Simulation of urban agglomeration ecosystem spatial distributions under different scenarios: A case study of the Changsha–Zhuzhou–Xiangtan urban agglomeration

Ecological Engineering 88 (2016) 112–121

Contents lists available at ScienceDirect

Ecological Engineering journal homepage: www.elsevier.com/locate/ecoleng

Simulation of urban agglomeration ecosystem spatial distributions under different scenarios: A case study of the Changsha–Zhuzhou–Xiangtan urban agglomeration Weiguo Jiang a,b,∗ , Zheng Chen a,b , Xuan Lei c,∗ , Bin He d , Kai Jia a,b , Yunfei Zhang a,b a

Academy of Disaster Reduction and Emergency Management, Beijing Normal University, Beijing 100875, China Key Laboratory of Environmental Change and Natural Disaster, Beijing Normal University, Beijing 100875, China c Tianjin University Research Institute of Urban Planning, Tianjin 300073, China d College of Global Change and Earth System Science, Beijing Normal University, Beijing 100875, China b

a r t i c l e

i n f o

Article history: Received 11 May 2015 Received in revised form 4 September 2015 Accepted 6 December 2015 Available online 30 December 2015 Keywords: Urban agglomerations Ecological security Scenario analysis CLUE-S

a b s t r a c t Changsha–Zhuzhou–Xiangtan is an urban agglomeration along the middle reaches of the Yangtze River, the development plan of which is listed as one of national strategic development. The eco-environmental quality of the Changsha–Zhuzhou–Xiangtan urban agglomeration concerns the ecological security of China. It is essential to discuss the method of creating urban agglomeration plans to increase ecological security. In this paper, the ecosystem distributions were mapped using the CLUE-S model incorporated with Gray Model (GM) (1, 1) and an auto-logistic regression model under the conditions of a natural increase scenario (NIS), a cultivated protection scenario (CPS) and an ecological protection scenario (EPS). We analyzed the change and conversion characteristics of ecosystems both in the study area and the junction of Changsha, Zhuzhou and Xiangtan. The results showed that the ecosystem change model developed in this paper performed well in mapping future urban ecosystem distributions. The change of the ecosystem spatial distributions showed us that the built-up ecosystem would expand in the future by transforming cultivated and green land ecosystems and that the boundaries between cities would be blurred. In the whole study area, the amount of converted area from cultivated and green land to built-up land was lower in CPS and EPS than in NIS. However, in the key area, the converted result was contrary to the whole study area. Although the CPS and EPS were beneficial to the eco-environmental protection of the whole study region, the sub-regional eco-environment should be given more attention to assure that the ecological security of the whole region remains safe. © 2015 Elsevier B.V. All rights reserved.

1. Introduction Urban agglomerations, which result from economic development and urbanization, are regions composed of a number of cities around regional economic core cities (He et al., 2013). The typical urban agglomerations in China include Beijing–Tianjin–Hebei, Pearl River Delta, Yangtze River Delta, and so on. Urban agglomeration expansions have changed the original ecological landscapes and ecosystem structures and threaten regional ecotopes (Zhou et al., 2015). As a result, regional ecological security is worsening. Therefore, it is essential to take measures to protect the

∗ Corresponding author at: Academy of Disaster Reduction and Emergency Management, Beijing Normal University, Beijing 100875, China. Tel.: +86 10 58809318. E-mail addresses: [email protected] (W. Jiang), lx [email protected] (X. Lei). http://dx.doi.org/10.1016/j.ecoleng.2015.12.014 0925-8574/© 2015 Elsevier B.V. All rights reserved.

environment and promote the harmonious development of human beings and nature. How can cities balance the contradiction between rapid development and ecological environmental protection to build an ecological urban agglomeration? Scenario analysis is an effective method to map the spatial distribution of urban agglomeration. Urban planners and policy makers can select a more reasonable mode of development and optimize the allocation of resources based on maps of urban agglomeration under different scenarios. Thus, scenario analysis is a significant method to maintain urban agglomeration ecosystem health. An ecosystem change model is the key to scenario analysis, such as the Cellular Automata model (CA), system dynamics model (SD) and the Conversion of Land Use and its Effects model (CLUE). Pijanowski et al. (2002) incorporated the BP model into the CA model to study urban expansion and urban structure change. Saysel et al. (2002) built an SD model to simulate the ecosystems spatial change and analyzed the variation trend. Veldkamp and Fresco,

