Land Use Dynamics and Optimization from 2000 to 2020 in East Guangdong Province, China

: Anthropogenic land-use change is one of the main drivers of global environmental change. China has been on a fast track of land-use change since the Reform and Opening-up policy in 1978. In view of the situation, this study aims to optimize land use and provide a way to effectively coordinate the development and ecological protection in China. We took East Guangdong (EGD), an underde-veloped but populous region, as a case study. We used land-use changes indexes to demonstrate the land-use dynamics in EGD from 2000 to 2020, then identiﬁed the hot spots for fast-growing areas of built-up land and simulated land use in 2030 using the future land-use simulation (FLUS) model. The results indicated that the cropland and the built-up land changed in a large proportion during the study period. Then we established the ecological security pattern (ESP) according to the minimal cumulative resistance model (MCRM) based on the natural and socioeconomic factors. Corridors, buffer zones, and the key nodes were extracted by the MCRM to maintain landscape connectivity and key ecological processes of the study area. Moreover, the study showed the way to identify the conﬂict zones between future built-up land expansion with the corridors and buffer zones, which will be critical areas of consideration for future land-use management. Finally, some relevant policy recommendations are proposed based on the research result.


Introduction
Humans are simultaneously confronting environmental problems on multiple fronts, such as climate change, loss of biodiversity, soil degradation, water pollution, and loss of ecosystem services, and each of these is caused either directly or indirectly by land-use changes [1][2][3][4][5][6][7]. It is estimated that~60% of global land-use changes are directly associated with human land-use activities and 40% with indirect drivers, such as climate change [4]. The expansion of built-up land in urban and rural areas, agricultural intensification, energy, and material consumption are the primary drivers of land-use change.
Land-use and land cover change (LULC) is a global subject of study. Most studies have focused on metropolitan areas [8][9][10], and other environmental targets of interest, such as the tropics [11,12], karsts [13], coastal zones [14], ecosystem services [5,15], climate change [16][17][18], etc. As a developing country, China has been on a fast track of land-use change since the Reform, and Opening-up Policy was established in 1978, an essential part of China becoming one of the fastest global economies [19][20][21][22][23]. In particular, rapid urban and rural expansion plays an important role in China's land-use change over the past 40 years. Although built-up land expansion is considered a sign of prosperity; however,

Study Area
Chaozhou and Shantou are located in the east of the Guangdong Province (23 • 21 N-24 • 14 N, 116 • 14 E-117 • 16 E), where the coastline is facing the South China Sea (Figure 1). This region encompasses the unique, local Chao-Shan culture; people in this region share a similar lifestyle, the same cuisine, and the same Chao-Shan dialect. The total combined area of these two cities is 6632 km 2 (including the sea territory), with a population of 8.32 million. Chaozhou is in the northeast of the EGD, while Shantou lays in the southwest. Chaozhou has three administrative regions at the county-level: Raoping County, Chao'an, and Xiangqiao Districts, while Shantou city has seven county-level regions, which are the districts of Jinping, Longhu, Haojiang, Chenghai, Chaoyang, Chaonan, and Nan'ao County. Even though the EGD is one of the most populous regions of the Guangdong Province, and many counties and towns are considered to be peri-urban areas, when comparing GDPs with the Pearl River Delta, it is apparent that EGD is the lesser developed area [70,71].
The terrain of the montane north and northwest regions in EGD is >1000 m and slopes gently down towards the southeast. There are three important rivers in the EGDthe Huanggang, Hanjiang, and Lianjiang-and the croplands and primary settlements concentrate along the Huanggang River valley, Hanjiang River plain, and Liangjiang River valley.

Data Sources
The land-use land cover data for 2000, 2010, and 2020 were obtained from the GLO-BELAND30 platform (http://www.globallandcover.com (28 September 2020)), at a 30 m resolution [72]. In the GLOBELAND30 platform, the classification system is as follows: cropland, forest, grassland, shrubland, wetland, water bodies, tundra (subclass are shrub tundra, grass tundra, wetland tundra and bare land tundra), artificial surfaces (builtup land), bare land, permanent snow and ice. The following classifications were found for EGD: cropland, forest, grassland, shrubland, wetland, waterbody, artificial surfaces (built-up land), bare land, and seawater. A 30 m digital elevation model (DEM) was acquired from the 91Weitu website (www.91weitu.com (25 September 2020)). Vectors of the EGD administrative region were drawn by the authors according to the regional maps of the Department of the Natural Resources of Guangdong Province's website (http://nr.gd.gov.cn/gdlr_public/map/3/index.html (25 September 2020)).

