The future urban growth under policies and its ecological effect in the Jing-Jin-Ji area, China.

Since 2016, the Chinese government has invoked some policies to make Jing-Jin-Ji (JJJ) a new urban agglomeration. However, there has been no research to study the effect of these new policies on future urban growth. This study assessed part of these new policies on JJJ urban growth in 2020-2050 using SLEUTH model. Then the ecological effects of the urban growth are evaluated. Results showed the policies had nearly no obvious impact on the whole JJJ urban growth, but affected sub-regional (Beijing, Tianjin and Hebei, respectively) urban growth. Under ecological protection in future, the value of ecological service in JJJ would increase to a maximum of 31.7×108 Yuan/km2 in 2031. The ecological elasticity also increased and the ecological risk was strongly reduced around the present urban area. This ecologically sustainable development is critical to the future urban growth, and should be considered more carefully by urban planners and managers. More policies should be evaluated for JJJ urban growth in future work.

Introduction

By 2050, 68% of the world population will live in urban areas, with 255 million of these new urban residents living in China (https://www.un.org/development/desa/en/news/population/2018-revision-of-world-urbanization-prospects.html). With the development of the world's economy and growth in population, the coverage of urban area is predicted to grow substantially (Miller and Small, 2003). The Government macro-control plays an important role in urban growth. Predicting urbanization under future policies would allow urban planners and managers to better develop urban (Goetz et al., 2011). However, there are still few studies focus on impact of diverse policies on future urban growth. The Jing-Jin-Ji (JJJ) metropolitan region, also known as the Beijing-Tianjin-Hebei region, is the largest urbanized region in Northern China. Since 2016, the Chinese government has implemented several policies to develop the JJJ into a new urban agglomeration, in addition to the Pearl River Delta agglomeration in southeast China and the Yangtze River Delta agglomeration in east China. The new policies include, for example, developing a new city (Xiong_An) to mitigate the growing Beijing population and transportation; building a JJJ regional distribution with the concept “one core, two cities, three axes, four regions, multiple nodes”; developing new railways and motorways. However, there has been no research to assess the impact of the latest policies on JJJ urban growth.

In addition, mostly studies modeled and forecasted future urban growth for studying climate change (Kim et al., 2016; Oo et al., 2019; Yang et al., 2017). However, urban spread and sprawl will bring ecological and environmental problems, e.g. urban heat islands, vegetation degradation, and reduction in biodiversity (Zhan et al., 2017). The ecological effect of the urban spatial-temporal growth is also important for urban plan and urban environment studies (Yang and Lo, 2003). Evaluation of the landscape pattern and ecological effects of urban growth will help to guide urban planning, and to protect and improve the regional ecological environment. Most studies have focused on the historical urban growth's ecological effect, not the future urban growth (Chen and Zhou, 2018; Zheng et al., 2019; Zhou et al., 2014), which is insufficient for future urban planning.

The SLEUTH (Slope, Land use, Exclusion, Urban growth, Transportation, and Hillshade) model is a cellular automata (CA) model that has become one of the most popular models for urban growth simulation (Chaudhuri and Clarke, 2013). There are some modified versions of SLEUTH model, for example, Guan and Clarke (2010) developed a parallel version of SLEUTH using the parallel Raster Processing Library (pRPL); Liu et al. (2012) utilized ant colony optimization to simplify the SLEUTH calibration procedures and also introduced sub-regional calibration to replace the entire study area calibration; Clarke (2018) replaced the brute force calibration method with a genetic algorithm (GA), which enhanced the computational speed and yielded the new SLEUTH-GA model. However, SLEUTH-GA is not very consummate and need future improvement. Finally, we selected SLEUTH, which has been widely used in the world, for modelling urban growth.

The aims of this study are: (1) predicting how government macro-control of the JJJ area will affect future urban growth during 2020–2050, by using SLEUTH 3.0 model and including the newest government policies; (2) assessing how the urban growth will impact on the JJJ ecological environment by studying the landscape pattern and ecological effect of future JJJ urban growth.

