The change of ecosystem resilience and its response to economic factors in Yulin, China

Ecosystem resilience at a regional scale is an important indicator and component of ecological health assessment. Ecosystem resilience refers to the self-regulating and restoring capacity of an ecosystem to restore itself to its initial state after deviation from the equilibrium state due to disturbances. In this study, Fragstats and ArcGIS software were used to calculate various indicators of ecosystem resilience, then a model of ecosystem resilience was used to evaluate and analyze the characteristics of ecosystem resilience in 12 counties of Yulin Prefecture, China. The ecosystem resilience–economy responses were discussed using Pearson correlation coefficients and multiple regression analysis accordingly. The results suggested that ecosystem resilience in Yulin increased steadily during 1995–2015, with the self-regulating and self-restoring capacity strengthened over time. The study also found that the total output value of agriculture, forestry, animal husbandry, and fishery, fiscal expenditure and gross investment in fixed assets were significantly and positively correlated with ecosystem resilience, with correlation coefficients of 0.716 (p = 0.000 < 0.01), 0.589 (p = 0.002 < 0.01) and 0.411 (p = 0.046 < 0.05), respectively. The proportion of primary industry were negatively correlated with ecosystem resilience, with correlation coefficients of −0.445 (p = 0.029 < 0.05). The research revealed the development and changes of ecological resilience in Yulin and the response to the social economy. The findings provided support for ecological health diagnosis and decision-making concerning sustainable development in the study area and beyond.


Introduction
The term 'resilience' was first introduced into the study of ecology by Holling (1973), and was defined as the ability of the ecosystem to withstand external disturbances and maintain a steady state (Junming and Matthias 2013). Later, some scholars (Stringham et al 2003) tended to define resilience more toward the concept of stability, that is, ability of the system to return to a stable state within a certain time after perturbance. With the continuous deepening of regional sustainable development research, the definition of ecosystem resilience has also been enriched and developed, and the viewpoint that ecosystems have self-regulation and self-restoration capabilities has been recognized gradually by most scholars (Sandra et al 2011). Among them, Walker et al (2004) and Takehiro et al (2015) believe that the perturbance resistance of an ecosystem that maintains its original structure and characteristics unchanged for a certain period of time is the ecosystem resilience. However, Gao (2001) and Wang et al believe that the self-regulation capacity of an ecosystem to return to its initial state after deviating from the equilibrium state denotes the resilience. Thus, ecosystem resilience includes two aspects: the ecosystem elastic strength coefficient and the ecosystem elastic limit, that is, the size and range of ecosystem resilience (Sterk et al 2013, Dou et al 2018. Among these, the ecosystem elastic strength coefficient refers to the nature of the ecosystem determined by its own conditions, such as topography, geomorphology, climate, soil, hydrology, and vegetation conditions (Zhang et al 2008). The ecosystem elastic limit refers to the ultimate capacity of ecosystem self-regulation, that is, the capacity of self-recovery after the maximum disturbance.
There are many research methods regarding ecosystem resilience (Ruth et al 2000, Karl-Göran et al 2003, Stephen et al 2013, Liao et al 2015. Some scholars used pasture model (Janssen et al 2004), state and transition model (Dardo et al 2013), land net primary productivity model (Ashutosh and Manish 2018) and elastic limit model (Zhang et al 2008) to quantitatively analyze the resilience of ecosystems in Australia, Argentina and China. Some scholars used the goal stratification method (Wang 2011, Luciano et al 2018 and comprehensive evaluation method (Wang et al) to evaluate the resilience of ecosystems in Beijing and Hunan in China from 2001 to 2013.
Comparing the above studies, it was found that the pasture model and the state and transition model had limitations in the study area in the study of ecosystem resilience. The model of land net primary productivity and elastic limit is mainly based on the quantitative calculation of vegetation photosynthesis and the proportion of land use types, which cannot comprehensively analyze the index factors affecting ecosystem change. Comprehensive evaluation method and the goal stratification method emphasize subjective qualitative analysis but lack quantitative research on ecosystem resilience.
With the rapid development of urbanization and industrial economy, the problem of ecological environment destruction has become increasingly prominent. How to control the impact of regional economic development on the ecological environment within the carrying capacity of the ecological environment, this has aroused wide concern of the international community. Therefore, studying the response relationship between ecosystem resilience and regional economic development can provide reference for the coordinated development of ecological environment and social economy.
Yulin City of Shaanxi Province is located in the semi-arid region of northwest China, with fragile ecological environment but rich mineral resources (Yan et al 2020). In 2021, Yulin ranked second in the GDP of northwest China, which is a typical resource-based city (Hao et al). At the same time, the geographical pattern and economic development of the north and the south of Yulin differ greatly. The six counties in the north are the aeolian sand and grassland area, which is mainly dominated by energy economy, while the six counties in the south are the loess hilly and gully area, which is mainly dominated by agricultural economy. Therefore, taking Yulin City as the research object with rapid economic development, based on the ecological environment changes of 12 counties (districts) in Yulin from 1995 to 2015. We used the ecosystem resilience model (Sterk et al 2013, Dou et al 2018 to calculate the change characteristics of the ecosystem resilience in Yulin area. Then, Pearson's correlation coefficients and multiple regression analyses were used to explore the relationship between ecosystem resilience and economic response. The findings will provide scientific support for the coordinated development of regional ecology and economy.

