Landscape ecological risk assessment in the Dongjiangyuan region, China, from 1985 to 2020 using geospatial techniques

Abstract

Anthropogenic activities can greatly affect the ecological environment. As an ecological protection area, it is necessary to scientifically evaluate the landscape ecological risk (LER) in the Dongjiangyuan region to provide scientific guidance for regional sustainable development. In this study, the LER was calculated, and the spatial and temporal characteristics of the LER from 1985 to 2020 were analysed using geospatial techniques. The results show that the proportion of low-risk and extremely low-risk areas increased from 87.65% to 94.26% during the 1985-2020 period. The extremely high-risk and high-risk areas had a decreasing trend, and the extremely high-risk areas were concentrated in areas with impervious surfaces and croplands. The rate of risk was negative, especially in south-eastern Xunwu County, southern Dingnan County and central Anyuan County, indicating that the ecological risk has been greatly improved. The LER centre gradually migrated to the geometric centre of the study area. However, the gravity centre of extremely high risk and high risk remained in Xunwu County. The spatial agglomeration of LER changed significantly, and the overall difference between cold areas and hot areas decreased. The hot spot areas in the Dongjiangyuan region are key areas for ecological governance in the future.

Keywords:

• Landscape ecological risk
• gravity modelling
• rate of risk change
• hot and cold spot
• Dongjiangyuan region

1. Introduction

Population growth and economic development have promoted an increase in human demand for land use and the exploitation of natural resources (Gong et al. 2012; Wang et al. 2018; Danish et al. 2020). Global land use and land cover (LULC) have undergone tremendous changes, and the ecological environment has also correspondingly encountered drastic changes, such as forest degradation (van Vliet 2019; Karki et al. 2018). The global urban land area increased from 33.2 to 71.3 million hectares between 1992 and 2015, resulting in a direct loss of 3.3 million hectares of forest (van Vliet 2019). The LULC change caused by excessive anthropogenic activities will also make the regional ecosystem more vulnerable (Fang et al. 2022). Ecological degradation and environmental destruction have seriously affected human well-being, particularly in regards to the high quality of life and the sustainable development of human society (Chopra 2016). Therefore, exploring the stability of ecosystem structure and sustainable development has become the top priority, and research on the quality and security of the ecological environment has become a global hot topic (Fu et al. 2020; Moarrab et al. 2022).

Ecological risk refers to the threats and risks borne by the ecosystem and its components under external pressures (e.g. anthropogenic activities), which have a negative impact on the structure and function of an ecosystem (Liu et al. 2018; Zhang et al. 2020). Generally, ecological risk assessment is used to evaluate the possibility and degree of ecosystem damage from the adverse impacts of natural and anthropogenic activities on regional ecosystems (Li et al. 2020; Zhang et al. 2020). The US Environmental Protection Agency (USEPA) first proposed the framework for ecological risk assessment in 1992 (USEPA 1992), and this framework served as the foundation for the EPA's 1998 ecological risk assessment (USEPA 1998). Subsequently, some scholars have conducted regional ecological risk assessments based on the USEPA framework (Cormier et al. 2000; Walker et al. 2001; Malekmohammadi and Blouchi 2014). Ecological risk assessment is an effective way to support ecosystem management and can provide theoretical support for environmental decision-making (Chen et al. 2013).

As an important part of ecological risk assessment, landscape ecological risk (LER) assessment has enriched and expanded ecological risk research and has played an important role in ecological assessment and environmental management (Peng et al. 2015; Tian et al. 2019). Specifically, LER assessment refers to the analysis of the response of landscape element mosaics, landscape pattern evolution and landscape ecological processes to internal risk sources and external disturbances (Peng et al. 2015). LER assessment is different from traditional ecological risk assessment methods. It relies on the perspective of coupling between ecological processes and spatial patterns in landscape ecology and particularly emphasizes the spatiotemporal heterogeneity and scale effects of ecological risks (Qiao et al. 2021; Han et al. 2022). Several geospatial techniques have been extensively used in the LER research (Li et al. 2018; Xie et al. 2021). Therefore, many scholars have conducted LER assessments based on landscape patterns to explore the spatiotemporal heterogeneity of ecological risk (Liu et al. 2018; Xie et al. 2021; Ran et al. 2022). Based on the theory of landscape pattern-process association, the LER assessment can directly reflect the ecological risks in the structure and composition of landscape patterns (Fu 2014; Zhang et al. 2020).