W. Jiang et al. / Ecological Engineering 88 (2016) 112–121

1996 proposed the CLUE model. Verburg et al. (2002) developed the CLUE model and proposed the Conversion of Land Use and its Effects at Small regional extent model (CLUE-S) to simulate ecological system changes at a small regional extent. The CLUE-S model has been developed and applied to many fields (Verburg et al., 2008; Overmars et al., 2007; Trisurat et al., 2010; Castella et al., 2007). For example, Fox et al. (2012) used the CLUE-S model to simulate the land-cover change in the montane mainland southeast of Asia. However, there are some limitations of these models. The CA model cannot simulate the competition between ecosystems and has difficulty incorporating expert knowledge (Mas et al., 2014). The ordinary regression models map the future ecosystem distribution according to past distributions, without considering the uncertainty. The CLUE-S model takes social–economic and natural factors into account and performs better than the other ecosystem change models (Jiang et al., 2015). A spatial correlation effect commonly exists in geographic space. The selected driving factors are likely to be incorrect when the spatial correlation is ignored (Overmars et al., 2003). Auto-logistic regression models (AL), which incorporate auto-covariates based on logistic regression models, are available to eliminate the spatial correlation effect. AL models have been commonly used in modeling ecological diversity (Augustin et al., 1996; Syartinilia and Tsuyuki, 2008). Studies show that incorporating an auto-logistic regression model is an effective method to improve the performance of the CLUE-S model (Lin et al., 2011; Wu et al., 2010). The CLUE-S model is widely used in landscape change studies of China’s urban agglomerations. Dai and Zhang (2013) simulated landscapes under five different scenarios in Zhangye city, China, using the CLUE-S model. Hu et al. (2013) incorporated a Markov model into the CLUE-S model and simulated the landscape of Beijing for 2015 under two scenarios. Zheng et al. (2015) used the CLUE-S model incorporated with a Markov model to simulate land use changes of urban renewal districts under four scenarios for 2018. The Changsha–Zhuzhou–Xiangtan region is one of the most important urban agglomerations. The ecological isolation zone between Changsha, Zhuzhou and Xiangtan has been destroyed and the cultivated areas and woodlands and wetlands are shrinking due to the expansion of the built-up ecosystem. The regional ecological security problem has become more prominent. Currently, construction of eco-cities is advocated in China. How to design the landscape to improve the ecological security of ecosystems in urban agglomeration is an important question to explore. This paper takes the Changsha–Zhuzhou–Xiangtan urban agglomeration, one of the most important agglomerations of the Yangtze River, as the study area. An auto-logistic regression model is incorporated into the CLUE-S model to map the ecosystem distribution of Changsha–Zhuzhou–Xiangtan urban agglomeration for 2014, 2019 and 2024 based on observed maps from 1995, 2000, 2005 and 2009 under the conditions of a natural increase scenario (NIS), a cultivated protection scenario (CPS) and an ecological protection scenario (EPS). The characteristics of the ecosystem patterns are analyzed to provide suggestions for urban planning and eco-city construction.

2. Materials and methods

113

Table 1 The definition of BE, GE, CE and WE (Ouyang et al., 2015). Ecosystem

Definition

Examples

BE

Human settlement, mainly of artificial surface Land surface is mainly of tree, bush and grass Land surface is mainly of crops Land surface with water

Residential area, urban green land Woodland, grass land

GE CE WE

Cultivated land, garden plot Rivers, lakes, marsh

of Changsha, Zhuzhou and Xiangtan (Yang et al., 2012), between 112◦ 36 –113◦ 17 E and 27◦ 37 –28◦ 33 N, with an area of 4588 km2 . The Xiangjiang River is the biggest river of the CZT, crossing the study area from south to north. The main urban area of Changsha and Zhuzhou is located along the right bank of the Xiangjiang River, and Xiangtan is located along the left bank. The distances from Changsha to Xiangtan and from Zhuzhou to Xiangtan are 40 km and 20 km, respectively. With the urban development, the distances between the cities are decreasing and the centrality of the cities is gradually expanding. The CZT is the regional economic core of Hunan province. The built-up area has expanded since the 1990s. From 2000 to 2008, the built-up area increased by 171 km2 at the cost of the cultivated land and woodlands. According to the “Changsha–Zhuzhou–Xiangtan urban agglomeration regional development plan” proposed in 2008, the ecotype of urban agglomeration is going to be constructed to improve regional ecological security. 2.2. Materials Two types of data were used in this paper, including ecosystem distribution maps and the collection of explanatory data. A series of Landsat TM/ETM data for CZT accepted in 1995, 2000, 2005 and 2009 were explored in this study at a resolution of 30 m. The CZT ecosystem has been classified into five classes (defined as Table 1), including the built-up ecosystem (BE), green land ecosystem (GE), cultivated ecosystem (CE), wetland ecosystem (WE) and others by the Support Vector Machine (SVM) method. The GE was composed of woodland ecosystems and urban green land ecosystems. The observed maps (Fig. 2) of the CZT ecosystem were used to simulate and validate ecosystem demand areas, simulate ecosystem distributions and analyze the characteristics of change under different scenarios for the years of 2014, 2019 and 2024. The natural resource data were used to select natural driving factors, including slope, aspect (−7∼639 m), and soil type (e.g., primary soil, semi-hydromorphic soil, anthropic soil, ferrasol, impervious surface, lake reservoir, rivers and island). Social–economic data (e.g., settlement, town center, river net and transportation data) were used to select social–economic driving factors based on distances from stations, city centers, town centers, county centers, village centers, express ways, railways and the Xiangjiang River and its branches. The limited developing regions and ecosystem areas were set in the CZT developing plan, and the cultivated protection scenario and ecological protection scenario were set according to the plan.