Data Sources
The land-use land cover data for 2000, 2010, and 2020 were obtained from the GLOBELAND30 platform (http://www.globallandcover.com(28th,September 2020)), at a 30 m resolution [72]. In the GLOBELAND30 platform, the classification system is as follows: cropland, forest, grassland, shrubland, wetland, water bodies, tundra (subclass are shrub tundra, grass tundra, wetland tundra and bare land tundra), artificial surfaces (built-up land), bare land, permanent snow and ice. The following classifications were found for EGD: cropland, forest, grassland, shrubland, wetland, waterbody, artificial surfaces (built-up land), bare land, and seawater. A 30 m digital elevation model (DEM) was acquired from the 91Weitu website (www.91weitu.com(25th,September 2020)). Vectors of the EGD administrative region were drawn by the authors according to the regional maps of the Department of the Natural Resources of Guangdong Province's website (http://nr.gd.gov.cn/gdlr_public/map/3/index.html (25 September 2020)).
Normalized difference vegetation index (NDVI) data were composed [73,74] using Landsat 8 data (the images downloaded from www.earthexplorer.usgs.gov(30th, September 2020), image date: October 2019, images resolution: 30 m, path/row: 120/43 and 120/44, Vectors of regional railways, roads, highways, town centers, and ecologically sensitive areas were obtained by the master city planning of Chaozhou and Shantou. Socioeconomic statistics were derived from provincial and municipal level data. All raster data were rescaled to 30 m.

Methods
First, we used four indexes to analyze the land-use dynamics in EGD. Second, the methods of the land-use simulation process were demonstrated. The Markov chain model, which predicted the quantity of land-use types in 2030, was introduced. Then the detail of the FLUS model and the validation process were presented. Third, the concept and the process of establishing the MCR model were also presented. Finally, the ESP can be synthesized.
The schematic methodology is shown in Figure 2.
First, we used four indexes to analyze the land-use dynamics in EGD. Second, the methods of the land-use simulation process were demonstrated. The Markov chain model, which predicted the quantity of land-use types in 2030, was introduced. Then the detail of the FLUS model and the validation process were presented. Third, the concept and the process of establishing the MCR model were also presented. Finally, the ESP can be synthesized.
The schematic methodology is shown in Figure 2.

Land Use Dynamics Analysis
With the characteristic of land-use change [42], two indices, the annual increase (AI, km 2 ·y −1 , Equation (1)), and annual growth rate of croplands and built-up land (AGR, %, Equation (2)) were introduced to reveal the dynamics of these four typical and signature land-use types of EGD [77][78][79][80]: where ( ) and are the areas of crop-(built-up) land at the initial and end times, respectively, and d is the analysis period in years (one decade). DEN stands for the proportion of built-up land in the study area. In Equation (3), denotes an artificial surface area, and A denoted the total area. The warning threshold for identification of

Land Use Dynamics Analysis
With the characteristic of land-use change [42], two indices, the annual increase (AI, km 2 ·y −1 , Equation (1)), and annual growth rate of croplands and built-up land (AGR, %, Equation (2)) were introduced to reveal the dynamics of these four typical and signature land-use types of EGD [77][78][79][80]: where A start (A built ) and A end are the areas of crop-(built-up) land at the initial and end times, respectively, and d is the analysis period in years (one decade). DEN stands for the proportion of built-up land in the study area. In Equation (3), A built denotes an artificial surface area, and A denoted the total area. The warning threshold for identification of "overdevelopment" according to international criteria are areas where DEN is >30%. Then we analyzed the increase in DEN increased rate (IR) where DEN 2010 and DEN 2020 are the DEN values in 2010 and 2020, respectively.

Markov Chain Model
To predict the quantity of each land-use type, a Markov chain model was introduced. Commonly used to predict geographical characteristics lacking after-effect events [81][82][83], this model has become an important prediction method in geographical research [84]. In this study, the Markov chain model calculated the probability matrix of land conversion and then analyzed the mutual transformation relationship of the different land-use types. By using a module integrated within FLUS software, the total amount of future land-use types in 2030 were predicted with a decade-long step of the transformation matrix.

FLUS Model
The FLUS model is software that combines the "top to bottom" system dynamics (SD) model and "bottom to top" CA model [36] and provided a multiple CA allocation model for simulating land-use change and scenario analysis. The "top to bottom" represents macroscale demands, such as political planning and background climate influences, while the "bottom to top" determines the system's evolution from a local perspective [36]. By considering various socioeconomic and natural environmental factors, the SD model was used to project the land-use scenario demands, while the CA model processed complex interactions among the different land-use types [36,37]. FLUS has shown to have a higher simulation accuracy than the CA and CLUE-S models [36]. The model and further information can be found online (http://www.geosimulation.cn/flus.html (28 October 2020)).
In this study, nine classes of land-use were used (Section 2.2), and for an ANN-based suitability probability estimation module, a random sampling of 1‰ of the land-use raster was selected with 10 hidden layers. The driving factors included environmental and socioeconomic factors: DEM, NDVI, ecologically sensitive areas, town centers, railways, and main roads. Moreover, the factors were trained by the model and land-use data to obtain a probability distribution map. Then self-adaptive inertia and competition mechanism CA module were used to simulate the land-use change in 2030 based on the starting year of 2020.
Simulation results were validated using both Cohen's Kappa coefficient and figures of merit (FoM) provided in the FLUS model. Some studies have shown that FoM is superior to the Kappa coefficient in assessing the accuracy of simulated changes [85][86][87]. The FoM is the ratio of the intersection of the observed change and predicted change to the union of the observed change and predicted change [86,88]. Equation (5) of FoM is as follows: where B is the area of correct due to observed change predicted as change, A is the area of error from observed change predicted as persistence, C is the area of error from predicted change to an incorrect category, and D is the area of error from observed persistence predicted as change. Our study focused on the validation by the FoM method, while the kappa coefficient was still demonstrated for references.

