- Email: [email protected]
Simulation of land use spatial pattern of towns and villages based on CA–Markov model
Mathematical and Computer Modelling 54 (2011) 938–943
Contents lists available at ScienceDirect
Mathematical and Computer Modelling journal homepage: www.elsevier.com/locate/mcm
Simulation of land use spatial pattern of towns and villages based on CA–Markov model Lingling Sang a , Chao Zhang a,∗ , Jianyu Yang a , Dehai Zhu a , Wenju Yun b a
College of Information and Electrical Engineering, China Agricultural University, Beijing, 100083, PR China
b
Land Consolidation and Rehabilitation Center, The Ministry of land Resources, Beijing 100035, PR China
article
info
Article history: Received 11 August 2010 Accepted 4 November 2010 Keywords: Land use change Spatial pattern Markov Cellular automata Fangshan district in Beijing
abstract Firstly, this paper analyzes the basic principles and processes of the spatial pattern changes of land use in towns and villages, and the result shows that the land resource demands of urban development and population growth lead to the spatial pattern changes. Secondly, in order to grip land use changes better, the paper proposes a method for the simulation of spatial patterns. The simulating method can be divided into two parts: one is a quantitative forecast by using the Markov model, and the other is simulating the spatial pattern changes by using the CA model. The above two models construct the simulative model of the spatial pattern of land use in towns and villages. Finally, selecting Fangshan which is a district of Beijing as the experimental area, both the quantity and spatial pattern changing characteristics are investigated through building a changing dataset of land use by using spatial analysis methods based on the land use data in 2001, 2006 and 2008; CA–Markov is used to simulate the spatial pattern of land use in Fangshan for 2015. © 2010 Elsevier Ltd. All rights reserved.
1. Introduction Rapid advances in geospatial models have made it increasingly possible to design and simulate spatial patterns for land use change (LUC). An approach commonly used in the simulation model for LUC is based on cellular automata theory and the GIS framework [1,2]. So far, the establishment of simulation models to monitor and adjust the changes of land use is a local and global issue of great concern. Mainly simulation methods include two aspects: the first one is the number of prediction, and the other one is simulating for the spatial patterns of the future. Spatial patterns of land use changes in the dynamic simulation and analysis relating to the impact of land use changes in many drivers is a very complex process [3]. Models based on the simulation of the spatial pattern of land use change processes are used to understand and explain the process of regional land use changes and trends in effective ways. However, the current real land use changes and the spatial distribution of the combination in different scales of land use, spatial and temporal dynamic model of the evolution rule are rare. Both cellular automata (CA) and the Markov model have great advantages in the study on land use changes, while both of them have respective disadvantages. The Markov model for land use changes has been widely used, but with the traditional Markov model it is difficult to predict the spatial pattern of land use changes. The CA model with powerful spatial computing can be used to simulate the spatial variation of the system effectively. A CA–Markov model is a robust approach in the spatial and temporal dynamic modeling of land use changes because geographic information systems (GIS) and remote sensing (RS) can be efficiently incorporated [4]. The CA–Markov model absorbs the benefits from the time series and spatial
∗
Corresponding author. Tel.: +86 10 62737855; fax: +86 10 62737855. E-mail address: [email protected] (C. Zhang).
0895-7177/$ – see front matter © 2010 Elsevier Ltd. All rights reserved. doi:10.1016/j.mcm.2010.11.019
L. Sang et al. / Mathematical and Computer Modelling 54 (2011) 938–943
939
predictions of the Markov and CA theory, and it can be used to carry out the Spatial–Temporal Pattern stimulation. The CA–Markov model also considers the land use changes’ suitability and the effect of natural, societal and economic factors about land use changes. Based on previous research and the recent trend towards LUC, this paper analyzed the land use changes for Fangshan district in Beijing which were based on the quantity of land use change characteristics of the structure and spatial pattern as the primary research content. The CA–Markov model combined with GIS is selected. The complex systems of spaces, to play its ability to change and Markov model for long-term forecasts of the advantages of both types of land use conversion increased prediction accuracy, and it can effectively simulate the spatial pattern of land use changes, with great science and practice. 2. Study area and data 2.1. Study area The Fangshan district is located in the southwest of the capital city (39°30′ –39°55′ S, 115°25′ –116°15′ E). The study area covers an area of approximately 198, 954.4 hm2 . There are a total of 22 district towns and more than 500 villages, and the population is nearly 90 million. The township in the district has a representation of urban fringe villages. States of land use in Beijing, as the edges of the area of Beijing are impacted by urbanization. Forest is a predominant land use accounting for 42.26% of the study area in 2001. According to the ‘‘Beijing Urban Master Plan (2004–2020)’’ which was approved by the State Council, Fangshan has been identified as the capital of urban development zone, and will become one of the economic centers of Beijing, which is the main carrier for development of modern manufacturing and modern agriculture, and becomes a center of an important regional industry and population. 2.2. Data sources and processing The Fangshan district in Beijing was selected as a study area because of its land use complexity and data availability. The main data sources are digital land use data (2001, 2006 and 2008), which were sourced from the local land department. Land use maps in 2001, 2006 and 2008, and land use change maps (2001–2006, 2001–2008) were produced at 1:5000 from Spatial overlay analysis. The digital map data for the study area were vectored to a manageable spatial resolution and contain seven classes: arable field, garden, woodland, grassland, construction land, waters and other unused land. The detailed processing procedure is as follows: (1) Firstly, the vector land use data at the three times were separately converted into raster data (grid units of 100 m) using the spatial analysis module by ArcGIS9.2. (2) Secondly, through the reclassifying module, all land use types were aggregated to the primary level. They were then used to derive the land changes/conversion matrix using a raster calculator. (3) Finally, models and quantitative methods of land use changes were applied to the derived land use change maps at different periods in an effort to determine the spatiotemporal pattern of land use changes in Fangshan. 3. Research methods 3.1. Markov model The Markov model is a theory based on the process of the formation of Markov random process systems for the prediction and optimal control theory method [5]. The Markov model not only explains the quantification of conversion states between the land use types, but can also reveal the transfer rate among different land use types. It is commonly used in the prediction of geographical characteristics with no aftereffect event which has now become an important predicting method in geographic research. Based on the conditional probability formula—Bayes, the prediction of land use changes is calculated by the following equation [3,5,6]: S (t + 1) = Pij × S (t )
