Spatiotemporal changes of poverty based on the perspective of ecological poverty reduction: evidence from guangdong province, China

To explore the evolving trends of ecological poverty reduction in impoverished counties within Guangdong Province, this study adopts the Driver-State-Response (DSR) model. We establish an ecological poverty evaluation index system, predict the poverty reduction index using a neural network, analyze the developmental relationship between the ecological environment and socio-economy within Guangdong’s impoverished counties utilizing an improved decoupling index, and finally employ a heatmap to dissect the spatiotemporal distribution features of poverty alleviation pressure. The main findings are as follows: (1) Fewer poor counties in Guangdong Province decouple natural ecology and social economy between 2013 and 2021, but overall poverty pressure is declining; (2) The pressure to reduce poverty and its changes vary greatly across different regions, with the pressure being concentrated in the north and less in the center and east. The essay makes appropriate policy recommendations for reducing poverty in various Guangdong Province regions based on the aforementioned findings.


Introduction
With increasing globalization and rapid economic development, poverty has become one of the most pressing and challenging issues globally, closely related to economy, education, health, environment, and sustainable development (Su et al 2022).Absolute poverty was completely eliminated in China by the end of 2020 (Wang et al 2021), the focus of poverty alleviation has shifted to reducing relative poverty, but research on how relative poverty is measured, how it changes over time in terms of patterns of space and time, and what factors contribute to it is generally lacking (Wang and Wang 2021), and the accurate identification of poverty levels and povertycausing factors of poor farm households is the first step in alleviating both absolute and relative poverty (Wang et al 2022).In China, poverty has been a major issue, particularly in the Guangdong Province's impoverished counties.As a sizable agricultural region in China, Guangdong has had extremely uneven development, with 'the richest and poorest places in the nation both being in Guangdong.'There are still 28 poor counties in Guangdong Province, according to data published on the Guangdong Poverty Alleviation Website.Most of these impoverished counties are located in northern Guangdong's hilly regions and western Guangdong's less developed regions, both of which have less established transportation systems.Guangdong's impoverished counties made up 49.12% of the province's total counties by 2021, yet their GDP made up only 3.46% of the GDP of the entire province.In 2022, urban inhabitants had a per capita disposable income of 56,905 yuan, compared to only 23,598 yuan for rural residents.There are mountains, plains, and hills among the 28 impoverished counties in Guangdong Province.Among them, Liannan, Lianshan, Zijin and other mountainous counties and cities have rugged terrain.Ecological poverty eradication is integral part of the road of poverty reduction with Chinese features (Zhang 2021).The so-called ecological poverty reduction is to protect and restore the ecological environment in impoverished areas, to change the mode of economic development in impoverished areas, to end the vicious cycle of poverty and ecological destruction, and to strike a balance between ecological improvement and the strategic goal of poverty reduction.Ahmed (2011) highlighted that extreme climate events can significantly increase the population living in poverty in Tanzania, consequently intensifying the existing pressure on food security and further exacerbating poverty levels.In recent years, Guangdong Province has proactively embraced ecological poverty reduction initiatives, fostering the development of emerging sectors such as eco-tourism and eco-agriculture to generate employment opportunities and augment revenue streams for impoverished regions.The critical interplay between poverty and ecological factors has been characterized as a reciprocal relationship by Jehan and Umana (2003).Addressing this complex interrelationship and effectively breaking the cycle of poverty and ecological degradation remains a topic of broad interest and ongoing discussion (Le and Leshan 2020).Poverty represents a pervasive global issue, necessitating tailored strategies in diverse locations to alleviate its impacts (Cobbinah et al 2013).It is important for the economic, social and environmental development of poor counties in Guangdong Province, and there are few studies on the relationship between ecology, environment and poverty in Guangdong Province (Cheng et al 2018).
In order to address this problem, the study use the DSR model and Pearson test to build a comprehensive measurement index system of poverty reduction index.After that, the neural network is used to calculate various types of poverty reduction/poverty-causing indices, and finally, it is used to determine the value of the comprehensive poverty reduction index of each poor county in Guangdong Province.Then, in order to provide a scientific foundation for the development of policy and the implementation of ecological poverty reduction in poor counties in Guangdong Province, decoupling index and heat map are used to analyze the spatial and temporal evolution of the comprehensive poverty reduction pressure in each poor county.