2.1. Study area 2.3. Methods The Changsha–Zhuzhou–Xiangtan region is an urban agglomeration located along the middle reaches of the Yangtze River. The core region of Changsha–Zhuzhou–Xiangtan is located in the middle-east of Hunan province, which is located in central China. The Changsha–Zhuzhou–Xiangtan urban agglomeration (CZT) (Fig. 1) is composed of the main part and the perimeter zone

This paper simulated ecosystem demand areas for 2000, 2005 and 2009 using the GM (1, 1) model based on observed maps for 1995, and the simulated demand areas were validated by observed maps for 2000, 2005 and 2009. If the simulated demand areas met the required accuracy, the ecosystem demand for 2014,

114

W. Jiang et al. / Ecological Engineering 88 (2016) 112–121

Fig. 1. Location of the study area.

Fig. 2. The ecosystem distributions data of 1995, 2000, 2005 and 2009.

2019 and 2024 was simulated according to the natural increase scenario, cultivated protection scenario and ecological protection scenario. An auto-logistic regression model was employed to select driving factors considering the spatial autocorrelation effect. The driving factors and ecosystem demands were used as spatial and non-spatial module input parameters to simulate ecosystem distribution maps in NIS, CPS and EPS. Finally, based on the simulated maps, the ecosystem conversion characteristics, both of the whole study area and the key region, were analyzed to provide urban agglomeration planning suggestions. A method flowchart is shown in Fig. 3.

2.3.1. Scenarios setting Setting the scenario conditions is the key to scenario analysis. In this work, scenarios were set and the demand areas were limited according to each ecosystem’s spatial characteristics. By considering the spatial distribution characteristics of the regional economic core cities and different urban development orientations, we simulated the spatial distributions of the CZT ecosystem for 2014, 2019 and 2024 under the NIS, CPS and EPS, respectively. Under the NIS, the ecosystem spatial distribution and demand was unaffected by natural conditions and state policy. Each ecosystem changed following the trend from 1995 to 2009, and the

W. Jiang et al. / Ecological Engineering 88 (2016) 112–121

115

Fig. 3. Flowchart of CLUE-S model based on auto-logistic regression model, the conditions are according to scenarios setting methods.

built-up ecosystem experienced rapid expansion. Under this scenario, ecosystem demand was predicted according to the historical tendency and the rivers were the only regions not allowed to develop. The CPS was set to assure food supply security on the condition that the CZT development is unaffected. A certain region of town was used as the future built-up ecosystem, and the cultivated land, located far from cities and main roads, was set as the limited development region to strictly control the conversion of cultivated land. According to the research of Liu et al. (2012), the limited region under this scenario includes a 1000 m buffer from the key planning regions, expressways and railways and a 500 m buffer from cities and main roads. According to the CZT developing plan, the built-up ecosystem area accounts for 26.3% of the whole region. The EPS ecosystems, which are important to regional ecological security, such as woodlands, grasslands, wetlands, ecological preservation areas and wetland reserves, were given strengthened protection based on the CPS. The EPS was significant to maintain and improve the regional eco-environment and stabilization of the urban agglomeration ecosystem. The limited regions include a 300 m buffer from the Xiangjiang River and its branches, a 500 m buffer from lakes and 100 m buffer from streams based on the limited regions of the CPS. The builtup ecosystem demand was set to be the lowest to reduce the conversion rate from green land ecosystems to built-up ecosystems. 2.3.2. The ecosystem change model The CLUE-S model was used to simulate future ecosystem spatial distributions under different scenarios. It assumed that the

ecosystem demand was driving the evolution of the ecosystem distributions and the relationship between the CZT spatial distribution and demand area and that the environment and social–economic conditions maintain a dynamic balance. The CLUE-S model consists of a spatial and non-spatial module (Verburg et al., 2002). The ecosystem demands were simulated in a non-spatial module based on ecosystem data, social–economic data and the historical area of each ecosystem. The demands are translated into each ecosystem to map ecosystem distributions based on the spatial characteristics of each ecosystem. The use of ecosystem demand calculation models is an effective way to improve the performance of the CLUE-S model (Liang et al., 2011; Aspinall, 2004). The GM (1, 1) model is a time series prediction method based on mathematical statistics theory. In this paper, the GM (1, 1) model was used to calculate the demand based on the following considerations: (1) if the state policy does not include a major change, the urbanization is a stabilized system developing process and the CZT ecosystem will change gradually; (2) the ecosystem area can be accumulated and is mutually independent; (3) only four periods of observed maps were used in this paper; and (4) the time span is only 10 years from 2014 to 2024. The GM (1, 1) model has the advantage of predicting the demand based on limited samples and is applicable to calculate ecosystem demand over short periods. The spatial autocorrelation effect is common in geographical research. The interaction in the ecosystem conversion process is significant. The auto-logistic regression model, which is incorporated with auto-covariates, is effective at improving the CLUE-S model by eliminating the bias caused by spatial autocorrelation (Jiang et al., 2015). In this paper, the auto-logistic regression model was employed to select the driving factors of the CZT ecosystem