(1)
where S (t ), S (t + 1) are the system status at the time of t or t + 1; Pij is the transition probability matrix in a state which is calculated as follows [3]:
P11 P21
P12 P22
0 ≤ Pij < 1 and
∑N
Pij = ··· Pn1
··· Pn2
··· ··· ··· ···
P1n P2n
(2)
· · ·
Pnn
j=1 Pij = 1, (i, j = 1, 2, . . . , n) .
3.2. CA model The behavior of CA models is affected by uncertainties arising from the interaction between model elements, structures, and the quality of data sources used as the model input [7,8]. It focuses mainly on the local interactions of cells with
940
L. Sang et al. / Mathematical and Computer Modelling 54 (2011) 938–943
distinct temporal and spatial coupling features and the powerful computing capability of space, which is especially suitable for dynamic simulation and display with self-organizing feature systems. The use of geographic cellular automata for land use change simulations not only takes into account comprehensive consideration soil conditions, climatic conditions, topography and other natural factors, but also considers a comprehensive policy, economy, technology and other human factors, and takes into account the historical trends of land use with strong applicability. The CA model can be expressed as follows [3]: S (t , t + 1) = f (S (t ), N )
(3)
where S is the set of limited and discrete cellular states, N is the Cellular field, t and t + 1 indicate the different times, and f is the transformation rule of cellular states in local space. 3.3. CA–Markov model CA–Markov is a combined Cellular Automata/Markov Chain/Multi-Criteria/Multi-Objective Land Allocation (MOLA) land cover prediction method that adds an element of spatial contiguity as well as knowledge of the likely spatial distribution of transitions to Markov chain analysis. The Markov model focuses on the quantity in predictions for land use changes. For this model, the spatial parameters are weak and do not know the various types of land use changes in the spatial extents [9]. The CA model has a strong space conception, which is a strong capability of space-time dynamic evolution with complex space systems. The CA–Markov model, which incorporates the theories of Markov and CA, is about the time series and space for the advantages of forecasting. It can achieve better simulation for temporal and spatial patterns of land use changes in quantity and space [10,11]. The CA–Markov module in IDRISI32 integrates the functions of cellular automaton filter and Markov processes, using conversion tables and conditional probability of the conversion map to predict the states of land use changes, and it may be better to carry out land use change simulations. The CA–Markov model to simulate land use changes has been put into use in this paper. Firstly, converting the vector data into raster data; Secondly, based on the CA–Markov module—GIS and image processing module in IDRISI software, spatial distribution of land use is achieved through the Markov model analysis of land use trends and the use of CA modeling. The specific process is as follows: (1) Determining the transition rules. With Markov chain analysis, future land use changes can be modeled on the basis of the preceding state; that is, a matrix of observed transition probabilities between maps in 2001 and 2008 can be used to project future changes from current patterns. Through spatial overlay analysis, the transition probability matrix and the transfer area of the matrix are achieved. Among them, the transition probability matrix reflects the various land use types into other types of probability; the transfer area of the matrix reflects the land use conversion to other land use types in the expected area in the next period. Note that the baseline is the land use map of 2008, which is superimposed with the map in 2001. The calculated transition probability matrix will serve as the transformation rules to put CA–Markov model simulations into practice. (2) Determining CA filters. CA filters can produce a clear sense of the space weighting factor, which can be changed according to the current adjacent cellular state. The standard 5 × 5 contiguity filter is used as the neighborhood definition in this study. That is, each cellular center is surrounded by a matrix space which is composed by 5 × 5 cellular to impact the cellular changes significantly. (3) Determining the starting point and the CA number of iterations. The study takes the year 2008 as a starting point. The number of CA iterations is selected at 15 in order to simulate the landscape spatial pattern for the study area in 2015. Using CA–Markov modules in IDRISI, the land use change simulation technical route is shown in Fig. 2. 4. Results and analysis 4.1. Reclassification of land use types Based on the land use maps in 2001 and 2008, the types of land use were reclassified with reference to remote sensing image characteristics. The land use classification adopted in the model consists of seven land use classes. Seven of these, namely: cultivated field, garden, woodland, grassland, construction land, waters and other non-utilization land are included. The result of the new reclassification is consistent with national land use classification criteria. The major tourist attractions and grassland will be added into other non-utilization land in the new classification. The quantity and spatial pattern of other non-utilization land basically remained stable. After reclassification, special lands were classified into other land use types according to basic characteristics of land use/land cover, which includes urban, woodland, water or independent industrial and mining. The result of the reclassifications and their distribution is as shown in Fig. 3. 4.2. Land use structure changes According to the spatial overlay analysis in GIS, from 2001 to 2008, there are changes in the land use types over these 7 years. The changes in area are as shown in Table 1.