Study area
Yulin City (107°28′-111°15′E, 36°57′-39°34′N) is located in the northernmost part of Shaanxi Province, on the southern edge of the Mu Us Desert, bordering the four provinces of Gansu, Ningxia, Menggu, and Shanxi. Yulin City has jurisdiction over Yuyang, Shenmu, Fugu, Dingbian, Jingbian, Hengshan, Jiaxian, Mizhi, Wubao, Suide, Qingjian, and Zizhou Counties (figure 1). The northern part of Yulin is a wind-sandy area, and the southern part is a loess hilly and gully area. Yulin is one of the most severely eroded areas in the middle reaches of the Yellow River (Jia et al 2020). The area has a large variation of temperature between day and night and annual average rainfall is 404 mm (Hui et al). It has a typical mid-temperate, semi-arid continental monsoon climate (Ge and Zhang, Wang et al 2022). The soil types in Yulin mainly include aeolian soil, yellow soil, new soil, chestnut soil, holene soil, and salinite. The zonal vegetation is steppe and forest steppe, and the main vegetation is forest, steppe, thicket, meadow, sandy vegetation, and halinite vegetation (Shi et al 2019). Yulin has plenty of wetlands and minerals (e.g., coal, natural gas, oil, and salt), which are the core of the national energy-chemical industry (Li and Zhang 2018 where, E is the ecosystem resilience, λ is the adjustment coefficient (Luciano et al 2018) (taken as 0.01), μ is the ecosystem elastic strength coefficient, and ECO res is the ecosystem elastic limit. Model determination is based on the determining factors of μ and ECO res , and selected properties (Gao 2001, Zhang et al 2008, Dardo et al 2013: where H is the landscape diversity index, V is the vegetation index, c 1 is annual temperature change, c 2 is annual precipitation variability, P i is the area coverage percentage of the land type, and S i is elasticity score of the land type, m is the number of land types.

Calculation of H
The H reflects the degree of diversification of landscape types in the study area and the changes in their proportions (Nancy et al 2008, Liu et al 2014, Chen et al 2019. The higher is H, the greater is the ecosystem resilience (Huang et al).
The H mainly depends on pixel size, geomorphological scale, and land-use classification. ArcGIS 10.1 software was used to rasterize land-use maps. Landscape index calculation software where, H is the landscape diversity index (Shannon's index), P i is the proportion of landscape types, and n is the number of landscape types in the study area.
where, NDVI is in the near-infrared band (0.841-0.876 μm) and Red is the red band of visible light (0.62-0.67 μm). The value of NDVI is between −1 and 1, with NDVI 0 denoting non-vegetated areas.
The data were preprocessed by the Maximum Value Composite method to achieve atmospheric correction, radiation correction, and geometric correction of influencing factors such as aerosol and solar altitude angle (Li et al 2011). 3.1.4. Annual precipitation variability and annual temperature variability Annual precipitation variability refers to the inter-annual variability of precipitation. Usually, the precipitation variability refers to the relative precipitation variability, i.e., the percentage of the absolute precipitation variability to the multi-year average precipitation (Xie and Shu): where, C 1 is the relative variability of the multi-year average precipitation in the study area, R i is the actual precipitation in a certain period of the year, and R is the average annual precipitation in the same period. The annual temperature variability refers to the inter-annual change of temperature, expressed in terms of the relative variability of inter-annual temperature: where, C 2 is the relative variability of the multi-year average temperature in the study area, T i is the temperature in a certain period of the year, and T is the average temperature in the same period.