116

W. Jiang et al. / Ecological Engineering 88 (2016) 112–121

Table 2 The validation results. c is the variance ratio, and p is error probability, and bias is the error between simulated demands and calculated demands, and a is the overall accuracy. c is the greater and p is the smaller, the accuracy is the better. When p > 0.95 and c < 0.35, the result is the best; when p > 0.80 and c < 0.50, the result is qualified. Year

BE

1995 2000 2005 2009

GE

CE

Bias%

a

Bias%

a

Bias%

a

Bias%

a

18.53 −11.15 −3.25 3.61

c = 0.26

−1.55 1.19 0.58 0.32

c = 0.33

−3.94 2.33 1.61 −5.65

c = 0.34

9.29 −5.05 −5.03 5.81

c = 0.57

−0.62 1.04 0.21 5.12

c = 0.37

p = 1.0

p = 1.0

Scenes

Year

BE

GE

CE

WE

Other

NIS

2014 2019 2024 2014 2019 2024 2014 2019 2024

987.82 1265.53 1514.60 982.82 1234.53 1474.60 952.82 1194.53 1434.60

1843.09 1720.62 1606.91 1843.09 1720.62 1606.91 1873.09 1760.62 1646.91

1434.08 1273.29 1130.68 1439.08 1304.29 1170.68 1439.08 1304.29 1170.68

226.57 240.07 252.86 226.57 240.07 252.86 226.57 240.07 252.86

96.23 88.46 82.69 96.23 88.46 82.69 96.23 88.46 82.69

EPS

spatial change. The most common auto-logistic model is defined as (Wu et al., 2009) log

 p  i 1 − pi

= ˇ0 + ˇ1 X1,i + ˇ2 X2,i +, ..., +ˇn Xn,i + ˇn+1 autocovi (1)

where, pi is the probability that a unit belongs to ecosystem I; and ˇ0 , ˇ,  are the variable coefficients calculated by the maximum likelihood method; and autocovi is the auto-covariate defined as (Jiang et al., 2015):

 autocovi =

Other

a

Table 3 The ecosystem demands (km2 ). NIS is short for natural increase scenario, and CPS is short for cultivated protection scenario and EPS is short for ecological protection scenario.

CPS

WE

Bias%

p = 0.96

p = 0.80

p = 0.95

3.2. Ecosystems spatial distribution maps The CZT ecosystem distribution in 2014, 2019 and 2024 were simulated by the CLUE-S model based on auto-logistic regression results and the observed map of 2009 for the NIS, CPS and EPS. The simulated maps are shown as Fig. 4. The simulated maps show that the built-up ecosystem tends to expand and that the area of green and cultivated land decreases. The centrality and clustering of Changsha, Zhuzhou and Xiangtan is significant in the developing process. The built-up ecosystem of Changsha is gradually expanding to the south and the built-up ecosystems of Zhuzhou and Xiangtan are expanding to the north. As a result, the boundaries of Changsha, Zhuzhou and Xiangtan are gradually diminished. However, because of the different limited conditions, the ecosystem distribution characteristics of the NIS, CPS and EPS are different. In the CPS and EPS, the cultivated, green land and wetland ecosystems are protected and the area converted to built-up ecosystem is reduced. The area of built-up ecosystem under the CPS and EPS is less than that under the NIS, and the area converted from cultivated ecosystem to built-up ecosystem is lowest under the EPS. The spatial distribution maps suggest that the expansion of the built-up ecosystem is more conspicuous under the NIS and that other ecosystems change so rapidly in a short time that the ecosystem stability decreases and are therefore vulnerable. In the NIS, the regional ecological security is more likely to be destroyed. As a consequence, producing an ecological urban plan is a significant way to reduce the regional ecological security risk.

wij yj

j= / i



(2) wij

j= / i

where, yj is the probability that pixel j belongs to a certain ecosystem. If j belongs to a certain ecosystem, then yi = 1; otherwise yi = 0, and wij is the weight coefficient according to the distance (d) between pixel i and pixel j. When the distance between pixel i and pixel j less than d, then wij = 1/d, otherwise wij = 0. 3. Results 3.1. Ecosystems demands prediction results The posterior variance test was used to validate the simulated demands for built-up ecosystems, green land ecosystems, cultivated ecosystems, wetland ecosystems and other ecosystems in the CZT from 1995 to 2009. The validation results of the simulated demands with the demands calculated from observed maps are shown in Table 2. A series of observed maps from remote sensing data (1995, 2000, 2005 and 2009) and land use data (2002, 2003, 2006 and 2007) were used as the basic data to simulate ecosystem demands for 2014, 2019 and 2024 according to the conditions of NIS, CPS and EPS. The ecosystem demands for 2014, 2019 and 2024 under the NIS, CPS and EPS are shown in Table 3.