MCR Landscape Analysis
In MCR landscape analysis, we used three steps: defining the ecological source, building the ecological resistance surface and identifying key ecological corridors. In the second step, the ecological resistance surface was established according to the MCRM.

Source Identification
Ecological source originally identified the habitat that species most relied upon [48,89], and recently this definition has been expanded to include those patches, which plays a crucial role in overall ecosystem health and are important for urban/regional ecological security [49]. In the present study, sources represent the main core areas for a particular ecological or artificial process. Two kinds of processes were acknowledged: natural or ecological processes and the expansion of human activity, categorized into the ecological sources (sources mode) and built-up modes, respectively. Ecological sources were obtained from the Chaozhou and Shantou master planning data, which identified the ecologically important and sensitive regions. Ecological sources included: forests of protected regions, protected drinking water reservoirs, and some important habitats for certain species. For sources of the built-up mode, we chose the built-up land from the 2020 land-use map >10 hm 2 .

MCRM Surface
Resistance means the difficulty for the species or ecological flows the move or travel to other places [89,90]. In the MCRM, the space of the study area was considered as a surface with different resistance values, and the resistance values were calculated in different evaluation systems. The MCRM is a mainstream assessment method, and it was widely applied to landscape design, which aims to protect natural ecosystems within specific secure borders [47], biodiversity conservation, and urban planning [62,90,91].
The MCRM calculates the required cost of species' moving from the source to the destination. It reflects the potential possibility and tendency of species' movement [48]. The MCRM can also be used to simulate different ecological flows across different landscape surfaces by establishing a minimum cumulative resistance path [49,92,93].
In our study, the MCRM surface was evaluated, and the evaluation contained two primary factors: natural and anthropogenic conditions. The natural factors included: DEM, NDVI, land-use types, geological hazards, risk (i.e., landslide or debris flow) and distance from water (rivers, ponds, reservoirs). Anthropogenic influences included: distance from built-up land, distance from roads (main city roads and highways). Elevation and NDVI data were standardized as the following equation [94]: where x * represents the standardized value, x represents the original pixel value, max and min represent the maximum and minimum pixel value, respectively. Geological hazard data were obtained from Chaozhou and Shantou geological planning, which defined risk into four levels: comparatively safe, low, medium, and high-risk. Using previous studies [48,49,91,94,95] and the conditions of the EGD, all factors were assigned scores from 0 to 100, where a higher score indicates a higher resistance. The weight of each factor was calculated using the analytic hierarchy process (AHP) method.
The equation of the MCRM is as follows (Equation (7)): where f represents the relative function to the minimal resistance; D ij represents the spatial distance of a target unit from the source point j to the land-use type i; and R i represents the ecological resistance of the spatial unit i to a target's migration. The MCRM contains two parts described as follows (Equation (8)) [94]: where the MCR represents the total MCR, MCR source represents the MCRM based on the ecological sources, and MCR built−up represents the MCRM based on the built-up land expansion. MCR source and MCR built−up represent the difficulty for the land-use type to expand or develop from the center of the source to the surrounding areas. The higher the MCR, the more difficult the expansion, and details of the factors and scores with the weight can be seen in Table 1.  After identifying the ecological sources and setting up the resistance surface, it was necessary to connect them based on the MCRM surface analysis. The least-cost path method was applied to establish the corridors, and the corridors connect the ecological sources. In our study, the least cost path was calculated by the tool of linkagemapper (https://circuitscape.org/linkagemapper/ (14 January 2021)) in ESRI ArcGIS (v.10.2), and the linkagemapper is a commonly used tool in ecological and habitat studies [96,97]. A 40 km threshold was used to truncate corridors between sources, with a maximum number of connected nearest neighbors set to four after tested. Key nodes, which represent the place that had equal resistance values between every two ecological sources, were considered to be crucial to the process of maintenance and migration of species [48]. Moreover, the key nodes were identified through the watershed method toolbox in ArcGIS. Buffer zones of 600 m were set up around the center of each corridor. Conflict zones are the most sensitive and fragile regions in need of future protection. We simulated the increase in built-up land for 2030 using the FLUS model and labeled the intersection of areas where new built-up land occurred and buffer zones as conflict zones. The conflict zones are the concern for future land-use management.