Literature review
A search of the literature reveals that current research by scholars can be broadly classified into three categories:

Application of the DSR model
The current speed of poverty reduction has been slowed down by the frequent occurrence of catastrophic weather events and the worsening of ecological challenges, which has drawn the attention of academics.Many academics have used the DSR model to build the corresponding indicator system for the relationship between economy, society, and ecology.This indicator system was developed based on the PSR (Pressure-State-Response) model, which only describes static environmental problems and does not have the capability of evaluating the problems associated with sustainable development as a whole.The DSR model's indicator system more thorough and current (Gao et al 2021) and more closely relates to sustainable environmental goals (Long 2008).It also emphasizes the causal link between environmental pressure and environmental degradation.Li et al (2018) constructed a time-series evaluation system for the redevelopment of abandoned industrial squares in coal mines based on the DSR model; Yang and Cui (2022) constructed a regional abandoned mine redevelopment with the requirements of regional mining and sustainable social development as the driver and GDP and disposable income as the response potential comprehensive evaluation framework.Bai et al (2022) and Zhang et al (2017) constructed a comprehensive evaluation index system for carbon audits, using total assets and total energy consumption of enterprises as drivers and major greenhouse gas emission reductions as responses.Chen (2017) used GDP per capita as the driver and senior talents in marine science and technology, etc as the response to construct an index measurement system for the marine strategic resource security index.In conclusion, there is a wealth of research based on DSR models, involving various social, economic, environmental and policy aspects, providing important theoretical and practical support for achieving sustainable development.

The coupling of ecology and poverty
The exploitative development and utilization of resources by communities have initiated a detrimental cycle, reinforcing the delicate ecological environment's interaction with poverty (Zhang et al 2001).This spatial coupling between the ecological environment and poverty holds significant implications.Understanding the mechanisms underlying their interaction is crucial for both ecologically restoring vulnerable areas and implementing targeted poverty alleviation efforts (Chen et al 2021).Similarly, Hooli (2016) advocated for a comprehensive consideration of the strategic importance of ecological environmental protection in impoverished regions.However, scholarly investigations into the coupling relationship have yielded diverse conclusions.Qin et al (2020) and Wang et al (2020) discovered that the coupling between poverty and the ecological environment has achieved a fundamental level of coordination in various regions.Conversely, Zhou et al (2021) observed a declining trend in the coupling coordination degree among China's densely concentrated poverty areas.Contrarily, Chen et al (2022) revealed a persistently high multidimensional poverty index in Guizhou.Wu and Zhao (2020) identified a pattern of 'high at both ends and low in the middle' in terms of poverty and ecological environment coupling within contiguous poor regions in Hebei.In another dimension, Cai et al (2020) noted a gradual decrease in the coupling degree between arable land pressure and rural poverty as one moves from the central region toward the east and west.Additionally, Li and Liao (2018) found that a considerable number of impoverished counties in Chongqing lack effective models for managing the interplay between poverty alleviation and the ecological environment.These discrepancies emphasize that the coupling relationship between ecology and poverty is diverse across different geographical contexts.
2.3.Spatial and temporal evolution of different poverty zones Meng and Cao (2020), as well as Zhang et al (2022), underscore the significance of proximity to main roads as a primary factor contributing to poverty.Luo et al (2021) similarly concluded that the distribution of impoverished counties is primarily associated with topographical factors such as altitude, slope, and elevation.Meanwhile, Yang et al (2021) propose that beyond the aforementioned factors, policy orientation also significantly molds the spatial distribution of pro-poor tourism villages.Tong and Kim (2019) shed light on the transformation of community poverty dynamics in Los Angeles, underscoring the impact of spatially explicit processes like wealth aggregation.Reiterating their stance, Ding et al (2023) reaffirm, based on the Tobit model, the divergent spatio-temporal evolution traits of urban and rural water poverty.
Currently, academics have developed a number of assessment methods based on DSR, and many of them have also carried out research on the spatial and temporal evolution of poverty areas as well as the linked relationship between ecological and poverty.When academic research is combined, it is simple to see that poverty no longer has a single economic form and that the study of poverty must take into account the comprehensive influence of many different elements.There is, however, little study on the connection between the environment, ecology, and poverty in Guangdong Province, and much less literature specifically on the application of multivariate evaluation methods.In order to thoroughly examine ecological poverty reduction in Guangdong Province's impoverished counties and to provide theoretical and practical support for ecological poverty reduction and sustainable development, this study will combine the DSR model with the ecological poverty reduction theory.The DSR model and Geographic Information System (GIS) technology are also combined in this study to create a spatial model for the analysis of the spatial and temporal evolution patterns and the features of the spatial distribution of poor counties in Guangdong Province.This study therefore offers some theoretical and applied innovative value.