Correlation derivation
For data with different magnitudes and units, the influence of dimension should be eliminated by standardization in mathematical statistics. The z-score standardization method is also called the standardized score. The processed economic and ecological data conformed to the characteristics of a normal distribution, that is, with mean 0 and standard deviation of 1. The function expression follows: where, μ is the mean of all sample data and σ is the standard deviation of all sample data.
In order to accurately evaluate the relationship between ecosystem resilience and economic development in Yulin City, based on related research achievements (Yang andTong 2013, Zhang et al 2014), considering that economic development is mainly reflected in the growth of economic aggregate, the improvement and optimization of economic structure and the promotion of economic quality. This paper focuses on the economic scale, economic structure and economic benefits of Yulin City. 11 selected economic factor to establish correlation with ecosystem elastic force (gross domestic product, total populationat the end of the year, per capita gross domestic product, total output value of agriculture, forestry, animal husbandry, and fishery, total industrial output value, Proportion of primary industry, Proportion of secondary industry, Proportion of tertiary industry, fiscal revenue, fiscal expenditure, gross investment in fixed assets).

Multiple regression analysis
Regression analysis is a statistical method used to predict and analyze one or more dependent variables through a set of independent variables (Ding, Zhu et al 2022). This method makes full use of the correlations among the data, and evaluates and predicts the response effect of the independent variable on the dependent variable.
After calculating the economic factors with good correlations with resilience of the ecosystem, a regression model was established between the variables of ecosystem resilience and economic variables, and R, F and T tests in the model were carried out. The predicted value of the elastic force of the ecosystem was obtained with the assumption that the regression equation was feasible. Expression of a multiple regression equation is as follows (Ding): where, X is an independent variable; Y is the dependent variable; a 0 is a constant; a n (n > 0) is the regression coefficient of the independent variable, which represents the influence of the independent variable on the dependent variable; and ε is random error, which indicates the degree of influence of other random factors on the dependent variable.

Data acquisition and processing
The land-use change map of Yulin City obtained from the China Land Use Remote Sensing Monitoring Database (in 1995 and 2015) was used to calculate the landscape diversity index. The LUCC production data were based on LandsatTM/ETM remote sensing images of each period as the main data source, and generated through manual visual interpretation. The resolution was 1 km. The China Land Use Remote Sensing Monitoring Database (https://resdc.cn/data.aspx?DATAID=97) is currently the most accurate remote sensing monitoring data product of land use in China (Wei et al 2022), with an overall accuracy of more than 90%. It has played an important role in national land resource survey, hydrological and ecological research.
In order to objectively reflect the operability of the landscape diversity index, a GIS database based on the classification of first-level land-use types were established, including six land-use categories: cultivated land, forest land, grassland, water area, construction land, and unused land. Supported by the Spatial Analyst module in the ArcGIS 10.1 system, the raster conversion of land-use type vector data was carried out.
The NDVI data were acquired from the International Scientific Data Mirror Website of the Computer Network Information Center of the Chinese Academy of Sciences (http://gscloud.cn). The NDVI vegetation index product of MODIS China (http://modis.gsfc.nasa.gov/) was processed by MOD (MYD) 09GA through inversion, splicing, cutting, projection conversion, unit conversion, and other processes. The coordinate system was EPSG:4326 (WGS84), and the spatial resolution was 1 km, which was used to calculate the vegetation index in Yulin. The research period was 1995-2015.
The rainfall and temperature change data were obtained from the Yulin Meteorological Bureau (http://sn. cma.gov.cn/dsqx/ylqxj/). The data series for 30 consecutive years of 1972-2001 were selected to calculate the annual precipitation variability and annual temperature variability in Yulin.
The economic data (gross domestic product, total populationat the end of the year, per capita gross domestic product, total output value of agriculture, forestry, animal husbandry, and fishery, total industrial output value, Proportion of primary industry, Proportion of secondary industry, Proportion of tertiary industry ,fiscal revenue, fiscal expenditure, gross investment in fixed assets) of Yulin City, including 12 counties (Yuyang, Shenmu, Fugu, Dingbian, Jingbian, Hengshan, Jiaxian, Mizhi, Wubao, Suide, Qingjian, and Zizhou) were obtained from the Yulin Statistical Yearbook of the Yulin Bureau of Statistics for 1995-2015 (http://tjj.yl.gov. cn/list/tjgb).

