values were: diffusion = 7, breed = 58, spread = 76, road gravity = 75, slope resistance = 65. The visual output of calibration was shown in Fig.2.

From the results of calibration, the projected urban area was larger than real urban area. A ratio of projected urban area and real urban area was 1.4 in 1996 and it decreased to 1.08 in 2005. This trend suggested that projected urban development was more consistent with real situation in 2005 than that in the past. The growth rates were 10.6%, 15.6% and 7.3% for three periods of 1991-1996, 1996-2000 and 2000-2005. The projected growth rates by SLEUTH model were 23.6%, 8.2% and 5.4%. This growth rate disparity indicated that SLEUTH model was less effective to simulate wave-like and nonlinear urban expansion during economic transitional period in Hangzhou.

In addition, polycentric urban form was found in Hangzhou due to linear growth along roads and edge growth around existing urban centers. However, dispersed urban growth pattern occurred in calibrated result for 1996. Finally, it is clear that SLEUTH model was less powerful to simulate the new type of urban development, such as State-Approved Development Zones of Xiasha subcenter. This may be explained by the fact that hotspots of urban development changed frequently due to the complex behaviors of local government and developers in transitional period.

Proc. of SPIE Vol. 7471 747104-3

Fig.2 Visual output of calibration results and actual situation in Hangzhou

4.1.2 Prediction results

Four scenarios in 2020 by SLEUTH model were shown in Fig.3. From this Figure, large amount of edge growth would occur around the existing urban centers. The MSQ had the largest newly developed urban area, mainly along road corridors, leading to urban sprawl in the future. The ER had limited newly developed urban area by narrowing down urban development towards those areas which have access to transportation infrastructure and services, such as the west of Linping center and the south of Main center. The MFP prevented farm lands in the north of Linping center and the south of Jiangnan center from urban encroachment. The CC had the most compact and regular urban morphology among four scenarios, which was regarded as a good urban form by urban planners.

4.2 Temporal process of urban expansion

4.2.1 Growth of urban area

Based on temporal Landsat images, urban area increased linearly between 1991 and 2005. From Fig.4 (a), urban area was 125 sq km in 1991 which was 11.7% of the total area. It increased to 426 sq km in 2005 which was 40% of the total area and was 3.4 times of that in 1991. In 2020, urban area will be up to 596 sq km by averaging the urban area in the four scenarios. Newly increased urban area in MSQ was as high as 189 sq km, while that in ER was cut down by 35 sq km by incentive polices. Those in MFP and CC were cut down by 18 sq km and 26 sq km respectively due to restrictive policies. An interesting finding is that urban growth pattern in Hangzhou was shifting from one main center to multiple subcenters. The newly increased urban area of the main center was larger than that of subcenters in the past. However, in ER, the newly developed area of the main center was decreased to 33% of the total new urban area. The consumption of cultivated land, forest, and water has been shown in Fig.4 (b). The loss of cultivated land area in ER and CC was at a speed of 7.5 sq km per year while the losses in MSQ and MFP were at a speed of 9 sq km and 8 sq km per year, respectively. Fragmentation of cultivated land earns increasing attentions during urban land conversion in the future.

Proc. of SPIE Vol. 7471 747104-4

Fig.3 Four projected scenarios of urban expansion in 2020

(a)

0

200

400

600

1991 1996 2000 2005 2020Year

Urb

an la

nd a

rea(

sq k

m)

(b)

0

40

80

120

160

MSQ ER MFP CC

Land

con

sum

ptio

n (s

q km

) Cultivated land Forest Water

Fig.4 Projected urban area and land consumption in four scenarios, the left figure (a) shows urban area between 1991 and

2005, the right figure (b) shows possible land consumption in scenarios between 2005 and 2020

Proc. of SPIE Vol. 7471 747104-5

4.2.2 Phases of urban expansion

Historically, urban land expanded along the north-south axis in the east of West Lake, leading to narrow urban form of Hangzhou [21, 22]. Recently, the urban form changed gradually from “palm shape” to “fan shape” [17]. As was seen from satellite imagery, residential land expanded towards suburban area along major roads, Beijing-Hangzhou Grand Canal and Qiantang River in the early 1990s. Industrial sites were relocated in the north nearby urban built-up area during the same period. The primary urban development plan of Hangzhou was revised in 1996, which proposed to build a polycentric city through developing Jiangnan and Xiasha subcenters [23]. The two districts adjacent to northern and southern suburban areas were annexed into Hangzhou in 2001. According to the scenarios by SLEUTH model (Fig.3), Hangzhou will witness dispersed urban form through polycentric urban development, which terms the era of ‘one main center and three subcenters’. There will be three hot-spots of the growth in the near future. The first hot-spot is both sides of Highway for residential development; The second spot is the west of the main center and Xiasha subcenter for university town; The last one is Qianjiang century CBD for commercial development.