L. Sang et al. / Mathematical and Computer Modelling 54 (2011) 938–943
941
Fig. 1. Study area and location.
Fig. 2. Flowchart of the technical route of land use simulation.
a
b
Fig. 3. The re-classified grid land use maps (a) 2001 land use map (b) 2008 land use map. Note: AL—arable land; GAL—garden; WL—wood land; GL— grassland; WT—water (lakes, rivers and artificial ponds); CL—construction land; UL—unused land.
The main types of land use in Fangshan are arable land and construction land, which accounted for 58% of the total. Construction land (including urban construction, rural settlements and other construction land) in Fangshan accounted for 28.9% of the total area from 2001 to 2008; arable land, grassland and unused land have a corresponding reduction, which ranges up to a 29.1% reduction in arable land, reflecting the quick decline in arable land resources; Unused land decreased by 18.2%, indicating reserve land resources are inadequate in the study area; gardens and water have smaller changes in the total area. Land use changes are significant in Fangshan District from 2001 to 2008. Fig. 1 and Table 1 reflect the land use of space and volume changes from 2001 to 2008. Comparing the land use changes, the most obvious trend is the expansion of urban and industrial land. However, the expansion of urban and industrial land in space is very different. Land use dynamic changes can quantitatively describe the speed of land use changes, which plays a unique role in comparing the regional differences of land use changes and in predicting the trend of land use changes [12]. Land use dynamic changes also indicate the level of human disturbance to land use types. The higher the disturbance level, the more intensively land use changes. In this paper, the area changes of land use types in Fangshan during the two periods were further
942
L. Sang et al. / Mathematical and Computer Modelling 54 (2011) 938–943
Table 1 Area of land use changes in Fangshan during 2001–2008.
2001 (hm2 ) 2008 (hm2 ) Changed area (hm2 ) Change ratio (%)
AL
GAL
WL
GL
CL
WT
UL
32 079.82 22 751.87 −9 327.95 −29.1
9110.03 9575.54 465.51 5.1
72 661.85 81 758.28 9 096.43 12.5
26 046.03 20 013.50 −6 032.53 −23.2
33 187.16 42 787.70 9 600.54 28.9
6757.57 6432.92 −324.65 −4.8
19 111.94 15 634.59 −3 477.35 −18.2
Table 2 Transition probabilities matrix of land use types in Fangshan during 2001–2008 (units /%). Land use types
AL
GAL
WL
GL
CL
WT
UL
AL GAL WL GL CL WT UL
87.21 1.72 0.44 4.58 0.44 0.15 2.75
4.07 90.43 0.51 1.24 0.86 0.07 0.05
0.28 1.61 97.53 0.45 0.62 0.18 4.83
0 0.41 0 89.82 0 3.20 0
7.96 5.26 1.29 3.43 98.01 0.15 16.83
0.08 0.28 0.02 0.21 0.07 96.25 0
0.40 0.29 0.21 0.27 0 0 75.54
calculated. The land use changes at these two stages showed marked fluctuations. Arable land, woodland and unused land experienced a reversed trend of changes during the two periods. 4.3. Model application The following procedures were performed using algorithms available in IDRISI Kilimanjaro and Image Processing software in order to implement the CA–Markov model in the study area: (1) computation of land use transition potential maps based on procedure, (2) computation of transition probabilities using Markov chain analysis, and (3) spatial allocation of simulation land use probabilities base on a CA spatial filter. Land use pattern changes in 2015 were predicted using a CA–Markov model based on Fangshan District of Beijing in 2008 as an initial state, for which the time interval was 7 years. The Markov transition matrix can not only quantitatively explain the conversion between the types of land status, but also reveal the transfer rate between the different types. The conversion matrix of land use types during 2001–2008 in Table 2 can be summarized as follows: (1) Arable land had a massive loss as a result of conversion to other land types, in the form of conversion from arable land to garden and construction land, of which 7.96% was transformed to construction land and 4.07% into corner. The main reason for the reduction of arable land is urbanization and agricultural structure adjustment, while the supplement of arable land comes mainly from grassland and unused land. (2) The increase in woodland mainly came from garden and unused land, which accounted for 80.8% of the total conversion gain, 1.29% of which were transformed into construction land, 0.51% into corner, 0.44% transformed into cultivated land, which increased some of the major sources of arable land and grassland. (3) 5.26% of garden was transformed into construction land, 1.72% was transformed into arable land, and 1.61% into woodland. There was little change in the total area of Garden, but there were increases and decreases, and there were significant changes in the spatial distribution. (4) There was a grassland loss of 3.62%, which was due mostly to conversion to arable land, garden and construction land. At the secondary level of classification, conversion to woodland is the main reason for the shrinkage of dense cover