Results and analysis
4.1. Elastic strength coefficient of the ecosystem 4.1.1. Land use and landscape diversity index Figure 2 and table 1 show the land use of various counties (districts) in Yulin for 1995 and 2015. Overall, the proportion of grassland area in Yulin was the largest (45.26% and 43.97%, respectively), followed by cultivated land (40.18% and 37.29%), and the proportion of construction land was the smallest (0.36% and 1.89%).
Temporally, from 1995 to 2015, the area of cultivated land in Yulin decreased by 2.94%, grassland decreased by 1.2%, water area decreased by 0.03%, and the area of cultivated land was mainly transformed into forest and construction land. Forest land, construction land and unused land continued to increase, by 1.79%, 1.53%, and 0.85% by 2015, respectively. The land use structure of Yulin tended to be stable, the level of complexity gradually improved, and the landscape types were gradually enriched.
In terms of geographical distribution, compared with the six northern counties in Yulin Prefecture, the southern six counties accounted for the largest proportion of cultivated land (54.17% and 48.7% in 1995 and 2015, respectively), followed by grassland and forest land. In the past 20 years, the proportion of cultivated land decreased by 5.47% (from 54.17% to 48.7%), while that of forest land increased by 4.15% (from 6.58% to 10.73%). The area of grassland in the six northern counties was the largest, followed by that of cultivated land, the proportion of grassland and cultivated land decreased by 2.23% and 1.78%, respectively, while the proportion of construction land increased by 1.87% (from 0.39% to 2.26%). On the whole, the cultivated land area in the two regions showed a declining trend during the 20 years, and there was a higher proportion of cultivated land area in the six southern than in the six northern counties.
Based on the classification of land use, the landscape diversity index of each of the 12 counties (districts) under the jurisdiction of Yulin City was calculated (table 2). The landscape diversity index of Yulin showed an increasing trend during 1995-2015. In 2015, it was 9.10% higher than in 1995, and the ecosystem resilience continued to increase.
Of the spatial changes in 12 counties (districts), the highest average landscape diversity index in the 20-year period was in Yuyang District (1.314) and the lowest in Mizhi County (0.738). As of 2015, Fugu County had the highest growth rate (29.4%), followed by Jia County (16.77%); Jingbian County had the smallest growth rate (4.74%) and Dingbian County had a negative growth rate (−6.26%). Among them, the landscape diversity indexes of Fugu and Jia counties, although greatly increased, were still lower than that of most counties; in 2015, the landscape corresponding diversity indexes were only 1.030 and 0.947. The overall landscape diversity index was greater for the six counties in the north than the six in the south, being about 0.144 higher in 2015. During 1995During -2015, the NDVI values of Yulin showed a trend of ups and downs but an overall increase in 2015 of 0.24 compared with 1995 (table 3). Among them, Qingjian County had the largest increase (0.38) and Jingbian County had the smallest (0.23).

Analysis of NDVI
In terms of geographical distribution, the average value of NDVI was greater in the southern six counties than in the northern six counties over the years, and vegetation index was higher in the eastern than the western region. According to the average precipitation of Yulin, except for Jiaxian, the precipitation of the other five counties in the six southern counties is higher than that in the six northern counties (table 4). Combined with the research of Luo et al (2016), it can be seen that the precipitation in the northern part of Shaanxi shows a decreasing trend from south to north. The annual temperature and annual precipitation variabilities of Yulin were 6.6% and 17.9%, respectively. The largest and smallest annual precipitation variabilities in Yulin (districts) were for Wubu (14%) and Jiaxian (4.8%), respectively. The largest and smallest relative annual temperature variabilities were for Dingbian (20.1%) and Suide (14.9%) Counties, respectively.
The annual precipitation and annual temperature variabilities were generally larger for the six northern than the six southern counties, and mainly related to the geographic characteristics of the two regions.