Synthesize the ESP
Apart from the built-up land (artificial surface) of 2020, the MCRM surface result was classified into four classes with natural breaks (Jenks) using ArcGIS, each representing the relative ecological importance: disturbed class, where the most frequent human activities occur; low resistance class, the most important areas, which may require protection any time; medium resistance class, average importance of the EGD; and high-resistance class, areas that need the least protection.
With the ecological sources, corridors, buffer zones, and key nodes ad the resistance classes, an ecological security pattern (ESP) was established. By analyzing the ESP, the corridors, key nodes and the spatial conflict between the land-use simulation results in 2030 of the FLUS model with these patterns, ecologically sensitive and fragile regions were identified, presenting the optimized land use. Table 2 shows the areas and proportion of land use for the three time periods. As shown in Figure 3, forests are the most distributed land-use type in EGD, located primarily in the north and southwest. Cropland is the second most distributed land-use type, mainly found in the north, east, central, and southeast of EGD. Croplands and built-up land received the most attention from previous studies of EGD land-use [42,98]. Although the quantity increased, the proportion of cropland remained nearly the same in the first decade (Table 2). Cropland AI and AGR were 239.67 hm 2 ·y −1 and 0.14%, respectively, from 2000 to 2010, while its quantity dropped down by ~5% from 2010 to 2020, where AI was −3856.44 hm 2 ·y −1 , and AGR was −2.52%. Built-up land was stable from 2000 to 2010, with an AI and AGR of −219.12 hm 2 ·y −1 and −0.37%, respectively, but it increased from 8.67% to 15.31% from 2010 to 2020, with an AI and AGR of 4386.82 hm 2 ·y −1 and 5.85%, respectively (Table 3). Since EGD is located along the South China Sea, the study here also focused on the coastline and the sea. Seawater area dropped to 19.75%, indicating that EGD acquired land from the sea during 2010-2020 (Table 2). Croplands and built-up land received the most attention from previous studies of EGD land-use [42,98]. Although the quantity increased, the proportion of cropland remained nearly the same in the first decade (Table 2). Cropland AI and AGR were 239.67 hm 2 ·y −1 and 0.14%, respectively, from 2000 to 2010, while its quantity dropped down by~5% from 2010 to 2020, where AI was −3856.44 hm 2 ·y −1 , and AGR was −2.52%. Built-up land was stable from 2000 to 2010, with an AI and AGR of −219.12 hm 2 ·y −1 and −0.37%, respectively, but it increased from 8.67% to 15.31% from 2010 to 2020, with an AI and AGR of 4386.82 hm 2 ·y −1 and 5.85%, respectively (Table 3). Since EGD is located along the South China Sea, the study here also focused on the coastline and the sea. Seawater area dropped to 19.75%, indicating that EGD acquired land from the sea during 2010-2020 (Table 2).

Land Use Transition Matrix Analyses
In the first analysis period (2000-2010), the most significant transformation was from built-up land to cropland, accounting for 5400 hm 2 ( Table 4). The transformation from cropland to a water body (4821 hm 2 ), and the reverse (4360 hm 2 ), were also high. The amount of area converted from grassland to forest was also nearly equal to the opposite conversion. Table 5 shows the transformation areas from 2010 to 2020, where the strongest pattern was cropland becoming built-up land (36,050 hm 2 ). Croplands were also being encroached upon by water bodies (>4500 hm 2 ). Over the same period, the sea area was also diminished by built-up land, at a scale of 1242 hm 2 .  5 show that during the two decades of analysis, croplands in EGD decreased from built-up land expansion and conversion to water bodies, where the latter can be explained by local farmers changing focus from traditional agriculture to aquaculture. Builtup land gains were derived primarily from croplands, and the sea and total conversion into built-up land took up a large proportion of all the land-use transformations. The tradeoff between cropland and built-up land was the most significant problem of the EGD over the past two decades, and it will likely continue to be a central priority of territorial planning for the future.

Land Use Change Characteristic Analysis
Two typical land-use types-built-up land and cropland-were chosen to analyze their changing spatial characteristics.
Over the previous two decades, transitioning into built-up land was the most significant land-use change of the EGD. From 2000 to 2010, built-up land expansion was scattered throughout EGD, and the relatively small-scale expansions were located in the central and southern Chao'an District, south of the Chenghai District, along the Lianjiang River valley in the southeast of EGD, and road construction in the south of Raoping County. In the second decade of analysis, built-up land expansion was at a significantly larger scale and distributed more broadly. The largest expansion during this period was located along the sea, in a reclamation area from Shantou. The south and middle of Chao'an County, the north and west of Chenghai District, and expansion along the Lianjiang River valley were also quite high.
The amount of cropland decreased during these two decades. From 2000 to 2010, the decrease in cropland was relatively small and mainly occurred around the Chaoyang and south of Chao'an Districts. For the second period, the decrease in cropland was broadly located throughout the entire study area, and these spatial distributions were highly consistent with the expansion of built-up land in the south and central Chao'an District, west and south of the Chenghai District, the center of Shantou, and along the Lianjiang valley. In the southwest of Raoping County, the decrease in cropland derived from conversion into water boded for the aquaculture.

Development Intensity
It was found that built-up land expanded significantly from 2010 to 2020. According to E (3), the DEN of the entire EGD in 2010 and 2020 were 10.73% and 18.92%, respectively. The DEN of all county levels was from 2.15% to 33.19% in 2010, with the highest intensities occurring mainly in the center of Shantou, particularly in the Longhu and Jinping Districts (Table 6). At the same time, the Chaoyang and Chaonan Districts were also in a relatively high-intensity location. In 2020, it was apparent that the DEN increased at a rapid pace, where Longhu and Jinping Districts still maintained the largest two DEN values, and five other district/counties were also >20%. Haojiang District and Nan'ao County increased by >200%, indicating that these two administrative regions developed at a significant pace. DEN at the township-level was also analyzed (Figure 4). In 2010, 37 of the 122 towns in EGD exceeded 30%. These towns are mainly distributed in the center of Shantou, south of the Chenghai District, the center of Chao'an District, and north of Chao'an District. In 2020, 57 of the 122 towns exceeded the international criteria, distributed from the central Chao'an District to the southeast of Shantou and along the Lianjiang River valley. Meanwhile, there were 16 additional towns > 20%, together indicating that the built-up land expansion was intensive over the previous decade.   Supplementary Table S1).