Data sources
This report examines the spatial and temporal evolution of pressure for poverty reduction in 28 poor counties throughout Guangdong Province, including prefecture-level cities like Shaoguan, Heyuan, and Meizhou.Guangdong Province's impoverished counties cover a combined area of roughly 64,689.88square kilometers and have a population of about 8,147,500 people.These counties' geography is primarily hilly and mountainous, with warm, muggy seasons and copious rainfall.The information utilized in the article came from the Guangdong Rural Statistical Yearbook, the prefecture-level city statistical yearbooks, and the county and district statistics bulletins.The 28 disadvantaged counties in Guangdong Province were the final data set chosen for this article's study aims and data availability from 2013 to 2021.

Selection of indicators
Based on the driving(D)-state(S)-response(R) model, this study established the evaluation index system of this paper with economic, social and environmental as the three driving forces, environmental and climate change as the state, and economic, social and ecological adjustment as the response.Building upon the works of scholars like Shao et al (2008), Tan and Yu (2012), Liu (2013), and Pang and Wang (2014), a refined evaluation index system for this study was established by carefully selecting and modifying indicators.The various indicators and their classifications are presented in table 1 for reference.In order to test the degree of influence of the selected regional rural poverty factors on regional poverty alleviation pressure, this study referred to scholars Xu and Li (2005) to conduct Pearson correlation analysis between each factor and GDP per capita, and classified each indicator into six major categories of factors based on the correlation results.Before the correlation analysis, the raw data were normalised using the polar difference method to eliminate the interference of the difference in magnitude on the results.The formula for normalization is shown in equation (1): 3.2.2.Identification of poverty reduction and poverty-causing factors X1-X5 are economic poverty reduction factors, and as can be seen from table 2, all five indicators are positively correlated with GDP per capita, indicating that all these factors have poverty reduction characteristics.X6-X7 are negatively correlated with GDP per capita, and many scholars have demonstrated that there is a negative correlation between the growth of employment and economic growth, which is necessarily a growth in GDP.
The results of Hao's (2007) study showed that a decline in per capita income will lead to employment falling into a low-level equilibrium trap, while an increase in food production will exacerbate the ecological carrying capacity gap in poor counties with mainly agricultural development, so X6-X7 are classified as economic poverty-causing factors to characterise the negative effects of economic growth in the poverty reduction process; X8-X11 are all positively correlated with GDP per capita and belong to the social poverty reduction factors X8-X11 are all positively correlated with GDP per capita, which are social poverty reduction factors, covering social and rural development, infrastructure, etc, indicating that social development is conducive to poverty reduction; however, X12-X14 are negatively correlated with GDP per capita.X22 are all negatively correlated with GDP per capita, suggesting that changes in production and pollution, climate and topography can exacerbate poverty to some extent.It is worth noting that the NDVI is positively correlated with GDP per capita, which does not contradict the assumption of natural ecological poverty.The hypothesis that ecological deterioration exacerbates poverty is supported by the fact that a decline in the vegetation cover index leads to a decline in farmers' disposable income, which in turn increases poverty levels.To further refine the secondary indicators, we grouped X15-X17 as natural ecological poverty-causing factors and X18-X22 as productive livelihood povertycausing factors.The correlation results are shown in table 2.