3.3. Analysis of ecosystem changes under different scenarios A transfer matrix (Table 4) was calculated by overlaying the ecosystem distribution maps of 2024 and 2009 to analyze the ecosystem spatial pattern characteristics after 14 years of development. Based on Table 3, we calculated the ratio of each ecosystem increased area and the total area under each scenario. The results are shown as Fig. 5. Under the NIS, 50.4% of the built-up ecosystem is composed of the increased area from 2009 to 2024. The increased area was converted from cultivated ecosystems (58.76%) and green land ecosystems (38.4%). There is no increased area of the green land ecosystem, although it accounts for the largest percentage of the whole region. However, the leading causes of green land ecosystem conversion were weakened, which is beneficial for the regional ecological service value. The percentage of the increased area of the cultivated ecosystem is 5.5%, converted from a small reservoirpond and urban green land. Although there is some increased area to make up for the decrease of cultivated land, the decreasing trend is still the most significant. From 2009 to 2024, the cultivated ecosystem decreased by 27.9%. The wetland ecosystem increased by 23.2%, and 77.9% of the increased area was converted from green land ecosystems. Due to the conversion from wetland to cultivated ecosystems, the total area of wetland ecosystems is unchanged. Other ecosystems decreased by 17% by converting to the built-up ecosystem. The natural increase scenario is set according to the past CZT development tendency, which lacks a reasonable ecological

W. Jiang et al. / Ecological Engineering 88 (2016) 112–121

117

Fig. 4. Ecosystem spatial distributions under different scenarios. (a), (b) and (c) are ecosystem distributions of NIS in 2014, 2019 and 2024, respectively. (d), (e) and (f) are ecosystem distributions of CPS in 2014, 2019 and 2024, respectively. (g), (h) and (i) are ecosystem distributions of EPS in 2014, 2019 and 2024, respectively.

protection mechanism. As a consequence, the region developed at the cost of green land ecosystems, destroying the eco-environment. This suggests that future rapid urbanization and the regional development pattern in the CZT is not suitable to the goal set by the CZT development plan. Under the CPS, 48.8% of the built-up ecosystem is the increased area from 2009 to 2024. The area converted to cultivated ecosystems and green land ecosystems decreased. The smallest amount of area was converted from cultivated ecosystems because of the protections of the cultivated ecosystem; thus, the cultivated ecosystem

increased by 3.8%. Because of the decreased converted area, the total ecosystem area in 2024 increased by 40 km2 and the percent of unchanged cultivated ecosystem increased from 68.1% under the NIS to 71.4%. This suggests that the cultivated protection scenario is effective at protecting cultivated land. A portion (16.5%) of the wetland ecosystem was from the increased area. However, the total wetland ecosystem area under the CPS was equal to that under the NIS, both of which showed a decrease. Table 4 and Fig. 5 show that under the CPS, both of the increased and decreased wetland areas are reducing and the percent of unchanged area increased from

118

W. Jiang et al. / Ecological Engineering 88 (2016) 112–121

Table 4 The transfer matrix from 2009 to 2024 and the unit is km2 . 2009

NIS

BE GE CE WE Other

CPS

BE GE CE WE Other

EPS

BE GE CE WE Other

2024 BE

GE

CE

WE

Other

755.91 295.29 451.35 0.99 20.43 BE 755.91 289.26 408.87 0.45 20.61 BE 755.91 258.12 399.42 1.08 21.33

0 1598.85 0.27 0.18 0 GE 0 1607.13 0 0.09 0 GE 0 1645.92 0.81 0.09 0

0 24.12 1069.11 28.35 9.9 CE 0 24.66 1126.98 11.52 7.74 CE 0 18.18 1135.35 9.81 7.56

0 12.51 44.82 190.8 0.18 WE 0 9.9 30.78 208.71 0.54 WE 0 8.73 31.5 209.43 0.45

0 0.36 2.97 1.17 79.83 Other 0 0.18 1.89 0.72 81.45 Other 0 0.18 2.07 1.08 81

Fig. 5. The ratios of each ecosystem increased area and the total area under NIS, CPS and EPS in 2024.