grassland. (5) The conversion gain of waters was contributed mainly by grassland. The changes were small. (6) The area converted to construction land mainly encompassed rural settlements and urban construction, which took up lots of arable land, garden and woodland. And changes to other land use types are less. (7) The main conversion was other unused land into construction land, woodland and cultivated land. Studies have shown that unused land and forest land were the dominant land use types in Fangshan. Since 2001, farm land had been shrinking while forest land and built-up areas had increased quickly. Among the various transformation types, the changes from farm land to built-up land spread most significantly. Regions along railways and main roads as well as rivers with intense human activity had a high local variability of land use distribution. Therefore they became more varied and the spatial pattern of land use changes became more complex. The simulation result by the CA–Markov method showed that its original rate of changes in trends and changes will keep constant from 2008 to 2015. Therefore, it is urgent to strengthen the protection of farm land and water-bodies, to prevent acts of indiscriminate use of farmland in order to promote the protection of farmland and the rational use of land. 5. Conclusion Using land use maps (2001, 2006 and 2008), the CA–Markov model that combines the Markov chain analysis and CA models successfully simulated land use changes in Fangshan. The number and spatial distribution of the area is analyzed,
L. Sang et al. / Mathematical and Computer Modelling 54 (2011) 938–943
943
and the overall results are satisfactory: the seven land use types will maintain their changes in the direction and rates of change. Therefore, it is urgent to strengthen the protection of arable land and water and prevent acts of indiscriminate use of cultivated land in order to promote land protection and rational use of land. This study represents an important contribution to land use modeling as shown by the CA–Markov land use simulation model. In addition, the model considers only the surrounding natural environment, and the cellular is not moving. No consideration is given to playing a decisive role in land use dynamics changes of the social environment and their interactions. As a result, the human decision-making model for the simulation is still a weakness which needs further study. Acknowledgements This research was supported by National Science and Technology Ministry (No. 2006BAJ14B01-02) and Graduate Research and Innovation Subject in China Agricultural University (No. kycx09125). The authors also thank Miss Gao for comments on earlier versions of this paper. References [1] Verda Kocabas, Suzana Dragicevic, Assessing cellular automata model behaviour using a sensitivity analysis approach, Computers, Environment and Urban Systems 30 (2006) 921–953. [2] B.C. Pijanowski, D.G. Brown, G. Manik, Using neural nets and GIS to forecast land use changes: a land transformation model, Computers, Environment and Urban Systems 26 (6) (2002) 553–575. [3] X.Y. Hou, B. Chang, X.F. Yu, Land use change in Hexi corridor based on CA–Markov methods, Transactions of the CSAE 20 (5) (2004) 286–291. [4] Courage Kamusoko, Masamu Aniya, Bongo Adi, et al., Rural sustainability under threat in Zimbabwe-simulation of future land use/cover changes in the Bindura district based on the Markov-cellular automata model, Applied Geography 29 (2009) 435–447. [5] G.H. Jiang, F.R. Zhang, X.B. Kong, Determining conversion direction of the rural residential land consolidation in Beijing mountainous areas, Transactions of the CSAE 25 (2) (2009) 214–221. [6] G.Q. Yang, Y.L. Liu, Z.F. Wu, Analysis and simulation of land use temporal and spatial pattern based on CA–Markov model, Geomatics and Information Science of Wuhan University 32 (5) (2007) 414–418. [7] M. Batty, Yichun Xie, Zhanli Sun, Modeling urban dynamics through GIS-based cellular automata, Computers, Environment and Urban Systems 23 (1999) 205–233. [8] L.K. Peterson, K.M. Bergen, D.G. Browna, et al., Forested land-cover patterns and trends over changing forest management eras in the Siberian Baikal region, Forest Ecology and Management 257 (2009) 911–922. [9] Rohan Chandralal Wickramasuriya, Arnold K. Bregt, Hedwig van Delden, et al., The dynamics of shifting cultivation captured in an extended constrained cellular automata land use model, Ecological Modelling 220 (2009) 2302–2309. [10] X.L. Wang, Y.H. Bao, Study on the methods of land use dynamic change research, Progress in Geography 18 (1) (1999) 81–87. [11] H. Ji, Yoshitsugu Hayashi, C. Xin, et al., Application of an integrated system dynamics and cellular automata model for urban growth assessment: a case study of Shanghai, China, Landscape and Urban Planning 91 (2009) 133–141. [12] L.Y. Guo, D.L. Wang, J.J. Qiu, et al., Spatio-temporal patterns of land use change along the Bohai Rim in China during 1985–2005, Journal of Geographical Sciences 19 (2009) 568–576.