The Dongjiangyuan region is the source area of the Dongjiang River and is an important ecological function protection area in Jiangxi Province. The ecological security of the Dongjiangyuan region is not only related to the ecological environment of Jiangxi Province but also has a direct impact on the drinking water security and sustainable development of Guangdong and Hong Kong (Zhou et al. 2012). Currently, ecological issues such as forest degradation, and soil erosion have become increasingly prominent in the Dongjiangyuan region, and these issues have vital roles in water recycling and biodiversity (Hu et al. 2008; Lei et al. 2019; Wang et al. 2022). Thus, it is urgent to further protect, restore and promote the sustainable use of terrestrial ecosystems and biodiversity and sustainable forest management to achieve regional sustainable development in the context of urbanization process. The application of geospatial techniques in LER can improve our understanding of the spatial and temporal distribution of LER dynamics in the Dongjiangyuan region.

Hence, LER assessment was carried out in this study based on the LULC change to explore the threat of anthropogenic activities to the regional ecosystem and provide decision-making for ecological governance and regional sustainable development in Dongjiangyuan. We conducted this study to explore the spatial and temporal patterns of the LER index of the Dongjiangyuan region from 1985 to 2020; furthermore, the ratios of risk change, gravity centre transfer, and hot and cold spots of the LER index were analysed. In addition, the spatial and temporal characteristics of LER in the Dongjiangyuan region and implications for sustainable regional development are discussed.

2. Materials and methods

2.1. Study area

The Dongjiangyuan region is located in southern Ganzhou city, Jiangxi Province, China; the region includes Xunwu, Anyuan and Dingnan Counties and neighbours Guangdong Province in the south and Fujian Province in the east. The geographical distribution of this region extends from 24°32'26′' N to 25°37'36′' N latitude and from 114°47'58′' E to 115°55'28′' E longitude (Figure 1). The GDP was 28.187 billion Yuan, and the total population was 975,000 as of 2020. The region is dominated by mountains and hills, with an altitude of 157-1491 m. The terrain is high in the central and south-eastern parts of the study area. Forestland is the most important LULC type in the Dongjiangyuan region, with a forest coverage rate of 83.49% (Xunwu Yearbook 2021; Dingnan Yearbook 2021; Anyuan Yearbook 2021).

Figure 1. Location of the Dongjiangyuan region, China.

The climate in the region is mild, and the annual precipitation in 2019 was 1630.77 mm (Xunwu Yearbook 2020; Anyuan Yearbook 2020; Dingnan Yearbook 2020). The rainfall is unevenly distributed, mainly from April to July, and the region is prone to soil erosion. In particular, red soil is widely distributed in the Dongjiangyuan region, and it is strongly acidic and not suitable for the growth of crops. The Dongjiangyuan region has rich mineral resources, and the development of mineral resources has played a role in promoting the economic development of the Dongjiangyuan region; however, it has also brought serious ecological and environmental problems (e.g. vegetation degradation) (Hu et al. 2008).

2.2. Data sources

The LULC data from 1985 to 2020 used in this study came from the 30-m annual China Land Cover Dataset (CLCD) derived from Landsat on Google Earth Engine (GEE). We retrieved and downloaded the dataset from the CLCD (http://irsip.whu.edu.cn/resources/resources_v2.php).

Compared with the existing annual land cover products (e.g. ESACCI_LC), the CLCD has a higher spatial resolution and a longer time coverage (Yang and Huang 2021). The overall accuracy of the CLCD reached 79.31%, which exceeded that of MCD12QI, ESACCI_LC, FROM_GLC and GlobeLand30 (Yang and Huang 2021). In addition, the CLCD products have been intercompared with several 30-m thematic products, and they have exhibited good consistencies with Global Forest Change, Global Surface Water, and three impervious surface products (i.e. GISA, GAUD, GAIA) (Yang and Huang 2021). These results indicate that the accuracy and reliability of the CLCD were higher than those of other Landsat-derived thematic datasets.

2.3. Methods

To explore the spatiotemporal characteristics of LER in the Dongjiangyuan region, the flowchart of LER research is shown in Figure 2. First, risk zones were divided by the Fishnet tool of ArcGIS 10.2 software. Second, the LER index of each risk zone was calculated by the landscape disturbance index, landscape vulnerability index and landscape loss degree index. Finally, the rate of risk change (RRC), centre of gravity transfer and spatial autocorrelation were used to reflect the spatial and temporal dynamics and characteristics of LER in the Dongjiangyuan region during the 1985-2020 period.

Figure 2. The framework of landscape ecological risk research.