86.1% to 94.2% compared with the NIS. This suggests that CPS is an effective way to protect wetland ecosystems. Although cultivated and wetland ecosystems are protected under the CPS, the green land ecosystem changes by the same amount as under the NIS. Compared with the CPS, the built-up ecosystem area is limited under the EPS and the area converted from green land ecosystems to built-up ecosystems is reduced. The increased built-up area was reduced by 41 km2 , and the other conversions were the same as under the CPS. Under the EPS, the green land ecosystem was still converted to other ecosystems, but the proportion converted to the built-up ecosystem decreased. The unchanged green land area increased to 1646 km2 , and the decline of woodland and grassland slowed down. By 2024, the wetland area increased because the wetland ecosystem was protected. Under the EPS, the region in the buffer from the wetland boundary was not allowed to develop. Although there were no other limitations compared to the CPS, the percentage of unchanged cultivated ecosystem area still increased and the percentage of converted land from cultivated ecosystems was reduced. This suggests that under the EPS, the stability of green land and wetland ecosystems was improved compared with the NIS and CPS, and because of the interaction between ecosystems, the stability of the cultivated ecosystem is also improved.

Consequently, under the EPS, the CZT ecological service value and eco-environmental quality was improved. 3.4. Ecosystem distributions change in the key region There is significant synergy in urban agglomeration ecosystem evolution. Ecosystems in the junction region of Changsha, Zhuzhou and Xiangtan changed rapidly and obviously. The junction of Changsha, Zhuzhou and Xiangtan was selected as the key region to analyze regional ecosystem conversion characteristics quantitatively and qualitatively. The spatial change process of ecosystems in the junction area (Fig. 6) shows that the built-up ecosystem increased from 2000 to 2024 in all three scenarios. The ecosystems distributed on the right bank of Xiangjiang River changed more obviously than that on the left bank. Changsha and Zhuzhou were centered near Xiangtan, and the ecosystems on the boundaries of Changsha–Xiangtan and Zhuzhou–Xiangtan changed rapidly. This suggests an obvious centrality and conglomeration. A megalopolis was formed due to the agglomerative development of Changsha, Zhuzhou and Xiangtan. Compared with the NIS, in the CPS and EPS, the contagion index of the built-up ecosystem distributed on the right bank of Xiangjiang

W. Jiang et al. / Ecological Engineering 88 (2016) 112–121

119

Fig. 6. The change of ecosystem distributions from 2000 to 2024 under different scenarios at the junction of Changsha, Zhuzhou and Xiangtan.

River was less than that on the left bank. Under the CPS, the built-up ecosystem distributed on the right bank of Xiangjiang River showed the lowest contagion index. The ecosystem areas in the junction in 2014, 2019 and 2024 were calculated (Fig. 7). The annual average increased area of the built-up ecosystem is above 100 km2 , and the built-up ecosystem, green land ecosystem and cultivated ecosystem changed rapidly since 2009. A portion (15%) of the increased areas was composed of the built-up ecosystem converted from green land and cultivated ecosystems. Different from the change characteristics in the whole study area, both of the increased built-up ecosystem and decreased green land ecosystem in the junction area under the NIS were less than that under the other two scenarios, and the cultivated ecosystem changed similarly. Under the CPS, the cultivated ecosystem in the key region that reduced the most, and under the EPS, the green land ecosystem lost the most area. It seems unreasonable to set the CPS and EPS conditions, but this is not the case. There are two reasons for this. First, most of the cultivated ecosystem is distributed northwest of Changsha and southwest of Xiangtan, and the cultivated ecosystem in the key region is limited. Thus, under the CPS, the protection of the cultivated ecosystem in the key region is not effective. Second, the EPS was set according to

the CZT development plan, and the woodland and grassland in the southwestern mountainous area are protected. However, a considerable amount green land ecosystems have been converted to built-up ecosystems to meet the urban development requirements. The ecosystem change characteristics of the key region imply that both the whole region and the key development region should be taken into account separately in the regional plan to ensure regional ecological security. 4. Discussion The CZT ecosystem spatial distributions in 2014, 2019 and 2024 were simulated using the CLUE-S model under the NIS, CPS and EPS, and the ecosystems change characteristics were analyzed. The simulated results in this paper are more reliable because an autologistic regression model was employed to select the driving factors considering the spatial autocorrelation effect, leading to selected driving factors that have a higher explanatory ability (Jiang et al., 2015). However, there are some disadvantages when using a linear model to calculate ecosystem demands (He et al., 2013). In this paper, we calculated ecosystem demands using the GM (1, 1), and the calculated results show that the GM (1, 1) is suitable to

120

W. Jiang et al. / Ecological Engineering 88 (2016) 112–121

Fig. 7. The changed area of ecosystem at the junction of Changsha, Zhuzhou and Xiangtan. NIS is short for natural increase scenario; CPS is short for cultivated protect scenario; EPS is short for ecosystem protect scenario. BE is short for built-up ecosystem; GE is short for green land ecosystem; CE is short for cultivated ecosystem. NIS BE stands for the curve of the change of built-up ecosystem area under the natural increase scenario, and by this analogy.