Contents lists available at ScienceDirect
Mathematical and Computer Modelling journal homepage: www.elsevier.com/locate/mcm
Simulation of land use spatial pattern of towns and villages based on CA–Markov model Lingling Sang a , Chao Zhang a,∗ , Jianyu Yang a , Dehai Zhu a , Wenju Yun b a
College of Information and Electrical Engineering, China Agricultural University, Beijing, 100083, PR China
b
Land Consolidation and Rehabilitation Center, The Ministry of land Resources, Beijing 100035, PR China
article
info
Article history: Received 11 August 2010 Accepted 4 November 2010 Keywords: Land use change Spatial pattern Markov Cellular automata Fangshan district in Beijing
abstract Firstly, this paper analyzes the basic principles and processes of the spatial pattern changes of land use in towns and villages, and the result shows that the land resource demands of urban development and population growth lead to the spatial pattern changes. Secondly, in order to grip land use changes better, the paper proposes a method for the simulation of spatial patterns. The simulating method can be divided into two parts: one is a quantitative forecast by using the Markov model, and the other is simulating the spatial pattern changes by using the CA model. The above two models construct the simulative model of the spatial pattern of land use in towns and villages. Finally, selecting Fangshan which is a district of Beijing as the experimental area, both the quantity and spatial pattern changing characteristics are investigated through building a changing dataset of land use by using spatial analysis methods based on the land use data in 2001, 2006 and 2008; CA–Markov is used to simulate the spatial pattern of land use in Fangshan for 2015. © 2010 Elsevier Ltd. All rights reserved.
1. Introduction Rapid advances in geospatial models have made it increasingly possible to design and simulate spatial patterns for land use change (LUC). An approach commonly used in the simulation model for LUC is based on cellular automata theory and the GIS framework [1,2]. So far, the establishment of simulation models to monitor and adjust the changes of land use is a local and global issue of great concern. Mainly simulation methods include two aspects: the first one is the number of prediction, and the other one is simulating for the spatial patterns of the future. Spatial patterns of land use changes in the dynamic simulation and analysis relating to the impact of land use changes in many drivers is a very complex process [3]. Models based on the simulation of the spatial pattern of land use change processes are used to understand and explain the process of regional land use changes and trends in effective ways. However, the current real land use changes and the spatial distribution of the combination in different scales of land use, spatial and temporal dynamic model of the evolution rule are rare. Both cellular automata (CA) and the Markov model have great advantages in the study on land use changes, while both of them have respective disadvantages. The Markov model for land use changes has been widely used, but with the traditional Markov model it is difficult to predict the spatial pattern of land use changes. The CA model with powerful spatial computing can be used to simulate the spatial variation of the system effectively. A CA–Markov model is a robust approach in the spatial and temporal dynamic modeling of land use changes because geographic information systems (GIS) and remote sensing (RS) can be efficiently incorporated [4]. The CA–Markov model absorbs the benefits from the time series and spatial
∗
Corresponding author. Tel.: +86 10 62737855; fax: +86 10 62737855. E-mail address: [email protected] (C. Zhang).
0895-7177/$ – see front matter © 2010 Elsevier Ltd. All rights reserved. doi:10.1016/j.mcm.2010.11.019
L. Sang et al. / Mathematical and Computer Modelling 54 (2011) 938–943
939
predictions of the Markov and CA theory, and it can be used to carry out the Spatial–Temporal Pattern stimulation. The CA–Markov model also considers the land use changes’ suitability and the effect of natural, societal and economic factors about land use changes. Based on previous research and the recent trend towards LUC, this paper analyzed the land use changes for Fangshan district in Beijing which were based on the quantity of land use change characteristics of the structure and spatial pattern as the primary research content. The CA–Markov model combined with GIS is selected. The complex systems of spaces, to play its ability to change and Markov model for long-term forecasts of the advantages of both types of land use conversion increased prediction accuracy, and it can effectively simulate the spatial pattern of land use changes, with great science and practice. 2. Study area and data 2.1. Study area The Fangshan district is located in the southwest of the capital city (39°30′ –39°55′ S, 115°25′ –116°15′ E). The study area covers an area of approximately 198, 954.4 hm2 . There are a total of 22 district towns and more than 500 villages, and the population is nearly 90 million. The township in the district has a representation of urban fringe villages. States of land use in Beijing, as the edges of the area of Beijing are impacted by urbanization. Forest is a predominant land use accounting for 42.26% of the study area in 2001. According to the ‘‘Beijing Urban Master Plan (2004–2020)’’ which was approved by the State Council, Fangshan has been identified as the capital of urban development zone, and will become one of the economic centers of Beijing, which is the main carrier for development of modern manufacturing and modern agriculture, and becomes a center of an important regional industry and population. 2.2. Data sources and processing The Fangshan district in Beijing was selected as a study area because of its land use complexity and data availability. The main data sources are digital land use data (2001, 2006 and 2008), which were sourced from the local land department. Land use maps in 2001, 2006 and 2008, and land use change maps (2001–2006, 2001–2008) were produced at 1:5000 from Spatial overlay analysis. The digital map data for the study area were vectored to a manageable spatial resolution and contain seven classes: arable field, garden, woodland, grassland, construction land, waters and other unused land. The detailed processing procedure is as follows: (1) Firstly, the vector land use data at the three times were separately converted into raster data (grid units of 100 m) using the spatial analysis module by ArcGIS9.2. (2) Secondly, through the reclassifying module, all land use types were aggregated to the primary level. They were then used to derive the land changes/conversion matrix using a raster calculator. (3) Finally, models and quantitative methods of land use changes were applied to the derived land use change maps at different periods in an effort to determine the spatiotemporal pattern of land use changes in Fangshan. 3. Research methods 3.1. Markov model The Markov model is a theory based on the process of the formation of Markov random process systems for the prediction and optimal control theory method [5]. The Markov model not only explains the quantification of conversion states between the land use types, but can also reveal the transfer rate among different land use types. It is commonly used in the prediction of geographical characteristics with no aftereffect event which has now become an important predicting method in geographic research. Based on the conditional probability formula—Bayes, the prediction of land use changes is calculated by the following equation [3,5,6]: S (t + 1) = Pij × S (t )