2.3.1. LER index

This research built an LER index based on the landscape structure of the regional ecosystem. (1) The landscape disturbance index (LDI) and (2) the landscape vulnerability index (LVI) were used to build the LER index. The Fishnet tool in ArcGIS 10.2 software was used to divide the study area into risk zones, and the LER value of each risk zone was calculated based on Fragstats 4.2 and ArcGIS 10.2 software. The calculation equation of the LER index (Chen and Pan 2003; Shi et al. 2015; Ju et al. 2021) is as follows: (1) $$L{\mathit{ERI}}_k=\sum _{i=1}^n\frac{A_{ki}}{A_k}{L}_i$$(1) where ${\mathit{LERI}}_k$ refers to the LER index of the $k$ th risk zone, $A_{ki}$ refers to the area of the ith landscape type of the $k$ th risk zone, $A_k$ is the area of the $k$ th risk zone, $L_i$ refers to the landscape loss degree index of the ith landscape type, and n refers to the number of different landscape types. The landscape loss degree index refers to the loss degree of natural properties of ecosystems represented by different landscape types under the interference of natural and anthropogenic factors (Li et al. 2017). The calculation equation of the landscape loss degree index (Xu et al. 2001; Liu et al. 2012; Ju et al. 2021) is as follows: (2) $$L_i=D_i\times V_i$$(2) where $L_i$ is the landscape loss degree index of the ith landscape type; $D_i$ is the LDI of landscape type i; and $V_i$ is the LVI of landscape type i.

Since there is no unified evaluation standard for the LER index, many previous studies have used the natural breakpoint method to classify the levels of the LER index (Zhu et al. 2022; Zhang et al. 2020; Ji et al. 2021). Hence, the LER index in 1985 was divided into five grades by using the natural breakpoint method. The specific division is shown in Table 1.

Table 1. Classification of landscape ecological risk index.

Value of landscape ecological risk index	Grade
LERI ≤ 0.07	Extremely low risk
0.07<LERI ≤ 0.16	Low risk
0.16<LERI ≤ 0.31	Moderate risk
0.31<LERI ≤ 0.55	High risk
LERI > 0.55	Extremely high risk

(1) Landscape disturbance index

The LDI is used to reflect the degree of external disturbance to the ecosystem represented by different landscapes (Li et al. 2017; Ji et al. 2021). The equation of the LDI (Chen and Pan 2003; Shi et al. 2015; Ji et al. 2021) is as follows: (3) $$D_i=a{LF}_i+b{LS}_i+c{LD}_i$$(3) where $D_i$ is the disturbance degree of the ith landscape type; ${LF}_i$ is the fragmentation degree of the ith landscape type; ${LS}_i$ is the separation degree of the ith landscape type; ${LD}_i$ is the dominance degree of the ith landscape type; and $a, b \mathrm{and} c$ are the weights of fragmentation, separation and dominance, respectively. Based on previous research results (Ji et al. 2021; Ying et al. 2022), the weights of fragmentation, separation and dominance in this paper were set to 0.5, 0.3 and 0.2, respectively.

(a) Landscape fragmentation degree

The landscape fragmentation degree indicates the fragmentation degree of the landscape, which can be used to reflect the complexity of the landscape spatial structure and can also reflect the degree of anthropogenic interference with the landscape to a certain extent (Zhang et al. 2020; Zhu et al. 2022). The landscape fragmentation degree equation (Li and Huang 2015; Ji et al. 2021; Zhu et al. 2022) is as follows: (4) $${LF}_i=\frac{n_i}{A_i}$$(4) where $n_i$ is the number of patches of landscape type i, and $A_i$ is the total area of landscape type i.

(b) Landscape separation degree

The landscape separation degree refers to the dispersion degree of the patch distribution of the same landscape type. The greater the separation degree is, the more dispersed the landscape type is in the region, and the worse its stability is (Li et al. 2017; Ran et al. 2022). The landscape separation degree equations (Han et al. 2010; Zhu et al. 2022) are as follows: (5) $${LS}_i=\frac{\sqrt{{DI}_i}}{2{RC}_i}$$(5) (6) $${DI}_i=\frac{n_i}{A}$$(6) (7) $${RC}_i=\frac{A_i}{A}$$(7) where ${DI}_i$ is the distance index of landscape type i; ${RC}_{i }$ is the relative coverage of landscape type i; $n$ is the total number of patches; and $A$ is the total area of the landscape.