calculate ecosystem demands in the CZT. We should note that the state policy and regional development plan have been considered in this paper, and the simulated maps are likely to be similar to the real distributions. The CZT development plan emphasizes that ecological environments should be protected, land should not be overused, and the connectivity between cities should be controlled in a range to ensure that the regional ecological environment is safe. Under the NIS, the CZT was developed following the past developing tendency by converting green land and cultivated ecosystem to built-up ecosystems, and the ecological security level is falling. The simulated maps indicated that the natural increase from an extensive development pattern is not suitable to CZT development. The protection of the cultivated ecosystem is considered under the CPS, and the converted area from cultivated ecosystems to built-up ecosystems is reduced. Under the EPS, parts of the cultivated, green land and wetland ecosystems are not allowed to be developed, therefore the development of the build-up is limited. The relationship between regional development and ecological security is improved, and the results show that the EPS meets the requirements of eco-environmental protection. The simulated maps provide a future outlook of urban agglomeration future for decision makers. Under each scenario, the built-up ecosystem underwent expansion into the future. Boundaries between cities were gradually obscured and green land and cultivated ecosystems were diminished. As a result, the simulations showed decreases in regional ecosystem stability and security. We noticed that there are some differences between ecosystem changes in the whole study area versus the key region. In the study area, cultivated and green land ecosystems decreased under the NIS more than under the CPS and EPS, but in the key region, the results show the opposite. The main reason is the difference in ecosystem distribution. Consequently, regional characteristics should be taken into account, an ecological protection red line should be set, and an ecosystem buffer should be assigned to improve regional ecological security.

In the past 30 years, China has experienced rapid development. In the future, urbanization and the number of urban agglomerations will increase. The relationship between human activities and the eco-environment is complex. They are not only mutually promoting but also mutually confining (Zhou et al., 2015). Different urban agglomerations have different circumstances; for example, the location and traffic conditions. Thus, it is necessary to select a suitable development plan. How should we design urban agglomerations to protect ecological security and eco-environment quality when sustainable development and eco-cities are required? This topic is worth discussing. 5. Conclusions In this paper, the future (2014, 2019 and 2024) ecosystem spatial distributions of CZT were simulated using the CLUE-S model under the NIS, CPS and EPS based on the observed map of 2009 and incorporating GM (1, 1) and an auto-logistic regression model and by accounting for the regional ecosystem distribution characteristics and regional policies. The ecosystem changes of the CZT and the key region were analyzed. We arrived at the following conclusions: (1) Incorporating the GM (1, 1) is an effective way to improve the performance of the CLUE-S model. According to the CZT development plan, it is reasonable to simulate future urban agglomeration spatial distributions by the CLUE-S model. The simulated maps conform to the real distribution. (2) In the future, built-up ecosystems will continue to increase, as green land and cultivated ecosystems are converted. Under the CPS and EPS, the areas converted from green land and cultivated ecosystems to built-up ecosystems within the study area are less than those under the NIS. However, in the key region, the reduced areas of the green land and cultivated ecosystems are larger than those under the NIS. Thus, not only should the green land and cultivated ecosystems in the whole study area be protected, but also the ecosystems in some special regions should be protected as well. (3) A significant centrality and aggregation occurs in CZT development. Boundaries between Changsha, Zhuzhou and Xiangtan were gradually obscured. In the future,

W. Jiang et al. / Ecological Engineering 88 (2016) 112–121

Changsha, Zhuzhou and Xiangtan will develop into a metropolitan city. Ecosystems in the urban agglomeration interact and influence each other. Compared with the CPS and NIS, under the EPS, ecological stability and security is improved and is suitable for CZT development. Acknowledgment This work is supported by funds from the National Natural Science Foundation of China (41171318 and 41571077) and the Fundamental Research Funds for the Central Universities. References Aspinall, R., 2004. Modelling land use change with generalized linear models-a multi-model analysis of change between1860 and 2000 in Gallatin Valley. Montana J. Environ. Manag. 72, 91–103. Augustin, N.H., Mugglestone, M.A., Buckland, S.T., 1996. An autologistic model for the spatial distribution of wildlife. J. Appl. Ecol. 33, 339–347. Castella, J.C., Kam, S.P., Quang, D.D., Verburg, P.H., Hoanh, C.T., 2007. Combining topdown and bottom-up modelling approaches of land use/cover change to support public policies: application to sustainable management of natural resources in northern Vietnam. Land Use Policy 24, 531–545. Dai, S.P., Zhang, B., 2013. Land use change scenarios simulation in the middle reaches of the Heihe River basin based on CLUE-S model—a case of Ganzhou district of Zhangye city. J. Nat. Resour. 28, 336–348 (in Chinese). Fox, J., Vogler, J.B., Sen, O.L., Giambelluca, T.W., Ziegler, A.D., 2012. Simulating land-cover change in montane mainland Southeast Asia. Environ. Manag. 49, 968–979. He, C.Y., Zhao, Y.Y., Tian, J., Shi, P.J., 2013. Modeling the urban landscape dynamics in a megalopolitan cluster area by incorporating a gravitational field model with cellular automata. Landsc. Urban Plan. 113, 78–89. Hu, Y.C., Zheng, Y.M., Zheng, X.Q., 2013. Simulation of land-use scenarios for Beijing using CLUE-S and Markov composite models. Chin. Geogr. Sci. 23, 92–100. Jiang, W.G., Chen, Z., Lei, X., Jia, K., Wu, Y.F., 2015. Simulating urban land use change by incorporating an autologistic regression model into a CLUE-S model. J. Geogr. Sci. 25, 836–850. Liang, Y.J., Xu, Z.M., Zhong, F.L., 2011. Land use scenario analyses by based on system dynamic model and CLUE-S model at regional scale: a case study of Ganzhou district of Zhangye city. Geogr. Res. 30, 564–576 (in Chinese). Lin, Y.P., Chu, H.J., Wu, C.F., Verburg, P.H., 2011. Predictive ability of logistic regression, auto-logistic regression and neural network models in empirical land-use change modeling—a case study. Int. J. Geogr. Inform. Sci. 25, 65–87.