This study is organized as follows: firstly is Introduction; secondly is Methodology: SLEUTH model and landscape metrics used in the present study; thirdly is Results and Discussion: spatio-temporal variability of JJJ future urban growth and ecological effect of future urban growth; the last part is Summary. The flowchart of this study can be seen in Figure 1.

Figure 1

Location of the Jing-Jin-Ji study area (left) and its land use and land cover in 2015 (right).

Methodology

Study area

The JJJ urban agglomeration, with an area of 218,000 km2, is located in the northeast of China and belongs to the Bohai Economic Rim. The GDP was equivalent to ca. 10% of China in 2014 (Zhao et al., 2017). In 2016, JJJ had total population of 112 million, similar to that of Mexico (http://data.stats.gov.cn/english/easyquery.htm?cn=E0103). The JJJ includes three provincial-level cities: Beijing (the capital of China), Tianjin (municipality), Shijiazhuang (the provincial capital of Hebei); and ten prefectural-level cities: Baoding, Chengde, Langfang, Cangzhou, Zhangjiakou, Tangshan, Qinhuangdao, Hengshui, Handan, and Xingtai, each incorporating many districts and countries. JJJ is continental monsoon climate, hot and rainy in summer and cold and dry in winter. About 80% of total yearly precipitation appears in summer.

JJJ development will be central to the country's economic development plan in the next century. The Chinese government is planning the JJJ area as home to 130 million people over the equivalent area of New England, by 2050 (https://www.nbcnews.com). According to the government, the biggest change for JJJ development will be in transportation. The government approved $36 billion to build 700 miles of rail in 2017–2019. Twenty-four intercity railways are planned to be completed by 2050, with eight finished by 2020. According to the government, the goal is a “one-hour commuting circle” across the JJJ region. According to a strategic report, the JJJ megalopolis development is one of the three key infrastructure projects, along with the Yangtze River Delta Economic Region and the “One Belt, One Road” program, aimed at boosting China's economy over the next 100 years (http://theory.people.com.cn/n1/2017/0703/c412914-29377905.html).

SLEUTH model

Modeling approach

SLEUTH predicts urban extent based on a summation of four growth types: spontaneous growth, diffusive growth, organic growth, and road-influenced growth (Clarke et al., 1996). The four growth types are controlled by five coefficients: diffusion, breed, spread, slope, and road gravity. The detailed interpretation of four growth types can be seen in Clarke et al. (1996). Each coefficient has an initial range of 0–100, and is narrowed down at each calibration stage by comparing the simulated and historical data (Jantz et al., 2010). Calibration yields a fixed five-coefficient set which is then applied to urban growth prediction. The SLEUTH model can reveal the urban growth characteristics affected by economic, social and political factors in the calibration stage based on the historical data of the JJJ region. The workflow of this study can be seen in Figure 2.

Figure 2

The flowchart of this study.

The five input data sets (six if land use is predicted) of this study are listed in Table 1. The present study only simulates urban growth and does not study land use changes, so the land use layer is not included. For optimal calibration of the model, at least four urban time periods must be used, and at least two time periods of road layers. The urban layer was extracted from China land cover/land use products with 1 km spatial resolution (=http://www.resdc.cn/data.aspx?DATAID=184). This land use map was derived from Landsat TM/ETM+ and Landsat 8 images in 1995, 2005, 2010 and 2015 using visual interpretation, with overall accuracy of classification of over 90%. The excluded layer included water bodies and 16 national nature reserves. All input layers were unified to 1km spatial resolution and projected to the WGS 84, Zone 50N in the UTM system.

Table 1

Details of the main input datasets for SLEUTH modeling.