(c) Landscape dominance degree

The landscape dominance degree refers to the degree of dominance of certain landscape types in the whole landscape pattern (Li et al. 2017; Ju et al. 2021). The higher the value of the landscape dominance degree is, the greater the influence of the landscape type on the whole landscape pattern (Zhu et al. 2022). The landscape dominance degree equations (Han et al. 2010; Ying et al. 2022) are as follows: (8) $${LD}_i=d{RD}_i+e{RC}_i$$(8) (9) $${RD}_i=\frac{n_i}{n}$$(9) where ${RD}_i$ is the relative density of landscape type i, ${RC}_{i }$ is the relative coverage of landscape type i, and d and e are the weights of ${RD}_i$ and ${RC}_i,$ respectively. Based on previous research results (Han et al. 2010; Ying et al. 2022), the weights of relative density and relative coverage in this paper were set to 0.4 and 0.6, respectively.

(2) Landscape vulnerability index

The landscape vulnerability index refers to the sensitivity of artificial or natural disturbance responses represented by different landscapes and can reflect the resistance of different landscape types to external interference (Ji et al. 2021; Ran et al. 2022). The higher the landscape vulnerability index is, the weaker the ability of the landscape type to resist external interference (Zhu et al. 2022). According to previous research results (Liu et al. 2012; Zhang et al. 2020; Ji et al. 2021; Zhu et al. 2022), the vulnerability of landscape types was divided into seven levels, from high to low, barren land > water > cropland > grassland > shrubland > forest > impervious surface, and then all the vulnerability values of all landscape types were normalized for further analysis.

2.3.2. RRC

The RRC was established to compare the difference in the rate of LERI change between units in different periods (Zhong et al. 2020). This index is the average annual increase in the LERI of a unit as a percentage of the initial LERI value, which is used to identify the spatiotemporal heterogeneity of risk change (Ran et al. 2022). The positive and negative RRC indexes indicate risk growth and decreased risk, respectively. Larger and smaller RRC indexes indicate severe risk change (Ran et al. 2022). The calculation of RRC is as follows: (10) $${\mathit{RRC}}_k=\left({\mathit{LERI}}_k^{t2}-{\mathit{LERI}}_k^{t1}\right)/{\mathit{LERI}}_k^{t1}\times \frac{1}{\Delta t}\times 100\mathrm{\%}$$(10) where ${\mathit{RRC}}_k$ represents the RRC of the $k$ th risk zone. ${\mathit{LERI}}_k^{t1}$ and ${\mathit{LERI}}_k^{t2}$ indicate the LER index of the kth risk zone at times $t1$ and $t2,$ respectively; $\Delta t$ represents the time span from $t1$ to $t2.$

2.3.3. Centre of gravity transfer

Gravity modelling can identify movement direction and distance to the gravity centre, which can reflect changes in the quantity and change trend of the targeted object over time (He et al. 2011). In this study, the centre of gravity transfer model was used to analyse the spatial and temporal changes in the LER index more visually and concretely and to summarize the spatiotemporal characteristics of the LER. The migration trajectory of the gravity centre in this paper can directly reflect the spatial pattern change of the LER index in the Dongjiangyuan region, and the migration distance and migration directions were then analysed. The equations for the centre of gravity are as follows: (11) $$X_t=\frac{\sum _{k=1}^n{x}_{tk}LER{I}_{tk}}{\sum _{k=1}^n\mathit{LER}{I}_{tk}}$$(11) (12) $$Y_t=\frac{\sum _{k=1}^n{y}_{tk}LER{I}_{tk}}{\sum _{k=1}^n\mathit{LER}{I}_{tk}}$$(12) where X_t and Y_t are the x and y gravity centre coordinates for LERI at time t. $x_{tk}$ and $y_{tk}$ represent the x and y gravity centre coordinates of the kth risk zone at time t, respectively. ${\mathit{LERI}}_{tk}$ indicates the LER index of the kth risk zone at time t.

The calculation of the movement distance for the centre of gravity is as follows: (13) $$D_{t1-t2}=C\times {\left[{\left(Y_{t1}-Y_{t2}\right)}^2+{\left(X_{t1}-X_{t2}\right)}^2\right]}^{\frac{1}{2}}$$(13) where $D_{t1-t2}$ refers to the distance of gravity centre movement from year t1 to t2. $X_{t1},$ $X_{t2},$ $Y_{t1},$ and $Y_{t2}$ refer to the gravity centre coordinates of the LERI for years t1 and t2. $C$ is a constant, which is the coefficient of converting geographical coordinate units into plane distance.