Ecosystem elastic strength coefficient
Temporally, the ecosystem elasticity coefficient in Yulin continued to increase with inter-annual changes (figure 3). In 2015, the ecosystem elasticity coefficient increased by 35.09 compared to 1995.
Spatially, Qingjian County (45.93-102.71) had the largest annual average increase in elastic strength coefficient, followed by Yuyang District (23.10-75.64), and Wubao County (13.60-27.04) had the smallest increase. In 1995, the elastic strength coefficient of the six southern counties except Wupu and Mizhi counties was higher than that of the six northern counties. In 2015, the elastic strength coefficient of Yuyang District in the six counties in the north increased significantly compared with that in 1995. Among the six counties in the

Analysis of ecosystem elastic limit
Various calculation indicators for elastic limit, including landscape diversity index (table 2), area percentage of land-use type, and elasticity score (figure 2) were used to calculate the elastic limit of ecosystems in the study area over the years (figure 4). The elastic limit of the ecosystem in Yulin increased steadily during 1995-2015, and reached 0.060 (figure 4). The average elastic limit was lower in the southern six than the northern six counties, but the elastic limit of the southern six counties increased by 0.078, which was 1.59 times the increase of the northern counties. Comparing different counties (districts) showed that the largest and smallest increases in ecosystem elastic limit in Fugu County (0.140) and Yuyang District (0.037), respectively.
The average elastic limits over the whole period were higher in the six northern than the six southern counties, consistent with the average change of the landscape diversity index of these regions. However, in terms of the elastic limit increase, the increase in the southern six counties was 1.59 times that of the northern counties, and the ecological system in the southern counties had good anti-interference capacity.

Ecosystem resilience
Combining the calculation results of various indicators of the resilience model (equation (1)) suggested that the resilience ranges of the ecosystem in Yulin in 1995 and 2015 were 0.011-0.310 and 0.068-0.796, respectively (table 6). The resilience of the ecosystem showed an overall increasing trend during the 20 years, with an average increase of 0.272.  The resilience of part of the ecological system in Qingjian and Suide counties was relatively high, with mean values of 0.722 and 0.626, respectively, suggesting that anti-interference capacity was strong. However, most of Wubu, Mizhi, Dingbian, and Shenmu Counties had low ecosystem resilience, with mean value of 0.051 of Fugu in 2015. Thus, the ecosystem was fragile and had poor anti-interference capacity.
The ecosystem resilience of the southern six counties was generally higher than that of the northern six counties, with an increase of 0.314 compared to 0.228, respectively.

Analysis of the relationship between ecosystem resilience and economic response
Ecological protection and economic development influence and restrict each other. Table 7 shows the correlation matrix for ecosystem resilience and economic indicators.
The total output value of agriculture, forestry, animal husbandry, and fishery, fiscal expenditure and gross investment in fixed assets were significantly and positively correlated with ecosystem resilience, with correlation coefficients of 0.716, 0.589 and 0.411, respectively (table 7). The proportion of primary industry were negatively correlated with ecosystem resilience, with correlation coefficients of −0.445. Other economic factors were weakly correlated with ecosystem resilience. To better describe the degree of influence of economic indicators on ecosystem resilience, the total output value of agriculture, forestry, animal husbandry, and fishery (X 1 ), proportion of primary industry (X 2 ), fiscal expenditure (X 3 ), and gross investment in fixed assets (X 4 ), which significant correlations, were selected for multiple linear regression analysis in predicting ecosystem resilience (Y) (table 8).
In the t-test, the p-values of the three economic indicators of total output value of agriculture, forestry, animal husbandry, and fishery, fiscal expenditure, and total fixed investment were 0.000, 0.045, and 0.001  0.411 * 0.845 ** 0.634 ** 0.800 ** 0.849 ** 0.777 ** −0.651 ** 0.790 ** −0.305 0.815 ** 0.820 ** 1 Note: X 1 gross domestic product; X 2 total population at the end of the year; X 3 per capita gross domestic product; X 4 total output value of agriculture, forestry, animal husbandry and fishery; X 5 total industrial output value; X 6 proportion of primary industry; X 7 proportion of secondary industry; X 8 proportion of tertiary industry; X 9 fiscal revenue; X 10 fiscal expenditure; X 11 gross investment in fixed assets. Y denotes ecosystem resilience. ** significant at p < 0.01. * significant at p < 0.05.
respectively, which are all less than 0.05. This suggested that these three indicators had a significant impact on ecosystem resilience. While the p-values of the proportion of primary industry was 0.846 > 0.05, it shows that the factor has little effect on the resilience of ecosystem The regression model (P < 0.05) was also effectively established.