(1)
where S (t ), S (t + 1) are the system status at the time of t or t + 1; Pij is the transition probability matrix in a state which is calculated as follows [3]:
P11 P21
P12 P22
0 ≤ Pij < 1 and
∑N
Pij = ··· Pn1
··· Pn2
··· ··· ··· ···
P1n P2n
(2)
· · ·
Pnn
j=1 Pij = 1, (i, j = 1, 2, . . . , n) .
3.2. CA model The behavior of CA models is affected by uncertainties arising from the interaction between model elements, structures, and the quality of data sources used as the model input [7,8]. It focuses mainly on the local interactions of cells with
940
L. Sang et al. / Mathematical and Computer Modelling 54 (2011) 938–943
distinct temporal and spatial coupling features and the powerful computing capability of space, which is especially suitable for dynamic simulation and display with self-organizing feature systems. The use of geographic cellular automata for land use change simulations not only takes into account comprehensive consideration soil conditions, climatic conditions, topography and other natural factors, but also considers a comprehensive policy, economy, technology and other human factors, and takes into account the historical trends of land use with strong applicability. The CA model can be expressed as follows [3]: S (t , t + 1) = f (S (t ), N )
(3)
where S is the set of limited and discrete cellular states, N is the Cellular field, t and t + 1 indicate the different times, and f is the transformation rule of cellular states in local space. 3.3. CA–Markov model CA–Markov is a combined Cellular Automata/Markov Chain/Multi-Criteria/Multi-Objective Land Allocation (MOLA) land cover prediction method that adds an element of spatial contiguity as well as knowledge of the likely spatial distribution of transitions to Markov chain analysis. The Markov model focuses on the quantity in predictions for land use changes. For this model, the spatial parameters are weak and do not know the various types of land use changes in the spatial extents [9]. The CA model has a strong space conception, which is a strong capability of space-time dynamic evolution with complex space systems. The CA–Markov model, which incorporates the theories of Markov and CA, is about the time series and space for the advantages of forecasting. It can achieve better simulation for temporal and spatial patterns of land use changes in quantity and space [10,11]. The CA–Markov module in IDRISI32 integrates the functions of cellular automaton filter and Markov processes, using conversion tables and conditional probability of the conversion map to predict the states of land use changes, and it may be better to carry out land use change simulations. The CA–Markov model to simulate land use changes has been put into use in this paper. Firstly, converting the vector data into raster data; Secondly, based on the CA–Markov module—GIS and image processing module in IDRISI software, spatial distribution of land use is achieved through the Markov model analysis of land use trends and the use of CA modeling. The specific process is as follows: (1) Determining the transition rules. With Markov chain analysis, future land use changes can be modeled on the basis of the preceding state; that is, a matrix of observed transition probabilities between maps in 2001 and 2008 can be used to project future changes from current patterns. Through spatial overlay analysis, the transition probability matrix and the transfer area of the matrix are achieved. Among them, the transition probability matrix reflects the various land use types into other types of probability; the transfer area of the matrix reflects the land use conversion to other land use types in the expected area in the next period. Note that the baseline is the land use map of 2008, which is superimposed with the map in 2001. The calculated transition probability matrix will serve as the transformation rules to put CA–Markov model simulations into practice. (2) Determining CA filters. CA filters can produce a clear sense of the space weighting factor, which can be changed according to the current adjacent cellular state. The standard 5 × 5 contiguity filter is used as the neighborhood definition in this study. That is, each cellular center is surrounded by a matrix space which is composed by 5 × 5 cellular to impact the cellular changes significantly. (3) Determining the starting point and the CA number of iterations. The study takes the year 2008 as a starting point. The number of CA iterations is selected at 15 in order to simulate the landscape spatial pattern for the study area in 2015. Using CA–Markov modules in IDRISI, the land use change simulation technical route is shown in Fig. 2. 4. Results and analysis 4.1. Reclassification of land use types Based on the land use maps in 2001 and 2008, the types of land use were reclassified with reference to remote sensing image characteristics. The land use classification adopted in the model consists of seven land use classes. Seven of these, namely: cultivated field, garden, woodland, grassland, construction land, waters and other non-utilization land are included. The result of the new reclassification is consistent with national land use classification criteria. The major tourist attractions and grassland will be added into other non-utilization land in the new classification. The quantity and spatial pattern of other non-utilization land basically remained stable. After reclassification, special lands were classified into other land use types according to basic characteristics of land use/land cover, which includes urban, woodland, water or independent industrial and mining. The result of the reclassifications and their distribution is as shown in Fig. 3. 4.2. Land use structure changes According to the spatial overlay analysis in GIS, from 2001 to 2008, there are changes in the land use types over these 7 years. The changes in area are as shown in Table 1.