2.3.4. Spatial autocorrelation analysis

Spatial autocorrelation shows whether features with high values or features with low values tend towards clustering (Le et al. 2022). When a feature with a high value is surrounded by others with high values, it can be referred to as a hot spot (Le et al. 2022). In this study, we conducted optimized hot spot analysis using ArcGIS 10.2 to explore the spatiotemporal characteristics of clustering and differentiation. This approach can identify the spatial clustering of high and low LERI values. The equation of $G_i^{*}$ is as follows: (14) $$G_i^{*}=\frac{\sum _{j=1}^n{w}_{i,j}{x}_j-\overline{x}\sum _{j=1}^n{w}_{i,j}}{S\sqrt{\frac{\left[n\sum _{j=1}^n{w}_{i,j}^2-{\left(\sum _{j=1}^n{w}_{i,j}\right)}^2\right]}{n-1}}}$$(14) (15) $$\overline{x}=\frac{\sum _{j=1}^n{x}_j}{n}$$(15) (16) $$S=\sqrt{\frac{\sum _{j=1}^n{x}_j^2}{n}-{\overline{x}}^2}$$(16) where $x_j$ is the attribute value of element $j,$ $w_{i,j}$ is the spatial weight between element $i$ and element $j,$ n is the total number of elements, $\overline{x}$ is the mean, and S is the standard deviation.

3. Results

3.1. Spatial analysis of LULC change

The LULC classification results of the Dongjiangyuan region from 1985 to 2020 are shown in Figure 3. Cropland and impervious surfaces were distributed in low-altitude areas, while forest was distributed in high-altitude areas. Forest was the major LULC type distributed in this region and showed a decreasing fluctuation trend. Cropland was mainly distributed in the area surrounding the county seat and showed a fluctuating growth trend. The impervious surfaces were mainly distributed in Xinshan and Tianxin towns of Anyuan County, Lishi town of Dingnan County, and Changning and Liuche towns of Xunwu County. The impervious surface of the Dongjiangyuan region gradually increased over the study period, and the impervious surface expanded by 32.58%, 23.24% and 30.06% in 2005-2010, 2010-2015 and 2015-2020, respectively.

Figure 3. Land use and land cover of the Dongjiang region from 1985 to 2020.

The LULC transfer matrix from 1985 to 2020 is shown in Table 2. This result shows that cropland, water, barren land and impervious surfaces increased, while forest, shrubland and grassland decreased during the 1985-2020 period. The impervious surface area increased from 13.75 km² in 1985 to 47.52 km² in 2020, an increase of 245.70%. The increase in impervious surfaces mainly resulted from the transformation of cropland and forest, and the area of cropland transformed to impervious surfaces between 1985 and 2020 was 64.88% of the total impervious surface in 2020. The cropland area increased from 561.89 km² in 1985 to 727.52 km² in 2020, and 329.69 km² of land was transformed from forest to cropland during the study period, which had a high contribution to the increase in cropland. It is worth mentioning that forest decreased by 182.00 km² from 1985 to 2020, which was the largest decrease among all the transformations, and the decrease in forest prominently changed to cropland and impervious surfaces.

Table 2. Land use transfer matrix from 1985 to 2020 (unit: square kilometre).

Year		1985
	Land use type	Cropland	Forest	Shrubland	Grassland	Water bodies	Barren land	Impervious surface	Total
2020	Cropland	382.70	329.69	0.24	9.90	1.36	0.02	3.61	727.52
	Forest	143.15	5064.78	3.09	12.44	2.67	0.00	1.12	5227.25
	Shrubland	0.04	1.80	0.69	0.15	0.00	0.00	0.00	2.68
	Grassland	1.85	3.31	0.19	0.87	0.02	0.00	0.02	6.26
	Water bodies	2.96	1.88	0.00	0.38	3.70	0.00	0.87	9.79
	Barren land	0.35	0.17	0.00	0.09	0.00	0.01	0.00	0.63
	Impervious surface	30.83	7.61	0.00	0.82	0.13	0.00	8.13	47.52
	Total	561.89	5409.25	4.21	24.63	7.88	0.03	13.75	6021.65