Temporal and spatial evolution characteristics of ecosystem resilience
The elastic strength coefficient of the ecosystem in Yulin steadily increased year by year ( figure 3 and table 6). However, the ecosystem elastic strength coefficients of Fugu, Shenmu, and Hengshan Counties showed negative growth. This was likely due to mining activities (e.g., natural gas and coal) that have restricted the sustainable development of the ecological environment of the north counties of Yulin (Wang).
The ecosystem elastic strength coefficient is mainly affected by landscape diversity index and vegetation index (equation (2)). The landscape diversity index of Yulin increased rapidly during 1995-2005, and ecosystem resilience continued to increase (tables 1-3). This may be attributed to the first construction phase of the Three North Shelterbelt Project in Northwest China in the 1990s (Kang). Wang (2011) evaluated the development status of Beijing's ecosystem resilience during 2001-2008, and suggested that forest coverage was an important factor affecting ecosystem resilience. Following the Chinese Government launching the Beijing-Tianjin sandstorm control project, the vegetation coverage rate in Beijing increased from 30.65% to 36.5% during 2000-2008, achieving a remarkable ecological effect. This is consistent with the ecological benefits achieved in Yulin since the end of the 20th century in the Mu Us Desert zone in regard to afforestation, wind-proofing, and sand-fixing work (Yang 2020). As the area of desertified land shrank year by year (Wang et al 2020), the Mu Us Desert gradually disappeared from Yulin, which has made a positive contribution to improving ecosystem resilience in Yulin.
Combining figures 2 and 4 showed that the ecosystem elastic limit of Fugu County increased significantly (0.092) during 1995-2015, while that of Yuyang District was the lowest (0.021), which was mainly related to the change of forest land area in land-use types. The area of forest land in Fugu County increased by 3.44% during 1995-2015 and ranked first in all regions (table 1), while Yuyang District had the smallest increase (0.42%). Thus, the percentage change of woodland area had a great influence on the ecosystem resilience and elastic limit.
In addition, although the six counties in the north of Yulin had higher diversity indexes than the six counties in the south, the vegetation growth status and main land-use types (such as woodland and grassland) were inferior. Coupled with the influence of the overall topographic pattern of the north and south regions, the ecosystem resilience of Yulin presented a pattern of high in the south and low in the north. Therefore, human activities such as planting trees and grasses, restoring vegetation, and changing land-use types were important measures to improve ecosystem resilience. Climate changes such as precipitation and temperature are also key factors affecting ecosystem resilience (equations (1) and (2)). This is consistent with the findings of Tian and Zhang,who studied the influence of moisture and thermal conditions on ecosystem resilience in the Guanzhong area of Shaanxi Province, and of Wang et al,who quantitatively analyzed the influence of precipitation factors on ecosystem resilience in Pingjiang County, Hunan Province.
In addition, the annual precipitation variability and annual temperature variability of the six northern counties in Yulin were greater than those of the southern six counties, but precipitation and temperature were lower. Precipitation and temperature play a key role in plant growth and human activities, and also directly contributed to the difference in ecosystem resilience between the north and south. Therefore, although the climate pattern in the north and south areas of Yulin was not affected by human activities, water ecological environment improved by the rational allocation of water resourc es (Ge and Zhang, Huai and Shang 2020, Hui et al), especially to improve the utilization rate and carrying capacity of water resources in the north counties in Yulin, thereby improving regional ecosystem resilience.