L. Sang et al. / Mathematical and Computer Modelling 54 (2011) 938–943
941
Fig. 1. Study area and location.
Fig. 2. Flowchart of the technical route of land use simulation.
a
b
Fig. 3. The re-classified grid land use maps (a) 2001 land use map (b) 2008 land use map. Note: AL—arable land; GAL—garden; WL—wood land; GL— grassland; WT—water (lakes, rivers and artificial ponds); CL—construction land; UL—unused land.
The main types of land use in Fangshan are arable land and construction land, which accounted for 58% of the total. Construction land (including urban construction, rural settlements and other construction land) in Fangshan accounted for 28.9% of the total area from 2001 to 2008; arable land, grassland and unused land have a corresponding reduction, which ranges up to a 29.1% reduction in arable land, reflecting the quick decline in arable land resources; Unused land decreased by 18.2%, indicating reserve land resources are inadequate in the study area; gardens and water have smaller changes in the total area. Land use changes are significant in Fangshan District from 2001 to 2008. Fig. 1 and Table 1 reflect the land use of space and volume changes from 2001 to 2008. Comparing the land use changes, the most obvious trend is the expansion of urban and industrial land. However, the expansion of urban and industrial land in space is very different. Land use dynamic changes can quantitatively describe the speed of land use changes, which plays a unique role in comparing the regional differences of land use changes and in predicting the trend of land use changes [12]. Land use dynamic changes also indicate the level of human disturbance to land use types. The higher the disturbance level, the more intensively land use changes. In this paper, the area changes of land use types in Fangshan during the two periods were further
942
L. Sang et al. / Mathematical and Computer Modelling 54 (2011) 938–943
Table 1 Area of land use changes in Fangshan during 2001–2008.
2001 (hm2 ) 2008 (hm2 ) Changed area (hm2 ) Change ratio (%)
AL
GAL
WL
GL
CL
WT
UL
32 079.82 22 751.87 −9 327.95 −29.1
9110.03 9575.54 465.51 5.1
72 661.85 81 758.28 9 096.43 12.5
26 046.03 20 013.50 −6 032.53 −23.2
33 187.16 42 787.70 9 600.54 28.9
6757.57 6432.92 −324.65 −4.8
19 111.94 15 634.59 −3 477.35 −18.2
Table 2 Transition probabilities matrix of land use types in Fangshan during 2001–2008 (units /%). Land use types
AL
GAL
WL
GL
CL
WT
UL
AL GAL WL GL CL WT UL
87.21 1.72 0.44 4.58 0.44 0.15 2.75
4.07 90.43 0.51 1.24 0.86 0.07 0.05
0.28 1.61 97.53 0.45 0.62 0.18 4.83
0 0.41 0 89.82 0 3.20 0
7.96 5.26 1.29 3.43 98.01 0.15 16.83
0.08 0.28 0.02 0.21 0.07 96.25 0
0.40 0.29 0.21 0.27 0 0 75.54
calculated. The land use changes at these two stages showed marked fluctuations. Arable land, woodland and unused land experienced a reversed trend of changes during the two periods. 4.3. Model application The following procedures were performed using algorithms available in IDRISI Kilimanjaro and Image Processing software in order to implement the CA–Markov model in the study area: (1) computation of land use transition potential maps based on procedure, (2) computation of transition probabilities using Markov chain analysis, and (3) spatial allocation of simulation land use probabilities base on a CA spatial filter. Land use pattern changes in 2015 were predicted using a CA–Markov model based on Fangshan District of Beijing in 2008 as an initial state, for which the time interval was 7 years. The Markov transition matrix can not only quantitatively explain the conversion between the types of land status, but also reveal the transfer rate between the different types. The conversion matrix of land use types during 2001–2008 in Table 2 can be summarized as follows: (1) Arable land had a massive loss as a result of conversion to other land types, in the form of conversion from arable land to garden and construction land, of which 7.96% was transformed to construction land and 4.07% into corner. The main reason for the reduction of arable land is urbanization and agricultural structure adjustment, while the supplement of arable land comes mainly from grassland and unused land. (2) The increase in woodland mainly came from garden and unused land, which accounted for 80.8% of the total conversion gain, 1.29% of which were transformed into construction land, 0.51% into corner, 0.44% transformed into cultivated land, which increased some of the major sources of arable land and grassland. (3) 5.26% of garden was transformed into construction land, 1.72% was transformed into arable land, and 1.61% into woodland. There was little change in the total area of Garden, but there were increases and decreases, and there were significant changes in the spatial distribution. (4) There was a grassland loss of 3.62%, which was due mostly to conversion to arable land, garden and construction land. At the secondary level of classification, conversion to woodland is the main reason for the shrinkage of dense cover