Land Use Prediction and Simulation Results
Based on the land-use data of 2000 and 2010, a Markov chain prediction module in the FLUS software was used to predict land use in 2030. For validation, the kappa coefficient and FoM were calculated as 0.86 and 0.35, respectively, which were the precision validation parameters of 2020 [35,36,87]. The parameters indicated that 2030 predictions  Supplementary Table S1).

Land Use Prediction and Simulation Results
Based on the land-use data of 2000 and 2010, a Markov chain prediction module in the FLUS software was used to predict land use in 2030. For validation, the kappa coefficient and FoM were calculated as 0.86 and 0.35, respectively, which were the precision validation parameters of 2020 [35,36,87]. The parameters indicated that 2030 predictions were reliable, and the simulation results can be seen in Table 7. The simulated results shown in Figure 5 indicated that the expansion of built-up land continued to take over cropland s of the EGD, particularly in the Lianjiang River valley, south and central Chao'an District, south Raoping County, and east of Shantou, all of which are presently occupied by tremendous croplands. Furthermore, the built-up land encroached upon the sea for all districts and counties along the South China Sea. Far fewer land-use changed were predicted in the north and central EGD.

Source Distribution and MCRM Results
As mentioned above, the ecological sources were obtained from the Chaozhou and Shantou city master plans. According to the land-use types, the primary ecological sources were forests and water bodies (i.e., rivers and reservoirs) scattered throughout EGD. Spatially, both were concentrated in the north and southeast, such as the northern Chao'an

Source Distribution and MCRM Results
As mentioned above, the ecological sources were obtained from the Chaozhou and Shantou city master plans. According to the land-use types, the primary ecological sources were forests and water bodies (i.e., rivers and reservoirs) scattered throughout EGD. Spatially, both were concentrated in the north and southeast, such as the northern Chao'an District, northern and central Raoping County, southern Chaonan District, central Chaoyang District, and two-thirds of Nan'ao County. These resources are currently protected by law and cannot be destroyed or encroached upon by any authorities or persons. Built-up mode sources were readily recognized in Figure 6, showing the MCRM calculation results of the two modes. The disturbing layer accounted for ~21.13% of the total study area (1401.34 km 2 ), located mainly around the Lianjiang River valley, the center of Shantou, southern and central Chao'an and Xiangqiao Districts, and a belt along the coastal sea (Figure 7). The low class accounted for ~10.67% of the study area (707.63 km 2 ) and was mainly concentrated in the north and southwest, corresponding to the montane and pelagic regions of EGD. The medium class proportion was 28.82% (1911.34 km 2 ), and located adjacent to the low class south of Chaonan District, central Chaoyang and Chao'an Districts, and northern Raoping County. The high class accounted for the greatest proportion of the EGD, 39.37% (2611.02 km 2 ), and surrounded the disturbing layer. This phenomenon showed that the high class had a strong correlation with human activities, and any further expansion of built-up land would encroach the high-resistance areas first. The disturbing layer accounted for~21.13% of the total study area (1401.34 km 2 ), located mainly around the Lianjiang River valley, the center of Shantou, southern and central Chao'an and Xiangqiao Districts, and a belt along the coastal sea (Figure 7). The low class accounted for~10.67% of the study area (707.63 km 2 ) and was mainly concentrated in the north and southwest, corresponding to the montane and pelagic regions of EGD. The medium class proportion was 28.82% (1911.34 km 2 ), and located adjacent to the low class south of Chaonan District, central Chaoyang and Chao'an Districts, and northern Raoping County. The high class accounted for the greatest proportion of the EGD, 39.37% (2611.02 km 2 ), and surrounded the disturbing layer. This phenomenon showed that the high class had a strong correlation with human activities, and any further expansion of built-up land would encroach the high-resistance areas first.

Ecological Corridors, Key Nodes, and Buffer Zones
One hundred seventeen corridors were established by the linkagemapper based on the MCRM surface analysis, and the lengths of the corridors were between 30 and 22,245 m ( Figure 8). The mountainous regions of the north, the southeast, and the southwest were concentrated with a greater number of corridors, which were also more curved than in the plain regions; whereas the plains, like the center of Raoping County, the south and center of Chao'an District, and the center of Shantou, maintained fewer corridors with a more direct morphology. The key nodes were mainly distributed in the middle and the north of Raoping County, the border between Chao'an District and Raoping County, north of Chao'an District, the center of Shantou, the south of Chaonan District and the Lianjiang River valley. Using the corridors as the center, 600 m buffer zones were also created and can be seen in Figure 8.