Research methodology
Figure 1 shows the research idea of the methodology used in this paper.
(1) BP neural Networks Based on the previously identified economic impoverishment factors (rural employment, X6, and total grain production, X7), the Economic Poverty Index (EPI) was computed utilizing a BP neural network in the Matlab 2018 environment.Similar approaches were employed to calculate the Economic Poverty Reduction Index (EPDI), Social Poverty Index (SPI), Social Poverty Reduction Index (SPDI), Natural Ecological Poverty Index (NEDI), and Production and Living Poverty Index (PLPI).In the BP neural network, the hidden layer neurons utilized the hyperbolic tangent function (tansig), while the output layer neurons employed the Purelin transfer function.Herein, each impoverishment and poverty reduction factor served as the input layer, and respective impoverishment and poverty reduction indices acted as the output layer.Under the defined parameter conditions, the BP network continuously learned and adjusted weights, simulating the intricate inherent correspondences between various influencing factors and indices.This facilitated the assessment of natural impoverishment and socio-economic poverty reduction, enabling effective evaluation (Xu et al 2006).
For the determination of the number of neurons in the hidden layer, this paper first refers to the empirical formula where n i is the number of input nodes, n 0 is the number of output nodes, and a is an arbitrary constant from 1 to 10. Through continuous experiments, the statistics for testing the optimal number of hidden layer neurons is assessed using the training speed, i.e. the least number of training times is optimal.The network topology of each index was determined as a × b × c (a, b and c represent the number of neurons in the input, hidden and output layers respectively): economic poverty reduction index: 5 × 5 × 1; economic poverty index: 2 × 3 × 1; social poverty index: 4 × 4 × 1; social poverty index: 3 × 3 × 1; natural ecological poverty index: 5 × 5 × 1; productive life poverty index: 3 × 3 × 1.
To commence, the neural network was subjected to training.The training data for the network typically comprise evaluation criteria for various research subjects.Given the skewed distribution of the data samples, this study utilized the non-equidistant natural break method to categorize the factors corresponding to each index into levels (table 3).Building upon this, Matlab 2018 was employed to perform third-order spline interpolation on the data samples, expanding the training samples and constructing the training data for the artificial neural network.Ultimately, each poverty reduction (or poverty) index was classified into 5 levels, represented by 1 to 5, signifying increasing degrees of impoverishment.The neural network was constructed and trained according to predetermined evaluation standards.The training function employed was the traingdx function, utilizing a variable learning rate momentum backpropagation (BP) algorithm.The fundamental parameters for network training were set as follows: learning rate of 0.5, maximum training epochs of 10000, maximum error of 10 −4 , with other parameters set to default values.
After normalizing the actual sample values, they were input into the well-trained network.Upon running the network, the resulting values for various poverty reduction (or poverty) indices were obtained, represented as EPI, EPDI, SPI, SPDI, NEDI and PLPI.Finally, the overall poverty reduction index of each poor county(CPDI ) in Guangdong Province is calculated based on these six indices , calculated as in equation (2):  Although economic and social development plays a greater role in poverty reduction, the impact of factors such as population growth and agricultural economic development on poverty reduction cannot be excluded.Therefore, this paper refers to the findings of scholar Wang (2008) and sets the economic and social offset coefficient at 0.2 to characterize the negative effect of such factors on poverty reduction.
(2) Decoupling analysis Decoupling Analysis (DA) is the process of analyzing the correlation between two or more variables and whether the effect of one of the variables on the other diminishes or disappears (Xu et al 2021).This paper draws on decoupling analysis to study the relative relationship between ecological environment and socioeconomic development in poor counties in Guangdong Province using the results predicted by BP neural networks.For the calculation of the decoupling index, the Tapio decoupling model is improved by referring to the approach of scholar Wang (2010) to obtain the ecological and economic decoupling index.The specific formula is shown under equation (3): Where DI are the decoupling indices, the EC is the ecological index, and EI is the economic index.In this paper, the growth rate of the ecological index and the growth rate of the economic index are constructed with the help of each index predicted by the BP neural network, the EC and EI are calculated by equations (4) and (5) respectively: In accordance with the Organization for Economic Co-operation and Development (OECD) criteria, absolute decoupling in this paper refers to a decoupling index close to 0 and relative decoupling refers to a decoupling index close to 1.When DI < 0, EC is positive and EI is negative, it means that the natural ecological index and economic growth are in a strong decoupling stage; if EC is negative and EI is positive, then the two are in a strong decoupling phase.When 0 < DI < 1, it means that the natural ecological index is growing slower than economic growth and is in the relative decoupling stage.Where EI is positive, it is expansionary relative decoupling; if EI is negative, it is declining relative decoupling.When DI = 0, it means that economic growth can still be maintained when the natural ecological index remains unchanged.When DI 1, it means that the natural ecological index grows at a rate equal to or faster than the economic growth rate, and the two are in the linkage stage.Where EI is positive and is expansion-linked; and EI is negative and is linked to decline.
This study will analyze the relative relationship between ecological environment and socio-economic development in the poor counties of Guangdong Province based on the Tapio decoupling model, during which the BP neural network prediction results will be used for the calculation of the decoupling index.
(3) Spatial autocorrelation analysis Moran's I index is a spatial autocorrelation measurement method used to assess the spatial clustering degree of correlated data (Chen et al 2003).To gain further insights into the spatial characteristics of the Comprehensive Poverty Reduction Index (CPDI) in impoverished counties within Guangdong province, we employed GeoDa software to compute the Moran's I index for the CPDI values.
(4) Geographic information system (GIS) In order to visually represent the spatiotemporal distribution of poverty alleviation pressure in impoverished counties across Guangdong province, we utilized ArcGIS to create regional spatiotemporal distribution maps.Considering the distribution pattern of the Comprehensive Poverty Reduction Index, we employed the natural breaks method to classify both the poverty reduction index and the alleviation pressure into distinct levels.The classification for different years is depicted in the legends of figures 2-4.