121

Liu, M., Hu, Y.M., Sun, F.Y., Li, C.L., Xu, Y.Y., Zhou, Y., 2012. Application of land use model CLUE-S in the planning of central Liaoning urban agglomerations. Chin. J. Ecol. 31, 413–420 (in Chinese). Mas, J.F., Kolb, M., Paegelow, M., Camacho Olmedo, M.T., Houet, T., 2014. Inductive pattern-based land use/cover change models: a comparison of four software packages. Environ. Model. Softw. 51, 94–111. Ouyang, Z.Y., Zhang, L., Wu, B.F., Li, X.S., Xu, W.H., Xiao, Y., Zheng, H., 2015. An ecosystem classification system based on remote sensor information in China. Acta Ecol. Sin. 35, 219–226 (in Chinese). Overmars, K.P., de Groot, W.T., Huigen, M.G.A., 2007. Comparing inductive and deductive modelling of land use decisions: principles, a model and an illustration from the Philippines. Hum. Ecol. 35, 439–452. Overmars, K.P., de Koning, G.H.J., Veldkamp, A., 2003. Spatial autocorrelation in multi-scale land use models. Ecol. Model. 164, 257–270. Pijanowski, B.C., Brown, D.G., Shellito, B.A., Manik, G.A., 2002. Using neural networks and GIS to forecast land use changes: a Land Transformation Model. Comput., Environ. Urban Systs. 26, 553–575. Saysel, S.K., Barlas, Y., Yenigun, O., 2002. Long term sustainability in an agricultural development project: a system dynamics approach. J. Environ. Manag. 64, 247–260. Syartinilia, Tsuyuki, S., 2008. GIS-based modeling of Javan Hawk-Eagle distribution using logistic and autologistic regression models. Biol. Conserv. 141, 756–769. Trisurat, Y., Alkemade, K., Verburg, P.H., 2010. Projecting land-use change and its consequences for biodiversity in Northern Thailand. Environ. Manag. 45, 626–639. Veldkamp, A., Fresco, L.O., 1996. CLUE: a conceptual model to study the conversion of land use and its effects. Ecol. Model. 85, 253–270. Verburg, P.H., Eickhout, B., van Meijl, H., 2008. A multi-scale, multi-model approach for analyzing the future dynamics of European land use. Ann. Reg. Sci. 42, 57–77. Verburg, P.H., Soepboer, W., Veldkamp, A., Limpiada, R., Espaldon, V., Mastura, S.S.A., 2002. Modeling the spatial dynamics of regional land use: the CLUE-S model. Environ. Manag. 30, 391–405. Wu, D.Q., Liu, J., Zhang, G.S., Ding, W.J., Wang, W., Wang, R.Q., 2009. Incorporating spatial autocorrelation into cellular automata model: an application to the dynamics of Chinese tamarisk. Ecol. Model. 220, 3490–3498. Wu, G.P., Zeng, Y.N., Xiao, P.F., Feng, X.Z., Hu, X.T., 2010. Using autologistic spatial models to simulate the distribution of land-use patterns in Zhangjiajie Hunan province. J. Geogr. Sci. 20, 310–320. Yang, B., Chen, Z.L., Zhang, C.S., Dong, J.H., Peng, X.C., 2012. Distribution patterns and major sources of dioxins in soils of the Changsha-Zhuzhou-Xiangtan urban agglomeration. China Ecotoxicol. Environ. Saf. 84, 63–69. Zheng, H.W., Shen, G.Q., Wang, H., Hong, J.K., 2015. Simulating land use change in urban renewal areas: a case study in Hong Kong. Habitat Int. 46, 23–34. Zhou, S., Huang, Y.F., Yu, B.F., Wang, G.Q., 2015. Effects of human activities on the eco-environment in the middle Heihe River Basin based on an extended environmental Kuznets curve model. Ecol. Eng. 76, 14–26.