Input data	Data sources	Year	Resolution
Urban	Landsat TM/ETM+ and Landsat 8	1995, 2005, 2010, 2015	Raster, 1 km
Transportation	National Geomatics Center of China	1995, 2015	vector
Hill-shade	SRTM DEM	2015	Raster, 30 m
Slope	SRTM DEM	2015	Raster, 30 m
Excluded	Landsat TM/ETM+ and Landsat 8	2015	Raster, 1 km

Model calibration

Four steps were applied in this study during the calibration progress: coarse, fine 1, fine 2 and final calibration. Each coefficient range was expected to have become harrower at the end of each step. The OSM index (Optimal SLEUTH metric) (Eq. (1)) (Dietzel and Clarke, 2007) was used to derive the best group of five coefficients for prediction; however, there is no uniform standard for determination of the best coefficients (Dietzel and Clarke, 2007; Zhang, 2013). In this study, the top 20 parameter combinations of the OSM metric were used for selecting coefficients. We also tested the top 5 and top 50 selections, but found the top 5 selection missed and limited coefficients variation ranges. The top 20 and top 50 could capture the broad patterns, while the top 50 also capture more detailed variations but increased the number of calibration stages in order to get high OSM values, thereby increasing running time. The top 20 were selected as an intermediate option for SLEUTH model calibration in this study.where F-match = 1 in this study when land use prediction was not included, and the other seven indices were produced after each calibration step. Compare is the ratio of modeled and actual population for the final year, pop is the least squares regression score of modeled and actual urbanization for the control year, edges is the least squares regression score of modeled and actual urban edge count for the control year, clusters is the least squares regression score of the modeled and actual urban clusters for the control year, slope is the least squares regression of average slope of modeled and actual urbanized cells for the control year, Xmean and Ymean are the least squares regressions of average x and y values for modeled and actual urbanized cells for the control year (Nigussie Tewodros and Altunkaynak, 2017).

Model validation

The calibrated model was validated with observed data by comparing modeled and observed urban extent in 2015. The Kappa coefficient and cell-by-cell matching methods were utilized for the validation. The Kappa coefficient defines the agreement between two classifications on ordinal or nominal scales (Al-shalabi et al., 2013; Fleiss et al., 2004; Watkiss, 2008). The cell-by-cell method involves the spatial matching of pixels and is a good accuracy evaluation technique (Torrens, 2011). The cell-by-cell matching index is defined as the ratio of the number of overlay pixels in both images to the number of pixels in the observed image (Kuhnert et al., 2005).

Scenarios settings for prediction

Four scenarios were established: current trends; managed trends; ecologically sustainable with farmland protection; and ecologically sustainable with farmland, forest and grassland protection (Table 2). This study will evaluate the potential impacts of these new policies on JJJ urban growth. The primary policies are as follows (Figure 3). (1) Construction of railways and motorways in the JJJ area (see Table 2 for detailed information). (2) Synergistic development of Beijing, Tianjin and Hebei province, and developing the Xiong_An new urban area (with an area of 2000 km2) in Hebei province. The goal of this policy is to develop the new JJJ urban agglomeration, and is ranked by the Chinese government as a major event for the next millennium. (3) General urban planning for Beijing during 2016–2035, developing a new super international airport - Beijing Daxing International Airport - with an area of nearly 30 km2. (4) Urban planning of the Beijing vice-center, located in the Tongzhou district of Beijing.

Table 2

The four scenarios for predictions.

Scenario	Name	Description
Scenario 1 (sce 1)	Current trend	This scenario had no additional restrictions and followed the current trend for future urban growth. The excluded layer included all water bodies and 16 national nature reserves with value of 100, which meant that all the water bodies and reserve areas were fully excluded from new urban growth. The road layer also had pixel values of 100.
Scenario 2 (sce 2)	Managed trend	A policy was included for future urban growth, such as the new roads, new city, and new wetland park. The new railways and motorways included in this study were all projects planned and under construction in JJJ and the Beijing vice-center development (Figure 3). The new roads were the Jing-Xiong railway, Jing-Gang-Tai railway, Jin-Xiong railway, Jing-Kun railway, Shi-Xiong railway, Jing-Xiong motorway, Rong-Wu motorway, Jing-De motorway, and Jin-Shi motorway for the synergistic development of JJJ. In addition, the Jing-Tang railway and Jing-Qin motorway in Beijing vice-center development were added to the road layer. The new urban areas added in the urban extent layer were Xiong_An new city and Beijing Daxing International Airport. The planned Wenyu River wetland park was added to the excluded layer.
Scenario 3 (sce 3)	Ecologically sustainable I	In addition to the future policy, the farmland was partly protected in this scenario. In this case, urban development was limited and ecological resources were better protected. The farmland pixel values were set at 50, which meant farmland was 50% protected.
Scenario 4 (sce 4)	Ecologically sustainable II	In addition to policies in Scenario 3, the forest and grassland were also partly protected in this scenario, with values of 80 and 20, respectively.