grassland. (5) The conversion gain of waters was contributed mainly by grassland. The changes were small. (6) The area converted to construction land mainly encompassed rural settlements and urban construction, which took up lots of arable land, garden and woodland. And changes to other land use types are less. (7) The main conversion was other unused land into construction land, woodland and cultivated land. Studies have shown that unused land and forest land were the dominant land use types in Fangshan. Since 2001, farm land had been shrinking while forest land and built-up areas had increased quickly. Among the various transformation types, the changes from farm land to built-up land spread most significantly. Regions along railways and main roads as well as rivers with intense human activity had a high local variability of land use distribution. Therefore they became more varied and the spatial pattern of land use changes became more complex. The simulation result by the CA–Markov method showed that its original rate of changes in trends and changes will keep constant from 2008 to 2015. Therefore, it is urgent to strengthen the protection of farm land and water-bodies, to prevent acts of indiscriminate use of farmland in order to promote the protection of farmland and the rational use of land. 5. Conclusion Using land use maps (2001, 2006 and 2008), the CA–Markov model that combines the Markov chain analysis and CA models successfully simulated land use changes in Fangshan. The number and spatial distribution of the area is analyzed,
L. Sang et al. / Mathematical and Computer Modelling 54 (2011) 938–943
943
and the overall results are satisfactory: the seven land use types will maintain their changes in the direction and rates of change. Therefore, it is urgent to strengthen the protection of arable land and water and prevent acts of indiscriminate use of cultivated land in order to promote land protection and rational use of land. This study represents an important contribution to land use modeling as shown by the CA–Markov land use simulation model. In addition, the model considers only the surrounding natural environment, and the cellular is not moving. No consideration is given to playing a decisive role in land use dynamics changes of the social environment and their interactions. As a result, the human decision-making model for the simulation is still a weakness which needs further study. Acknowledgements This research was supported by National Science and Technology Ministry (No. 2006BAJ14B01-02) and Graduate Research and Innovation Subject in China Agricultural University (No. kycx09125). The authors also thank Miss Gao for comments on earlier versions of this paper. References [1] Verda Kocabas, Suzana Dragicevic, Assessing cellular automata model behaviour using a sensitivity analysis approach, Computers, Environment and Urban Systems 30 (2006) 921–953. [2] B.C. Pijanowski, D.G. Brown, G. Manik, Using neural nets and GIS to forecast land use changes: a land transformation model, Computers, Environment and Urban Systems 26 (6) (2002) 553–575. [3] X.Y. Hou, B. Chang, X.F. Yu, Land use change in Hexi corridor based on CA–Markov methods, Transactions of the CSAE 20 (5) (2004) 286–291. [4] Courage Kamusoko, Masamu Aniya, Bongo Adi, et al., Rural sustainability under threat in Zimbabwe-simulation of future land use/cover changes in the Bindura district based on the Markov-cellular automata model, Applied Geography 29 (2009) 435–447. [5] G.H. Jiang, F.R. Zhang, X.B. Kong, Determining conversion direction of the rural residential land consolidation in Beijing mountainous areas, Transactions of the CSAE 25 (2) (2009) 214–221. [6] G.Q. Yang, Y.L. Liu, Z.F. Wu, Analysis and simulation of land use temporal and spatial pattern based on CA–Markov model, Geomatics and Information Science of Wuhan University 32 (5) (2007) 414–418. [7] M. Batty, Yichun Xie, Zhanli Sun, Modeling urban dynamics through GIS-based cellular automata, Computers, Environment and Urban Systems 23 (1999) 205–233. [8] L.K. Peterson, K.M. Bergen, D.G. Browna, et al., Forested land-cover patterns and trends over changing forest management eras in the Siberian Baikal region, Forest Ecology and Management 257 (2009) 911–922. [9] Rohan Chandralal Wickramasuriya, Arnold K. Bregt, Hedwig van Delden, et al., The dynamics of shifting cultivation captured in an extended constrained cellular automata land use model, Ecological Modelling 220 (2009) 2302–2309. [10] X.L. Wang, Y.H. Bao, Study on the methods of land use dynamic change research, Progress in Geography 18 (1) (1999) 81–87. [11] H. Ji, Yoshitsugu Hayashi, C. Xin, et al., Application of an integrated system dynamics and cellular automata model for urban growth assessment: a case study of Shanghai, China, Landscape and Urban Planning 91 (2009) 133–141. [12] L.Y. Guo, D.L. Wang, J.J. Qiu, et al., Spatio-temporal patterns of land use change along the Bohai Rim in China during 1985–2005, Journal of Geographical Sciences 19 (2009) 568–576.