3.2. LER analysis

The LER index in the Dongjiangyuan region is presented in Figure 4. During the study period, the extremely high-risk areas were concentrated mainly in the south-eastern Dongjiangyuan region. A small number of extremely high-risk areas were distributed in western Dingnan County, but since 2010, western Dingnan County has transitioned from extremely high risk to high risk. The high-risk area is surrounded by the extremely high-risk area, and the high-risk areas are scattered in the northeast and south of Anyuan County. From 1985 to 2020, the proportion of extremely high-risk areas decreased from 1.03% to 0.20%, while the proportion of low-risk and extremely low-risk areas increased from 87.65% to 94.26%. It is worth noting that, except in 2010, the low-risk and extremely low-risk areas gradually increased, and the high-risk and extremely high-risk areas gradually decreased. From 1985 to 2020, Tianxin town of Anyuan County gradually changed from high risk to moderate risk. The town of Liuche in Xunwu County also gradually changed from extremely high risk to high risk and moderate risk; however, the town of Changning in Xunwu County had a worsening trend of the LER. Lishi town of Dingnan County declined from extremely high risk to high risk. Overall, the LER index of impervious surfaces gradually declined during the study period.

Figure 4. Classification of the landscape ecological risk index of the Dongjiangyuan region from 1985 to 2020.

3.3. Spatiotemporal pattern of the rate of risk change

The spatiotemporal pattern of the RRC in the Dongjiangyuan region from 1985 to 2020 is shown in Figure 5. The RRC showed a negative value in a large area from 1985 to 2020 (Figure 5(h)), indicating that the LER index decreased in these areas, especially in south-eastern Xunwu County, southern Dingnan County and central Anyuan County. Compared with the classification of the LER index (Figure 4), the negative value area of the RRC coincided with the high-risk and extremely high-risk areas. Furthermore, the spatial pattern of the RRC shifted from a large-scale decline (especially in the 1985-1995 period) to a small-scale decline, and the RRC was positive and sporadically distributed in the study area. From 2000 to 2005, Figure 5(d) shows that the RRC increased significantly, mainly in Kongtian town of Anyuan County, Chengjiang town of Xunwu County and Lingbei town of Dingnan County. Although the areas with a negative RRC were much larger than the areas with a positive risk change rate, it is worth noting that the extent of the positive RRC showed an increasing trend across the study periods.

Figure 5. Rate of risk change in the Dongjiangyuan region from 1985 to 2020.

Note: the red color indicates the rate of risk change beyond than 0, and the blue color indicates the rate of risk change less than 0.

3.4. Analysis of the centre of gravity transfer

To explore the spatial shift trend of the LER index in the Dongjiangyuan region, the centres of gravity of the LER index and each risk grade from 1985 to 2020 were calculated, the centre of gravity transfer map was drawn as shown in Figure 6, and the centre of gravity transfer distance and direction were calculated, as shown in Tables S1–S6. The centre of gravity of the LER index from 1985 to 2020 was always in Anyuan County, and the extent of the gravity centre was between 115°26'15′'- 115°26'48′'E and 24°58'41′' − 24°59'04′' N, which is located in the southeast of the geometric centre of the Dongjiangyuan region (115°25'25′' E, 25°9'14′' N), indicating that the LER index of Xunwu County was higher than that of the other two counties.

As shown in Figure 6(a) and Table S1, the gravity centre of the LER shifted to the northwest by 0.63 km, indicating that the ecological situation of Xunwu County improved, and the gravity centre of the LER was close to the geometric centre of the Dongjiangyuan region. The gravity centre of the LER moved longitudinally by 1.02 km during the 2005-2015 period. The gravity centre of LER moved to the northeast in the 1985-1990, 2000-2005 and 2015-2020 periods because the reduction in LER in Dingnan County and Xunwu County was greater than that in Anyuan County. However, the gravity centre of the LER index moved westwards for a long distance, which was caused by the increase in the LER index in Dingnan County from 2005 to 2015. The moving distance of the gravity centre from 2010 to 2015 was the longest, moving 0.71 km to the northwest, indicating that the LER index of Xunwu County was significantly reduced during this period.

Figure 6. The gravity centres of the landscape ecological risk index and each risk grade from 1985 to 2020.

The gravity centre of extremely high risk was consistently located in Xunwu County, moving 17.21 km to the northeast. During 2000-2020, the gravity centre of extremely high risk gradually moved to the northeast. The gravity centre of high risk was also located in Xunwu County, and it moved to the southwest overall. This movement can be divided into two stages: from 1985 to 2000, the gravity centre moved to the south, and from 2000 to 2020, the gravity centre moved to the west. The gravity centre of high risk moved the longest distance among all risk grades, at 19.31 km. The gravity centre of moderate risk was located in Anyuan County at the beginning and ultimately moved to Xunwu County. The overall direction of this transfer was 10.11 km to the southeast. The gravity centres of both low risk and extremely low risk were in Anyuan County. Their migration directions were northeast and southwest, respectively, with a small transfer distance.