Figure 3

(a) The planned railways and motorways and (b) the new city, airport and park in the synergistic development of Jing-Jin-Ji during 2015–2035, according to the online policy report (http://www.gov.cn/).

From the planning report, the planned railways, motorways, Xiong_An main city (with an area of nearly 100 km2) and the airport (with area of nearly 30 km2) were digitized using ArcGIS 10.3 software, and added to the road layer and urban extent layer for 2015. The planned new Wenyu River wetland park was included in the excluded layer for 2015. The 2015 input layers were used as seed layers for prediction.

Landscape metrics

Landscape pattern index

Three commonly used landscape metrics of the Compactness Index (CI), Patch density (PD) and Mean Shape Index (MSI) were selected to analyze landscape pattern of urban growth under the four scenarios (Table 3).

Table 3

Description of landscape pattern metrics and ecological effect metrics used in this study.

	Index	Equation	Ecological significance
Landscape pattern metrics	Compactness index (CI)(Ye et al., 2012)	CI=∑i=1nS_i/∑i=1nP_iwhere, Si and Pi are the area and perimeter of patch i, respectively. n is the total number of urban patches in the study area.	The larger the CI value, the more compact the land space pattern.
	Patch density (PD) (Chen and Fu, 1996)	PD=n/A where, A is the total urban area, n is same as above.	PD is the opposite of CI.
	Mean shape index (MSI) (Tian et al., 2003)	MSI=∑i=1nP_i4S_in	A larger MSI value indicates the space form is less coherent and the land space pattern is more dispersed.
Ecological effect metrics	Value of Ecological Service (VES) (Hou and Qiao, 2012)	VES=∑j=1mV_jS_jwhere Vj and Sj are the VES of land use type j per unit area, and area of land use type j, respectively. m is the total number of land use types in the study area, with m = 1 (urban type) in this study. Vj (10000 Yuan/km2) for urban, farmland, forest, grassland, water body, and unused land are 0, 6, 19, 64, 542, and 3.71.
	Ecological elasticity (ECO) (Xu et al., 2010)	ECO=∑j=1mr_je_jwhere rj and ej are the area ratio of land type j to the total study area and elasticity value of land type j, respectively. The values of ej for urban, farmland, forest, grassland, water body, and unused land are 0.4, 0.5, 0.9, 0.6, 0.8 and 0.3.	The ability of the ecosystem to return to its original state after disturbance. A higher ECO value indicates the ecosystem is more stable. ECO ranges from 0 to 1.
	Ecological risk index (ERI) (Zhou et al., 2014)	ERI=∑j=1mS_jω_jA_0where Sj is the area of land use type j, A0 is the sample area, A0 = 5 km × 5 km in this study, ωj is ecological risk intensity of land used type j, the values of ωj for urban, farmland, forestland, grassland, water body are 0.29, 0.13, 0.03, 0.05 and 0.04.	The lower ERI value, the better for ecosystem. ERI value ranges from 0 to 1.