Decoupling analysis
Table 4 shows the pattern of changes in the decoupling index between economy and ecology in poor counties in Guangdong Province from 2013 to 2021.
The poor counties of Guangdong Province are currently developing economically and socially at the expense of the region's natural ecological environment due to an expansionary linkage between ecology and socio-economics.The northern region's decoupling situation is not encouraging.The state has improved and then deteriorated in six of the poorer counties, culminating in an expansionary linkage or a strong reconnection, with the environment's carrying capacity declining and the economy contracting, which hastens the issue.The two impoverished counties in the North that saw no change in their decoupling status from 2014 to 2021 as well as those that saw an improvement each show that these differences were significant.The center region's decoupling status change is more consistent.While the remaining six counties and districts remain in the expanded pegged status, the decoupling status of Fogang, Lianshan, and Liannan counties has all improved, with Lianshan and Liannan achieving decoupling.This shows that more impoverished counties continue to face greater environmental pressures when it comes to economic development.By 2021, every impoverished county in the east has been tagged.Guangdong Province's poorer counties have succeeded in decoupling to a lesser extent overall.Locals have engaged in predatory resource  management and exploitation as a result of the fragile ecological balance and less developed natural resources, which may have in turn made human productivity and behavior more constrained.

Time series changes in the composite poverty reduction index of poor counties in guangdong province
The trends of the composite poverty reduction index for poor counties in Guangdong Province from 2013 to 2021 are shown in table 5 and figure 2. The horizontal coordinates of figure 2, 1-28, represent the 28 poor counties in northern, central and eastern Guangdong Province respectively (the specific ranking is the same as in table 4).The number of poor counties with an overall poverty reduction index exceeding 2 is 1 in 2013, 11 in 2018 and 18 in 2021, accounting for 64.28%.In general, poverty pressure in Guangdong Province is decreasing and poverty reduction is becoming more effective.
The pressure to reduce poverty is focused in poor counties in the north, where poverty reduction is rather weak, as evidenced by the average value of the combined poverty reduction index for poor counties in the north, center, and east.The average value of its index won't surpass 2 until 2021, but both the central and eastern portions did so in 2017.Guangdong Province's middle and northern regions, which have poor geological characteristics, are where you'll find the majority of the province's mountains and hills.For instance, the steep topography of the northern city of Heyuan results in poorly developed transportation, and Heyuan's role as a water supplier to Hong Kong, Macao, and the Pearl River Delta region restrains industrial growth to some extent.Shaoguan, on the other hand, is more difficult to develop as a transport hub city because it is surrounded by mountains on all sides and the basin is primarily made up of agricultural or residential land.
The Composite Poverty Reduction Index shows the highest rise in the Central region, going from 1.27 in 2013 to 3.17 in 2021-exceeding the East for the first time in 2020.A couple of center counties worth mentioning include Liannan and Lianshan autonomous counties, neither of which will have a combined poverty reduction index of greater than 2 by the year 2020.These counties have a lower combined poverty reduction index than other central counties.The most mountainous region of Guangdong, Lianshan and Liannan counties have few plains and numerous mountains, the majority of which are still limestone areas, challenging transportation development, and limited educational coverage.The planning for reducing poverty in Liannan and Lianshan must receive the center's full attention.
The effect of poverty reduction in the eastern region is relatively optimistic, especially in Rao Ping County, where the composite poverty reduction index in 2013 was 3.18, surpassing the index value of Pingyuan and Wuhua counties and districts in 2021.Overall, the composite poverty reduction index for poor counties in the north, centre and east showed an upward trend in fluctuation.

Spatial autocorrelation analysis of the composite poverty reduction index of poor counties in guangdong province