3.5. Spatial clustering of the LER index

The hot and cold spot analysis of the LER index in the Dongjiangyuan region is shown in Figure 7. The hot spot areas were mainly distributed in south-eastern Xunwu County, north-eastern and southern Anyuan County, western Dingnan County, and the areas where cropland land and impervious surfaces were concentrated. The cold spot areas were mainly concentrated in eastern Anyuan County, western Xunwu County, and the areas where forestland was located. The hot spot area in north-eastern Anyuan County has been shrinking continuously, and its accumulation changed from Tianxin town and Chongshi town to the north-western part of Tianxin town. The hot spots of Lishi town in Dingnan County became gradually concentrated between 1985 and 2010 and spread to the outskirts from 2010 to 2020. The cold spot area largely showed a decreasing trend across the study period.

Figure 7. Spatial distribution of hot and cold spots of the landscape ecological risk index in the Dongjiangyuan region from 1985 to 2020.

4. Discussion

4.1. Spatiotemporal characteristic of the LER

In general, the LER index of the Dongjiangyuan region is relatively low, approximately 90% of the areas are at a low risk or an extremely low risk level, and the ecological environment is excellent in these regions, especially in the high-altitude area (e.g. Sanbaishan Mountain). In high-altitude areas, anthropogenic activities (e.g. production and construction) are restricted by landforms and climatic conditions, resulting in low anthropogenic disturbance and low LER (Teng et al. 2020). The areas with high and extremely high risk levels, which were located in the southeast, southwest and north of the Dongjiangyuan region because of intensified anthropogenic activities (Zhou et al. 2021), mainly included low-altitude areas. These low-latitude areas are normally covered with a concentrated population and urbanization, and the pressure effect of urbanization on the ecological environment has caused environmental pollution, resource destruction and ecosystem degradation (Dewan and Yamaguchi 2009; Mohan et al. 2011; Fang et al. 2016).

The change in the LER index in the Dongjiangyuan region was divided into three stages: 1985-2005, 2005-2010 and 2010-2020. The percentage of these high and extremely high risk levels decreased from 3.5% to 1.4% between 1985 and 2020. However, those in the 2005-2010 period increased from 1.6% to 2.1%. From 1985 to 2005, the high-risk and extremely high-risk areas in the Dongjiangyuan region gradually decreased, indicating that the Dongjiangyuan region strengthened its environmental protection and controlled the trend of ecological environment deterioration. The proportion of extremely high-risk areas decreased during the 1985-2020 period, which can be observed especially in Xunwu County. Some areas have been converted from cropland to orchard (e.g. orange plantation) in Xunwu County, meaning that the LULC type has varied from low ecosystem service value to high ecosystem service value (Lei et al. 2019), and the conversion between cropland and orchard has greatly improved the natural environment.

The gravity centre of the LER in the Dongjiangyuan region is close to its geometric centre. The centre of gravity transfer was divided into two stages: the centre of gravity transferred to the right from 1985 to 2005 and the centre of gravity transferred to the left from 2005 to 2020. From 1985 to 2005, the gravity centre of LER transferred to the right, mainly due to the significant reduction in LER in Anyuan County and Dingnan County. The gravity centre of extremely low risk obviously transferred to the left. From 2005 to 2020, the gap between Xunwu County and the other two counties narrowed. The gravity centre of high risk moved to the left and the gravity centre of extremely high risk moved to the north over a long distance. Xunwu County has achieved remarkable results in returning farmland to forests, and Jiangxi and Guangdong Provinces have signed agreements on horizontal ecological compensation for the upper and lower reaches of the Dongjiang region to increase ecological investment (Wang et al. 2022). The large ecological source patches have been fragmented in Anyuan and Dingnan Counties because of urbanization development. Hence, the centre of gravity transferred to the left in the Dongjiangyuan region during this period.

4.2. Implications for regional sustainable development

The spatial clustering of LER reflected that the overall difference between cold and hot spot areas decreased, especially in the 2010-2020 period. Chen et al. (2022) also showed that healthy development in the Dongjiangyuan region occurred in the 2010-2018 period. In recent years, the ecological environment in areas with high LER values has been improved through a series of joint efforts (such as construction of the ecological forest) by the government and the public. The Dongjiangyuan region has rich forest resources, which causes this area to have a low spatial ecological risk. Therefore, forest management can be further strengthened so that the high LERI areas can be improved and advanced in the future. The regions with positive RRC values were mainly distributed in the nonsignificant regions, which can be easily neglected in sustainable management. In addition, the conversion from cropland to orchard has been a major way to improve the natural environment in the region.