Ecological effects index

The Value of Ecological Service (VES), ECOlogical elasticity (ECO) and Ecological Risk Index (ERI) were selected to analyze ecological effects of future urban growth (Table 3). These three indices are all considering all land use types (e.g. urban, farmland, forest, grassland, water, etc.); however, there was only “urban” type and no other land use types in the present study. Therefore, we used urban area change under scenarios to study ecological effect change. For example, ΔVES, ΔECO and ΔERI of farmland change between scenario 2 (managed trend) and scenario 3 (farmland protection) were studies here (Eqs. (2), (3) and (4)). The farmland area change between scenario 2 and scenario 3 is equal to the urban area change.where, Vfarmland = 610000 Yuan/km2, and Surban_sce2 and Surban_sce3 are urban areas in scenario 2 and scenario 3.where, efarmland = 0.5, A is the study area.where, ωfarmland = 0.13, A0 = 5 km × 5 km. The study area was divided into 16905 sampling grid units of size A0. ΔERI was calculated for each sample grid unit, then interpolated by ordinary Kriging and grouped into three grades (low, medium, high) using the Natural Breaks classification method.

Results

Calibration and validation of the SLEUTH model

Calibration

Calibration was carried out under phases termed coarse, fine 1, fine 2, and final. After each calibration phase, the OSM index was calculated based on Eq. (1) and the top 20 highest OSM values were used to narrow down the coefficient ranges. The ranges of the five coefficients after each calibration phases are listed in Table 4. The highest OSM index value after the final calibration was 0.7, which is higher than that achieved in other studies, e.g. 0.5 (Nigussie Tewodros and Altunkaynak, 2017; Zhang, 2013), 0.49 (Sakieh et al., 2015b), 0.44 (Sakieh et al., 2015a). A higher OSM indicates that the urban extent was simulated more accurately when compared with the historical urban extent data. The best-fit coefficients after final calibration were (1, 1, 1, 99, 48); these values were used for prediction. The slope and road coefficients had large values, which showed slope and road extent had the greatest impacts on urban growth between 1995 and 2015.

Table 4

Selected top 20 of each coefficient for the next calibration stage, corresponding to the top 20 highest OSM values.

Coefficient	Coarse calibration			Fine1 calibration			Fine2 calibration			Final calibration			Best-fit	100 MC average	Highest OSM
	Start	Step	Stop	Start	Step	Stop	Start	Step	Stop	Start	Step	Stop
Diffusion	0	25	100	0	15	75	0	5	30	0	3	30	27	1	0.7
Breed	0	25	100	0	20	100	0	12	60	0	10	50	40	1
Spread	0	25	100	0	1	25	0	1	5	0	1	5	2	1
Slope	0	25	100	0	20	100	0	10	100	10	10	90	90	99
Road	0	25	100	0	20	100	0	12	60	0	10	50	50	48

Validation

The simulated urban growth in 2015 using the best-fit coefficients of (1, 1, 1, 99, 48), with 2010 as a seed year, was compared with the observed urban extent in 2015 (Figure 4); this showed strong similarity between modeled and observed extent, with only a few pixels being over- or under-estimated. The over-estimate pixels represented the non-urban areas in the historical data which were simulated as urban, and vice-versa for under-estimated pixels.

Figure 4

Comparison of simulated and observed urban extent in 2015 showing overlay pixels (red), over-estimated urban pixels (blue), and under-estimated urban pixels (green).

The error matrix showed that 187841 km2 of non-urban land and 19397 km2 of urban land were predicted correctly. The accuracy of prediction based on the cell-by-cell matching method was (187841 + 19397)/215188 × 100% = 96.3%. The Kappa coefficient (calculated based on the error matrix) was 0.81 (Table 5). Generally, a Kappa coefficient value of 0.41–0.6 implies moderate simulation performance, with 0.61–0.8 being superior (Li et al., 2009). Therefore, the best-fit coefficients could accurately reveal the historical urban growth trend and can be used for future prediction.

Table 5

The error matrix for simulated and historical urban extent in 2015.