Ecological Corridors, Key Nodes, and Buffer Zones
One hundred seventeen corridors were established by the linkagemapper based on the MCRM surface analysis, and the lengths of the corridors were between 30 and 22,245 m ( Figure 8). The mountainous regions of the north, the southeast, and the southwest were concentrated with a greater number of corridors, which were also more curved than in the plain regions; whereas the plains, like the center of Raoping County, the south and center of Chao'an District, and the center of Shantou, maintained fewer corridors with a more direct morphology. The key nodes were mainly distributed in the middle and the north of Raoping County, the border between Chao'an District and Raoping County, north of Chao'an District, the center of Shantou, the south of Chaonan District and the Lianjiang River valley. Using the corridors as the center, 600 m buffer zones were also created and can be seen in Figure 8.

Conflict and Fragile Zone Identification
After obtaining the buffer zones, we analyzed the spatial relationship between the ESPs and the buffer zones and found that the low-resilience, medium resilience, high-resistance and disturbing classes accounted for 6.16%, 35.20%, 48.22%, and 10.42%, respectively. The buffer zones in conflict with the low-class ESPs were mainly distributed in the north of Raoping County and connected to Na'nao County. The medium class conflict zones were mainly located in the north of Chao'an District, the northeast of Raoping County, the areas between Raoping and Na'nao Counties, and the middle of Chaoyang District.
To optimize future ecological land use, it is important to identify conflicts between the buffer zones and human interface; to this end, we extracted the increase in built-up land from 2020 to 2030 based on the simulated results of the FLUS model and analyzed the spatial fragility and conflict areas. The buffer zones with increasing built-up land were mainly distributed in the areas around the Lianjiang River valley in the Chaoyang and Chaonan Districts (Figure 9). The statistics indicated that the Chaoyang-Chaonan areas represented > 80% of the conflict zones in the whole study area (35.32% and 45.21%, respectively). For the remainder of the conflict zones, the Jinping, Xiangqiao, and Chenghai Districts maintained > 8.61%, 5.38%, and 3.75%, respectively. The Longhu District and Raoping County maintained 1.53% of the conflict zones of the whole study area. 89.97% of the increased built-up land were within the disturbed ESP zones, while 9.04% and 0.99% were within the high and medium ESP classes, respectively.

Conflict and Fragile Zone Identification
After obtaining the buffer zones, we analyzed the spatial relationship between the ESPs and the buffer zones and found that the low-resilience, medium resilience, highresistance and disturbing classes accounted for 6.16%, 35.20%, 48.22%, and 10.42%, respectively. The buffer zones in conflict with the low-class ESPs were mainly distributed in the north of Raoping County and connected to Na'nao County. The medium class conflict zones were mainly located in the north of Chao'an District, the northeast of Raoping County, the areas between Raoping and Na'nao Counties, and the middle of Chaoyang District.
To optimize future ecological land use, it is important to identify conflicts between the buffer zones and human interface; to this end, we extracted the increase in built-up land from 2020 to 2030 based on the simulated results of the FLUS model and analyzed the spatial fragility and conflict areas. The buffer zones with increasing built-up land were mainly distributed in the areas around the Lianjiang River valley in the Chaoyang and Chaonan Districts (Figure 9). The statistics indicated that the Chaoyang-Chaonan areas

Land Use Optimization in EGD
Conflict and fragile zones of the corridors, as well as built-up land (including 2020 measurements and 2030 simulated results), which encroached upon the low class of the ESP, should be reclaimed or remain in their original land-use type to best preserve ecological resources. There were 24.4 hm 2 of built-up land in the low class observed in 2020 scattered in the northern Chao'an District and central Raoping County. There were 369.70 hm 2 and 94.32 hm 2 of built-up land in the medium class of 2020 and 2030, respectively. The transitional nature of the medium class implies that they are also important to protect. Thus, the areas mentioned above should be the focus of land conservation plans in all future efforts.
The simulated results showed built-up land expansion into the water along the coastal South China Sea (Figure 8), with the greatest changes located on coastlines of Chaoyang, Haojiang, Longhu Districts, and Raoping County; however, in 2015, the MNR enacted a policy to ban reclamation from the sea to protect the pelagic and coastal ecology. Thus, the simulated results are not plausible, and all planned reclamation of the sea in the predicted territorial planning can be ignored if the present conditions are upheld. Sustainability 2021, 13, x FOR PEER REVIEW 19 of 25

EPS Optimization and Ecological Conservation
Environmental problems, such as water, soil, and air pollution, have had negative impacts on the EGD due to unsustainable developmental practices [106][107][108][109], and some of their negative impacts may be irreversible. In the background of the New Era, ecological civilization, and rapid expansion of the built-up land in EGD, it is crucial to protect the environment and preserve the ecological resources. The ESP, which was established according to the MCRM, demonstrated the spatial distribution of different ecological security levels. It provides basic regional protection for the necessary ecological areas and also guides the future built-up land expansion. The spatial management rules and the optimization in the study area should be categorized according to the resistance classes and the elements of the ESP (i.e., the corridors, the buffer zones or the areas around them).
The MCRM, which combined the natural conditions and anthropogenic factors, showed that the highest resistance areas are focused on the cluster of built-up lands. After establishing the ESP of the region, the low and medium resistance classes were identified that require prioritized protection and preservation. Low resistance areas, such as the north and middle of Chao'an District and Raoping County, the coastline around Na'nao County should be prioritized. This will present certain challenges based on the disturbances caused by human activities. We suggest that these areas should be strictly protected from any future land-use change or disturbance, and regulations should be set up to protect these areas. The present built-up land in the low resistance areas and the conflict zones should be identified in the consolidation plan. The medium resistance areas, especially We also analyzed the spatial relationship between the ecological sources and the built-up land. No simulated built-up land was found to occupy the ecological sources in 2030, while 129.01 hm 2 were recorded in 2020. Although the patch scales were relatively small (average area, 1.6 hm 2 ), they were scattered nearly throughout the entire study area: 44.83% located in the Chaonan District, primarily in the south and montane regions; 15.66% in Chaoyang District, which were also confined to the mountains; patches of the Chao'an District, Nan'ao, Raoping Counties, and Haojiang District were also in the mountainous areas, with proportions of 10.17%, 7.38%, 6.69%, and 4.81%, respectively; and in Chenghai and Longhu Districts, the patches occupied the protected drinking water reservoirs, comprising 3.16% and 5.19% of all patches observed. The results indicated that all of these patched areas should be reclaimed to their non-artificial surface conditions in future territorial plans to best preserve the region's ecological resources (Figure 8).