Zhou and Fan (2011) found that the degree of ecological degradation in Dongjiangyuan from 1985 to 2005 was −3.90, which also confirmed that the ecological environment had improved. From 2005 to 2010, the high- and extremely high-risk areas increased, and this period was in the early stage of urbanization. Urban land has largely expanded, and the corresponding basic infrastructure has been developed, which has caused the deterioration of the quality of the ecological environment, and the remote sensing-based ecological index decreased from 0.538 in 2004 to 0.332 in 2009 (Zhou et al. 2021). From 2010 to 2020, the high- and extremely high-risk areas gradually decreased. Anyuan County and Dingnan County had no extremely high-risk areas, and the extremely high-risk areas in Xunwu County decreased significantly. Xunwu County will still be the major area for ecological risk governance in the near future.

Overall, the gravity centres of extremely high risk and high risk were still located in Xunwu County. Xunwu County still needs to maintain the governance results from recent years, continue to regulate land use and reduce the LER. The gravity centre of extremely high risk gradually approached the county seat (i.e. Changning Town) of Xuanwu County. Hence, Xunwu County should take measures to implement ecological governance in its county seat later. Although the LER of Anyuan County and Dingnan County was lower than that of Xunwu County, their reduced change rate, i.e. the RRC, was lower than that of Xunwu County, so the implementation of policies can be strengthened to improve the ecological environment. Some areas in Xunwu County require long-term ecological management and restortation.

Water source protection and setting apart hills, including sand areas for tree growth, have been conducted to alleviate the ecological pressure on the Dongjiangyuan region (Zhou et al. 2021; Wang et al. 2022). By the end of 2018, this region had divided 32300 hectares of set-apart hills, including sand areas, for tree growth (Ganzhou Statistical Bureau 2019). These ecological protection policies have been effectively implemented to improve the natural ecosystem and recover the earth’s environment. A good natural environment in the region will help to further improve the water quality to better realize ecological compensation and then realize a ‘win-win’ outcome between the ecological environment and economic development.

4.3. Limitations of the study

In this study, we assigned the weight of calculating the LERI based on previous studies (Han et al. 2010; Zhu et al. 2022; Ji et al. 2021), which mainly depends on subjective expert experience. Thus, objective methods (e.g. principal component analysis) could be considered to allocate the weights for the calculation process in the future. In addition, the scale effect is inevitable in geographical research, and this study did not explore the spatiotemporal differentiation of the LER index at different scales. Therefore, we can further set different scales for selecting a suitable evaluation unit to optimize the results of the LER index in the future.

5. Conclusion

This study mapped and analysed the LER pattern of the Dongjiangyuan region using LULC data from 1985 to 2020. During the study period, the area of forest decreased the most, and the area of cropland increased the most, followed by the area of impervious surfaces. The LER index in the Dongjiangyuan region changed significantly from 1985 to 2020, and the areas of extremely high risk and high risk had a decreasing trend. The area of extremely high risk was concentrated in impervious surfaces and cropland, i.e. areas with high anthropogenic activities. The LER greatly improved in Xunwu County, which may have resulted from the conversion from cropland to orchard. Thus, the conversion from cropland to orchard can be considered in future management.

The RRC of the Dongjiangyuan region was negative in a large area, and the areas with a large decrease in the risk change rate were concentrated in the eastern part of Xunwu County and part of the north-eastern of Anyuan County, but there were still some areas with a high increase in the risk change rate. Xunwu County will still be the major area for ecological risk governance in the near future. The LER centre of the Dongjiangyuan region gradually migrated to the geometric centre. The gap between the LER of Xunwu County and the other two counties gradually narrowed. The gravity centre of extremely high risk gradually moved to the county seat of Xunwu County. The spatial agglomeration of LER in the Dongjiangyuan region changed significantly, and the overall difference between cold and hot areas decreased. It is worth emphasizing that the hot spot areas will still be key areas for ecological governance in the future.

Disclosure statement

No potential conflict of interest was reported by the authors.

Data availability statement

The datasets generated during and/or analysed during the current study are available from the corresponding author on reasonable request.

Additional information

Funding

This study was supported by the Science and Technology Project of Jiangxi Education Department (No. GJJ200542), and the Humanities and Social Sciences Project of Jiangxi Education Department (No. JC20201).