		Historical in 2015
		Non-urban	Urban	Sum
Simulated in 2015	Non-urban	187841	624	188465
	Urban	7326	19397	26723
	Sum	195167	20021	215188
Kappa coefficient		0.81

Spatial and temporal distributions of future urban growth

The urban areas of JJJ and its sub-regions (Beijing, Tianjin, Hebei) during 2020–2050 would continue to grow under all four scenarios, reaching about 4 × 104 km2 by 2050 in the JJJ area (Figure 5a-d). Ecological protection (sce 3 and sce 4) obviously restrained the urban expansion, resulting smaller urban areas than those without ecological protection (sce 1 and sce 2) (Figure 5a-d). The ratios of urban area difference between sce 1 and sce 2 in each sub-region area are shown in Figure 5e. The JJJ urban area under sce 2 was similar to that under sce 1, with a ratio close to 0 %; the urban area in Beijing decreased (ratio less than 0.5%), in Tianjin increased during 2020–2023 and then decreased after 2023 (ratio less than 0.4%). However, the Hebei urban area increased under sce 2, with a ratio of less than 0.1% indicating a partial migration of the future population in Beijing and Tianjin towards to Hebei province under the government's policies. With ecological protection, the urban area decreased in the order Tianjin > Beijing > Hebei > JJJ (Figure 5f). That showed the urban area of Tianjin would increase larger than Beijing and Hebei in 2020–2050 without ecological protection. Furthermore, ecological environment of Tianjin would also the worst under the new policies.

Figure 5

Urban area growth in different regions under four scenarios during 2020–2050 (a. JJJ, b. Beijing, c. Tianjin, d. Hebei) and the difference in urban area between different scenarios (e. between sce1 and sce2, f. between sce2 and sce4). The ratios in (e) and (f) are the contributions to changes in urban area from each-subregion.

The spatial patterns of urban area differences under different scenarios were compared (Figure 6). The pattern “sce2-sce1 < 0” indicated where urban growth was restricted by government policies, while “sce2-sce1 > 0” represented where urban growth was increased under the policies; overall, the area of lost urban land in JJJ by 2050 (2211 km2) was only 41 km2 larger than area of gained urban land (2170 km2) (Figure 6a), with net decreases of 23 km2 in Beijing, 16 km2 in Tianjin and 2 km2 in Hebei (Table 6). These changes indicate that government policies would only have a very small impact on JJJ urban growth by 2050. However, under the farmland protection scenario, 4380 km2 of protected farmland was retained in JJJ; this would have been changed to urban area by 2050 without environmental protection (Figure 6b). The largest protected area was in Hebei province, covering 3671 km2 (Table 6). Under sce 4, 4754 km2 farmland, forest and grassland area was protected from conversion to urban land, with the largest protected area again being in Hebei (3882 km2) (Figure 6c).

Figure 6

The spatial distribution of the differences in urban growth under different scenarios in 2050. (a) scenario 2 minus scenario 1, (b) scenario 2 minus scenario 3, (c) scenario 2 minus scenario 4.

Table 6

The urban area difference in 2050 under different scenarios (negative values indicate reduced urban area).

	Beijing (km2)	Tianjin (km2)	Hebei (km2)
sce2-sce1	-23	-16	-2
sce2-sce3	326	383	3671
sce2-sce4	470	402	3882

Ecological effect of future urban growth

The CI increased annually from 2020 to 2050 under the four scenarios, while the PD decreased (Figures 7a, 7b), showing that the future JJJ urban growth was characterized by compact development meeting the current planning strategy for intensive urban development. However, the government policies had no obvious effect on CI or PD. The MSI decreased during 2020–2050, showing the spatial pattern of urban growth was becoming increasingly regular and compact (Figure 7c). Nonetheless, the MSIs under scenarios 1 and 2 before 2027 were evidently different. Between 2020-2025, the spatial pattern of urban areas in JJJ under scenario 2 was more regular than that in scenario 1; however, the distribution was irregular and dispersed, with a larger MSI, than that in scenario 1 during 2026–2027 (Figure 7c).

Figure 7

Temporal changes in (a) compactness index (CI), (b) patch density (PD) and (c) mean shape index (MSI) during 2020–2050, under the four scenarios in the JJJ region (sce1, sce2, sce3 and sce4).