Table 6 presents the Moran's I index values for the Comprehensive Poverty Reduction Index.It is not difficult to find that the Moran indexes for 2013-2021 all passed the 1% significance test, i.e. there is a spatial autocorrelation of the composite poverty reduction index for all poor counties in Guangdong Province, but this correlation is weak.The Moran index is greater than 0 in all years, indicating that there is a spatial aggregation of 'high' or 'low' in the composite poverty reduction index of each county.From the change of Moran index, the spatial autocorrelation of poor counties in Guangdong Province shows a fluctuating increase, from 0.226 to 0.256 in 2013, but the index continues to decline in 2019-2021, indicating that the change of poverty reduction pressure in one county has a decreasing impact on the poverty reduction pressure in its neighbouring counties.
The increase of poverty reduction level no longer shows obvious spatial aggregation effect.The degree of economic linkages and interactions among poor counties in Guangdong Province is increasing, the pace of development in poor areas tends to be consistent, and the process of poverty reduction is smoother and more balanced across the province.
4.2.3.GIS-based analysis of the regional spatial and temporal distribution of poverty reduction pressure in poor counties in guangdong Province Figures 3-5 plots the regional spatial and temporal distribution of poverty reduction pressure in poor counties in Guangdong Province, where different colors represent the poverty reduction pressure in each county and district.The larger the poverty reduction index, the lower the poverty reduction pressure in that county or district.
In order to visualize the regional spatial and temporal distribution of poverty reduction pressure in poor counties in Guangdong Province, we used Arcgis to map the regional spatial and temporal distribution of poverty reduction pressure in poor counties in Guangdong Province (see figures 3-5).Taking into account the distribution characteristics of the composite poverty reduction index, we used the natural breakpoint method to classify the poverty reduction index and the poverty reduction pressure into grades, and the grades for different years are shown in the legend of figures 3-5.The higher the poverty reduction index, the lower the poverty reduction pressure in that county.Furthermore, utilizing the method of unnatural breakpoints to segment the poverty reduction index, we illustrate in figure 6 the temporal evolution of class thresholds for poverty reduction pressure across the years 2013, 2008, and 2021.Evidently, the thresholds corresponding to distinct levels of pressure exhibit a gradual increment as the poverty pressure undergoes changes.To illustrate, the threshold for higher poverty reduction pressure ascends from 0.089997 to 1.122866 in 2013, progressing to 1.122814 to 2.007355 in 2021.This trend serves to reinforce the observation that the overall poverty reduction pressure within impoverished counties in Guangdong Province is declining.
The number of counties with high pressure to reduce poverty is declining over time, while counties with low pressure to do so is rising.In 2013, there were only three counties with low pressure to do so; by 2021, however, there will be nine, showing that Guangdong has been quite successful in recent years at reducing poverty.Additionally, between 2013 and 2021, the pressure to reduce poverty has decreased to varying degrees in many counties and districts.For instance, Zijin and Luhe counties went from a high to a moderate level of pressure, while Longchuan and Guangning went from a high to a low level of pressure.However, there are some counties, such as Lianping in the north and Liannan in the center, where the demand to reduce poverty is still significant, indicating that poverty reduction in these areas needs to be strengthened more.Moreover, Heping in the center has seen an increase in demand to lower poverty.In Guangdong Province, there is a spatial concentration of poverty in impoverished counties.While three counties and districts are already under reduced pressure to reduce poverty in 2021, the pressure to reduce poverty in Liannan and Lianshan in the center part of the country is still strong.Among them, the pressure to reduce poverty in the central part of the country is falling faster.Pressures to reduce poverty are concentrated in northern Guangdong Province, and there is contiguous poverty in the counties of Nanxiong, Shixing, and Wongyuan, which are primarily steep and hilly with a dearth of natural resources like land and water.Contiguous poverty can hinder the region's economic growth and cause a significant brain drain, both of which further the problem of poverty.Hence, regional governments should choose the objectives and tasks for each of their individual efforts to reduce poverty in accordance with their own resources and unique qualities in order to create a synergy and execute resource sharing.In order to increase the effectiveness of poverty reduction, the government should modify its strategies for reducing poverty in the poor counties in eastern Guangdong Province where the pressure to do so is more stable.