4.3 Spatial analysis of urban expansion

As is seen from Fig.5 (a), the growth intensity curves of urban expansion from CBD had a peak at 3 km before 2000, but while the curve approached a peak at 6 km in 2005. These curves reflected an important impact of CBD on urban expansion. In Fig.5 (b), the growth intensity curves showed a peak at 12 km with proximately normal distribution in four scenarios. The decline of growth intensity to 1.5% indicated the weakening impact of CBD in the future. MSQ had higher growth intensity around the main center than others.

(a)

0

1

2

3

4

5

1 2 3 4 5 6 7 8Distance to CBD (km)

Gro

wth

inte

nsity

from

CB

D (%

)

1991-19961996-20002000-2005

(b)

0

0.5

1

1.5

2

2 4 6 8 10 12 14 16Distance to CBD (km)

Gro

wth

inte

nsity

from

CB

D (%

)MSQ ERMFP CC

Fig.5 Spatial differences of growth intensity among the buffer zones from CBD. The left figure (a) shows growth intensity between

1991 and 2005; The right figure (b) shows growth intensity of scenarios between 2005 and 2020

The growth intensity from urban fringe to outside showed a different trend. During the study period, the growth intensity declined in power function with an inflection at 500 m from urban fringe. According to Fig.6 (a), the power of the function was -0.528 for the period of 1991-1996, reflecting compact urban form. The power increased to -0.211 for 1996-2000, reflecting leapfrog urban development. It was -0.309 for 2000-2005, suggesting constrains of urban sprawl. In Fig.6 (b), each scenario had a power function with the power ranging from -0.3 to -0.7, which indicated dispersed or compact urban form. Specifically, CC with a power of -0.7 had compact urban form while MSQ with a power of -0.3 had extended urban area. Fortunately, the average growth density of different scenarios in the future will be lower than that in the past.

4.4 Policy assessment

Based on different assumptions, SLEUTH model has been used to generate four scenarios for policy assessment. For example, ER got the roads in transportation plans involved into the simulation of urban expansion and MFP considered the effect of the farmland protection policy. From the results, MSQ showed a dispersed urban pattern. Primary urban development planning or land use planning may help save farmland and adjust urban morphology; MFP restricted urban development by involving farmland protection policy; ER had the least impervious surfaces [24] by clustering population and industry in subcenters; CC showed more compact polycentric urban form than others. It is essential to make use of

Proc. of SPIE Vol. 7471 747104-6

the advantages of ER, MFP and CC to compromise multiple objectives of urban planning, such as urban growth, farmland protection, and ecological maintenance.

(a)0

2

4

6

8

10

12

0 0.5 1 1.5 2 2.5 3Distance to urban fringe (km)

Gro

wth

inte

nsity

from

urba

n fr

inge

(%)

1991-19961996-20002000-2005

(b)

0

1

2

3

4

0 0.5 1 1.5 2 2.5 3Distance to urban fringe (km)

Gro

wth

inte

nsity

from

urba

n fr

inge

(%)

MSQ ER MFP CC

Fig.6 Spatial differences of growth intensity among the buffer zones from urban fringe. The left figure (a) shows growth intensity

between 1991 and 2005; the right figure (b) shows growth intensity of scenarios between 2005 and 2020

5. CONCLUSIONS This paper has demonstrated the differences and similarities among different scenarios by SLEUTH model. Based on temporal land cover maps derived from remotely sensed data, this paper used SLEUTH model to simulate the urban development of Hangzhou in the past. Through the calibration, the suitability of SLEUTH model for the simulation of urban expansion has been discussed. Since the Compare statistic and Lee-Sallee statistic were up to 0.95 and 0.59 respectively, SLEUTH model was found to be useful to simulate urban growth in Hangzhou. But some assumptions about neighbor rules prevented SLEUTH from modeling government-led behaviors such as establishment of new subcenter.

The experiments in this paper show that SLEUTH model is useful for policy assessment. In this research, four scenarios of MSQ, ER, MFP, and CC were projected. The MSQ scenario showed that 189 sq km land would be developed, which had negative impacts on cultivated land and ecological land. The other scenarios represented urban growth under incentive or restrictive policies. From the results, spatial-temporal patterns of urban growth in these scenarios were found to be similar with those in the past. However, the growth intensity decreased and hotspot area moved towards subcenters in the future.

ACKNOWLEDGEMENTS

This work was supported by the Ph.D. Foundation of Southwest University (No. SWUB2008064) and National Key Technology R&D Program of China (No. 2007BAD87B10).