The value of ecological service would be increased during 2020–2050 under scenario 3 with farmland protection when compared to scenario 2 without environment protection, to a maximum value of 31.7×108 Yuan/km2 in 2031 (Figure 8a). The ecological elasticity was also increased under scenario 3 when compared to scenario 2 (Figure 8b). The proportion of land with a high reduction of ecological risk increased from 2020 to 2050, while the area with a low reduction decreased and the area with moderate reduction showed no obvious change (Figure 8c). The region of high reduction of ecological risk was distributed in farmland surrounding urban areas, showing that farmland around cities was more exposed to ecological risk than farmland far away from cities.

Figure 8

The differences in (a) value of ecological service (ΔVES), (b) ecological elasticity (ΔECO) and (c) the ecological risk index (ΔERI) between scenario 2 and scenario 3 in the JJJ region. c1 is the ecological risk reduction in 2020 under farmland protection (sec3), and c2 is the equivalent value in 2050.

Conclusions

In order to develop JJJ area as a new urban agglomeration, the Chinese government has implemented some policies since 2016. Some of these polices are clear (for example, a new city, wetland boundary, motorway route, etc.), but some are undefined (for example, the nine wedge-shaped green corridors have no explicit boundary at present). This study assimilated the explicit city planning policies based on the JJJ synergy development report and Beijing general plan report (2016–2035), including railways, motorways, Xiong_An new city, a new airport, wetland park, and Beijing vice-center. We investigated the impact of these policies on JJJ future urban growth and resulting ecological effects. The temporal and spatial patterns of JJJ future urban growth during 2020–2050 were simulated by the SLEUTH model under four different scenarios. The urban area in JJJ increased annually during 2020–2050 under all four scenarios, but the urban areas in scenarios 3 and 4 (with ecological protection) were much smaller than those under current (scenario 1) and managed policies (scenario 2). The new policies had a negligible effect on the JJJ urban growth, with very similar urban areas under scenario 2 and scenario 1. However, sub-regional urban growth showed some small differences: the urban areas in Beijing and Tianjin decreased, while that in Hebei increased, under the new policies. With ecological protection, urban area reduced in the order Tianjin > Beijing > Hebei > JJJ, which showed that Tianjin would undergo the greatest conversion of natural surfaces to urban areas if there was no environmental protection.

The landscape pattern and ecological effects of future urban growth under the four scenarios were also analyzed. This study utilized the compactness index (CI), patch density (PD) and mean shape index (MSI) to characterize variations in the landscape patterns, and value of ecological service (VES), ecological elasticity (ECO) and ecological risk index (ERI) to assess ecological effects. Results showed that future urban growth in JJJ was characterized by compact development, with increased CI and decreased PD and MSI. However, the government policies showed no obvious impact on changes in landscape patterns, with very similar CI, PD and MSI values under scenarios 1 and 2. With farmland protection, VES in the JJJ region was increased during 2020–2050, reaching a maximum value of 31.7×108 Yuan/km2 in 2031; the ecological risk was obviously reduced around urban areas and the ecological elasticity was also improved. Therefore, the urban planners and managers should consider not only the economic benefits of urban growth, but also the ecological effect. The ecological environment is very important for human health and comfort.

The accuracy of the present results may be increased when more policies are included, e.g. green corridors, parks around Beijing's 5th and 6th roads, permanent farmland, etc. However, these boundaries are not clear at present and it is unable to include that in the present study. As more policies are established, we will continue to study the urban growth and ecological effects of these additional policies, so that we can provide more comprehensive guidance for urban planners and managers. The climate effect of the future JJJ urban growth under government policies should also to be studied in next step.

Declarations

Author contribution statement

Nana Li: Conceived and designed the experiments; Contributed reagents, materials, analysis tools or data; Wrote the paper.

Shiguang Miao: Performed the experiments.

Yaoting Wang: Analyzed and interpreted the data.

Funding statement

This work was supported by the (Z201100008220002) and the Youth Beijing Scholars Program under grant number (2018-007).

Data availability statement

Data will be made available on request.

Declaration of interests statement

The authors declare no conflict of interest.

Additional information

No additional information is available for this paper.