Relationship between ecosystem resilience and socio-economic indicator responses
The socio-economic indicators, i.e., the total output value of agriculture, forestry, animal husbandry, and fishery, fiscal expenditure, and total fixed investment had a significant impact on the ecosystem resilience of Yulin (table 7). Further analysis suggested that the total output value of agriculture, forestry, animal husbandry, and fishery was highly correlated with landscape diversity index, NDVI, forest and grassland areas. Fiscal expenditure and total fixed investment were highly correlated with landscape diversity index, grassland area, and construction land area. This shows that Yulin has increased investment in ecological management such as afforestation and has begun to pay attention to the coordinated development of the economy and ecology. In the process of urbanization, investment in ecological environment governance has also been continuously strengthened. The ecosystem resilience in Yulin has been steadily improved, and ecosystem stability continuously strengthened. For instance, the Mu Us Desert has gradually disappeared due to afforestation activities in Yulin. This is also consistent with results of Zheng et al (2019) in a study of the relationship between landscape pattern and socio-economic response in the Danjiang River Basin in Shaanxi Province. Finally, from the perspective of economic development, the total GDP of Yulin City has continued to grow since 2005, and has maintained a ranking of second place in Shaanxi Province (Yan et al 2020). Although the current trend of ecosystem resilience is positive, the development of urbanization and industrialization will inevitably affect the ecological environment due to environmental pollution, reduction of vegetation coverage, and increases in construction land, Therefore, there should be future studies of the response relationship between ecosystem resilience and economic development, so as to better predict the development trend of ecosystem resilience.

Trends and threshold analysis of ecosystem resilience
In Yulin City, the proportion of primary industry and the total output value of agriculture, forestry, animal husbandry and fishery have significant negative and positive correlation with the resilience of ecosystem respectively. It can be seen that although the industrial structure of Yulin is changing from agriculture to industry and service industry, with the continuous growth of the agricultural economy, the resilience of the regional ecosystem is also strengthening. This is mainly because the primary industry mainly takes green plants as the production object, which is conducive to improving and optimizing the ecological environment (Zhu). While the secondary and tertiary industries have caused environmental pollution and forest area reduction due to the increase of energy consumption and urban land use, which have caused damage to the ecosystem. However, this negative impact is still within the threshold of ecosystem resilience, which means that the ecological and economic development of Yulin City is still in a balanced state at the present stage.
However, Yulin is China's national energy and chemical industry base. With the acceleration of industrialization and urbanization in Yulin since 2005 (Yan et al 2020), the positive impact of agricultural economy on ecological environment began to weaken (Tian and Zhang). Moreover, the economic structure dominated by industrial economy cannot be changed in a short time, which is bound to lead to the slowing down of the growth rate of ecosystem resilience and gradually reaching the upper limit of the threshold. Therefore, we should continue to study the change trend and threshold prediction of ecosystem resilience under the condition of rapid economic development in Yulin City, which is very important for guiding the sustainable development of regional ecology and economy.
In the following research, we can further analyze the coordinated development relationship between ecology and economy and the upper limit of ecosystem resilience in Yulin City by combining coupling coordination relationship model (Zhang et al 2014) and other methods, and provide reference basis for regional government decision-making. In addition, we suggest that while maintaining the economic development advantages of the northern region of Yulin, further promote the development of the southern region. At the same time, we should strengthen the development of agriculture and forestry economy and the investment of ecological construction funds in the northern region, so as to achieve diversified and balanced development.

Conclusion
(1) Spatio-temporal evolution of ecosystem resilience in Yulin Ecosystem resilience in Yulin showed an increasing trend with inter-annual changes. The self-regulation and disturbance-resistance ability of the ecosystem increased year by year, the ecosystem resilience in Yulin ranged from high in the south to low in the north, which was generally related to geographical differences between the two regions. The northern six counties are mostly sandy and grassy areas with fragile ecological environments and weak ecosystem restoration, while the southern six counties are mostly hilly and gully areas with good ecological environments and strong ecosystem resilience.