REFERENCES

[1] Turner, B. L. and Meyer, W. B., "Global land use and land cover change: an overview", in Meyer, W. B. and Turner, B. L., "Changes in land use and land cover: a global perspective", Cambridge University Press, New York, 1-11(1994). [2] Lambin E. F., Turner B. L., Geist H. J., Agbola S. B., Angelsen A., Bruce J. W., Coomes O. T., Dirzo R., Fischer G., Folke C., George P. S., Homewood K., Imbernon J., Leemans R., Li X., Moran E. F., Mortimore M., Ramakrishnan P. S., Richards J. F., Skanes H., Steffen W., Stone G. D., Svedin U., Veldkamp T. A., Vogel C. and Xu J., "The causes of land-use and land-cover change: moving beyond the myths", Global Environ. Change. Papers 11(4), 261-269 (2001). [3] United Nations, "World urbanization prospects: the 2003 revision", United Nations Press, New York (2004).

Proc. of SPIE Vol. 7471 747104-7

[4] Batty, M., "Urban evolution on the desktop: simulation with the use of extended cellular automata", Environ. Plan. A. Papers 30(11), 1943-1967 (1998). [5] Wu, F., "Calibration of stochastic cellular automata: the application to rural-urban land conversions", Int. J. Geogr. Inf. Sci. Papers 16(8), 795-818 (2002). [6] Li, X., Ye, J., Liu, X. P. and Yang Q. S., "Geographical simulation systems, cellular automata and multi-agents systems". Science Press, Beijing (2007). [7] Torrens, P. M. and O'Sullivan, D., "Cellular automata and urban simulation, where do we go from here? ", Environ. Plan. B-Plan. Des. Papers 28(2), 163-168 (2001). [8] Fragkias, M. and Seto, K. C., "Modeling urban growth in data-sparse environments, a new approach", Environ. Plan. B-Plan. Des. Papers 34(5), 858-883 (2007). [9] Clarke, K. C., Hoppen, S. and Gaydos, J., "A self-modifying cellular automaton model of historical urbanization in the San Francisco Bay area", Environ. Plan. B-Plan. Des. Papers 24(2), 247-261 (1997). [10] Silva, E. A. and Clarke, K. C., "Calibration of the SLEUTH urban growth model for Lisbon and Porto, Portugal", Comp. Environ. Urban Sys. Papers 26(6), 525-552 (2002). [11] Clarke, K. C. and Gaydos, L. J., "Loose-coupling a cellular automaton model and GIS, long-term urban growth prediction for San Francisco and Washington/Baltimore", Int. J. Geogr. Inf. Sci. Papers 12(7), 699-714 (1998). [12] Jantz C. A., Geog D., Goetz S. J., and Shelly M., "Using the SLEUTH urban growth model to simulate the impacts of future policy scenario on urban land use in the Baltimore-Washington metropolitan area", Environ. Plan. B-Plan. Des. Papers 31(2), 251-271 (2004). [13] Ding, H., "Land use/Land cover changes, driving force analysis and prediction in the developed coastal region in China, a case study in Zhejiang province", Ph. D. Dissertation, Zhejiang University, Hangzhou (2006). [14] Wu, X. B., Zhao, J., Wei, C. J., Huang L. F., Yan S. Y. and Xiao C. S., "The urban expanding simulation with the cellular automata model in Haikou", Urban Planning. Papers 26(8), 69-73 (2002). [15] Zhang, Y., Li, J. and Cheng, Y. H., "Simulation of Beijing urbanization using SLEUTH", Remote Sens. Inf. Papers (2), 50-54 (2007). [16] Feng, J., "Spatial-temporal evolution of urban morphology and land use structure in Hangzhou", Acta Geogr. Sinica. Papers 58(3), 343-353 (2003). [17] Zhou, G. P., "Presumption of urban spatial structure in Hangzhou", Zhejiang Construction. Papers 23(11), 1-10 (2006). [18] Li, W. M. and Li, S. B., "Research on the characteristic and the strategy of the development of new town in Hangzhou metropolitan area", J of Zhejiang University (Science Edition). Papers 32 (1), 108-114 (2005). [19] Xu C., Liu M., Zhang C., An S., Yu W., and Chen J., "The spatiotemporal dynamics of rapid urban growth in the Nanjing metropolitan region of China", Landscape Ecology. Papers 22(6), 925-937 (2007). [20] Zhu, H. Y. and Li, X. B., "Discussion on the index method of regional land use change", Acta Geogr. Sinica. Papers 58(5), 643-649 (2003). [21] Yao, S. M., "Urban expansion in metropolitan areas of China", China Science & Technology Press, Hefei, 465-483 (1997). [22] Feng, J. "Restructuring of urban internal space in China in the transition period", Science Press, Beijing (2004). [23] Department of Hangzhou Planning and Center of Urban Planning in Hangzhou, "Stepping into the Era of Qiantang River", Tongji University Press, Shanghai (2002). [24] Xian, G. and Crane, M., "Assessments of urban growth in the Tampa Bay watershed using remote sensing data", Remote Sens. Environ., 97(2), 203-215 (2005).

Proc. of SPIE Vol. 7471 747104-8