EGD Built-Up Land Expansion and Management
As shown in the previous section and from Figure 9, many areas in EGD underwent intensive built-up land expansion over the past decade. Many towns exceeded development intensity by 30%, some located in the city centers, but many others in rural areas. For example, some of the towns in Chao'an and Chaonan Districts were close to exceeding 60% development. By combining the development intensity of built-up land and the land-use map of EGD, we revealed three hotspots of built-up land: (1) the spread and sprawl belt of built-up land from the middle of the Chao'an District to the southeast towards Shantou's city center; (2) the intensive development along the Lianjiang River valley, located north of Chaonan and Chaoyang Districts, starting from the west of the Chaonan District and spreading east to the South China Sea coastline; and (3) starting from the south of Shantou Bay, and traveling along the coastline until reaching the northeast of Raoping County. These vibrant belts are the focus of EGD's manufacturing industry, most of which belong to small businesses or private companies. The primary industry of Chaozhou City and the surrounding suburbs is pottery. The south of Chao'an District, to the west of Shantou, is dominated by food industries, package printing, and machine manufacturing. The north of Shantou, like the Chenghai District, is a center of toy manufacture and assembly. To the south of Shantou, such as the Chaonan and Chaoyang Districts, the main industries are textiles and printing.
After analyzing the development intensity of the EGD over the past decade, we concluded that built-up land expansion was mainly concentrated in the suburban areas around the cities of Shantou and Chaozhou (Figure 9). It is widely acknowledged that the greater the intensity of built-up land expansion, the greater the correlated ecological and social problems [10,99,100]. Empirical studies have indicated that the scale of built-up land in these areas should be constrained [52,53,101,102] and restored to natural land cover types whenever possible for a more sustainable level of development.
The rapid expansion of built-up land allowed the economy to thrive in these areas over the past decade; however, it came at the sacrifice of the cropland area, as revealed by the land-use transition matrix. These plains areas made a significant contribution to the local agricultural industry and essential to China's policy of protecting croplands, especially those of high productivity. Thus, it is essential to prevent the further spread of built-up land encroaching on the high-quality croplands of the EGD.
Built-up lands also encroached upon on the South China Sea, and during 2010-2020, the conversion from sea to built-up land was > 1242 hm 2 , with an additional > 800 hm 2 of grassland converted from the sea as well. According to the master city plans of Chaozhou and Shantou, conversion from the sea was one of the major paths to increase built-up land. The conversion of the sea in Shantou occurred in three areas: Chaoyang-Chaonan, Haojiang, and the Longhu-Chenghai. For Chaozhou, sea conversion occurred exclusively in Raoping County. Reclamation in Chaoyang-Chaonan and Raoping County was mainly for the manufacturing industry, whereas conversion in Longhu, Haojiang, and Chenghai were for the construction of the new Shantou city center.
Studies showed that the sea reclamation had a negative impact on the coastline. First, it slows down the flow of the sea, and the self-purification capability of the sea would also go down. Second, the sea reclamation destroys the natural coastline, and it lowers the capability facing the tides and storm [103]. Historical, Shantou City lost over 500 hm 2 from 1950s to 1980s due to the sea reclamation, and it affected the biodiversity around the coast [104,105]. All planned reclamation projects from the sea before 2020 were carried out; however, the simulated results from 2030 showed that sea reclamation was still a hotspot for the increase of built-up land, despite the Chinese federal government, the MNR, and the Guangdong provincial government announced that reclamation from the sea is prohibited except in special circumstances. Moving forward, it will be nearly impossible for the EGD to reclaim any further area from the sea.
New policies regarding the renewal of old and low-efficiency neighborhoods will be necessary for future, sustainable spatial planning. It is recommended that the governments conduct full investigations to find these low-efficiency, dated, and built-up areas, which lack necessary facilities and are in need of increased residences. The specific policy for renewal and land consolidation should be formulated so as to meet the needs of the local residential demand. Old and inefficient areas usually are located within the city center, county, or towns, such as the southwest center of Shantou, south of Chao'an District, the center of Chaozhou, and the south of Raoping County.

EPS Optimization and Ecological Conservation
Environmental problems, such as water, soil, and air pollution, have had negative impacts on the EGD due to unsustainable developmental practices [106][107][108][109], and some of their negative impacts may be irreversible. In the background of the New Era, ecological civilization, and rapid expansion of the built-up land in EGD, it is crucial to protect the environment and preserve the ecological resources. The ESP, which was established according to the MCRM, demonstrated the spatial distribution of different ecological security levels. It provides basic regional protection for the necessary ecological areas and also guides the future built-up land expansion. The spatial management rules and the optimization in the study area should be categorized according to the resistance classes and the elements of the ESP (i.e., the corridors, the buffer zones or the areas around them).
The MCRM, which combined the natural conditions and anthropogenic factors, showed that the highest resistance areas are focused on the cluster of built-up lands. After establishing the ESP of the region, the low and medium resistance classes were identified that require prioritized protection and preservation. Low resistance areas, such as the north and middle of Chao'an District and Raoping County, the coastline around Na'nao County should be prioritized. This will present certain challenges based on the disturbances caused by human activities. We suggest that these areas should be strictly protected from any future land-use change or disturbance, and regulations should be set up to protect these areas. The present built-up land in the low resistance areas and the conflict zones should be identified in the consolidation plan. The medium resistance areas, especially the mountainous areas in Chaoyang District and Chaonan District, and the middle of Chao'an District should also be monitored by the local government. In addition, construction or potentially harmful activities should be constrained, as the preservation of these areas with low-resistance is vital. Negative impacts on high-resistance areas should also be regulated, although less strictly. Due to the critical importance of the corridors and buffer zones, they should be treated as low-resistance areas of the ESP.
The corridors play an important role in maintaining the landscape. In our study, the buffer zones were also identified. These two elements improve landscape connectivity, and landscape connectivity is a key factor in protecting regional ecosystem stability [110].
Conflict zones should be restricted from conversion to built-up land, thus preserving their original land-use type. This optimization can prevent anthropogenic interference and maintain landscape connectivity to promote ecological management. Local governments should strictly monitor and enforce any illegal encroachment into these areas because these areas are hotspots for built-up land expansion, and the likelihood of encroachment is greatest in these areas.
We should emphasize that the low-resistance class of the ESP, corridors, and the conflict zones should be considered as the ecological baselines in territorial planning. This does not mean, however, that these should be the only areas that are protected. All future built-up land expansion should be considered carefully to avoid negative ecological effects.
The built-up and low-resistance lands in 2020 should be categorized into the land consolidation and spatial territorial plans. Initially, the local governments should conduct ownership investigations and establish the ownership timeline of these areas, to distinguish any illegal claims under the current laws. Then, all applicable areas should be reclaimed to their pre-built-up status to preserve the ecological importance while fining the owners for illegal operations and encouraging the legal owners to move to less ecologically sensitive areas. Finally, the governments can also install specific reclamation plans according to the specific characteristics of these areas.

Study Limitations and Future Prospects
Our study used the MCRM to classify the ESP, combining natural and socioeconomic factors. Although we had eight factors included in the MCRM, due to the limitation of the data, additional environmental parameters, such as pollution and meteorological data, would likely increase the accuracy of the model. The restricted area of the FLUS model provides users an option to restrict land-use changes in certain areas with additional spatial data input. In this study, the ecological resources were used as this restriction, implying they cannot be changed to other land-use types. Permanent basic farmlands are included in the cropland area, and they are relatively high-quality cropland. These croplands are protected by law and should not be encroached on under any circumstances. The permanent basic farmlands should also be considered as restricted areas in the FLUS model. If we can obtain this data, the simulation of cropland dynamics would be more accurate.
In this study, the quantity of built-up land in 2030 was predicted by the Markov chain model based on the land-use dynamics pattern in the past. However, the predicted result neglected other factors, such as the built-up land per person in the region, which was an important index in territorial spatial planning, the spatial unbalance of development and so on. In the future, further studies should focus on multiobjective decision prediction in the quantity of land-use types, which can be applied in spatial territorial planning.

Conclusions
Using the land-use land cover products from GlobeLand30, we analyzed the land-use dynamics of the EGD, including the cities of Chaozhou and Shantou, from 2000 to 2020.
The conversion matrix showed that from 2000 to 2010, the quantity of cropland and built-up land was stable, whereas from 2010 to 2020, cropland dropped dramatically, and built-up land increased by 5.85%. The coastlines were also encroached upon by built-up land during the same period. We simulated land use in 2030 using a FLUS, and based on an MCRM, revealed the following distributions of ESPs in EGD: the disturbed class, 21.13%; low resistance class, 10.67%; middle-resistance class, 28.82%; and high-resistance class, 39.37%. The ecological resource locations were obtained from the master city plans for both urban areas, and a least-cost path to establishing the corridors and the buffer zones was proposed to improve landscape connectivity. By comparing the increased built-up land of 2030, the corridors, and the buffer zones, we have outlined an optimization plan by focusing on the hotspots for future built-up land expansion for the EGD; this can be applied in the spatial territorial planning. Additional measures and policies, such as preservation of the ecological sources, protection of cropland and coastlines, and consolidation of built-up land with low-resistance, were proposed for this and other similar areas in China with comparable development levels.