Discussion
Academic approaches to poverty measurement are transitioning from unidimensional to multidimensional.The interdependent and constraining nature of the complex relationship between ecology and socioeconomics also profoundly affects the poverty alleviation trajectory in China (Cheng et al 2018).In this paper, a comprehensive quantitative index system to measure the comprehensive poverty index is constructed by integrating economic, social and ecological variables on the basis of the DSR model.BP neural network is used for measurement purposes.The method visualises the relative balance between economy and ecology in the reservoir area and the spatial and temporal distribution of poverty alleviation pressure in the reservoir area.The DSR index and GIS are used to dynamically and intuitively portray the spatial and temporal changes in the comprehensive poverty alleviation pressure in the poor counties of Guangdong Province, and to provide a theoretical basis for accurate poverty alleviation in the counties and districts.However, the research methodology employed here requires large data requirements, but data access is limited due to factors such as hydrology, geological hazards and human dynamics, which contribute to poverty and poverty reduction but remain under-exploited.In future research, the lack of socio-economic data at the village level in China can be addressed by obtaining more comprehensive data through field research and exploring the distribution of poverty areas in more detail.Sustainable development is a highly complex multidisciplinary research field (Urbaniec et al 2017) and a dynamic process affected by multiple factors and their complex interactions (Zhang et al 2017), and System Dynamics Model (SDM) is a very effective tool that has obvious advantages in analysing complex system dynamics and simulating interactions and responses under different conditions (Sahin et al 2015).Therefore, future research should try to combine methods such as system dynamics to further explore the complex interactions of economy, society and ecology.In addition, third-party assessment plays an important role in improving the precision of poverty alleviation (Yang and Liu 2021), therefore, when researching precision poverty alleviation, attempts can be made to establish a more comprehensive poverty index measurement system and develop a more comprehensive and universally applicable indicator framework.

Conclusions and recommendations
Comprehensive research indicates that certain impoverished counties in Guangdong Province have achieved a decoupling between natural ecology and socio-economic development, thereby alleviating overall poverty to a certain extent.However, this decoupling phenomenon is not universal, and many regions still adversely affect the natural environment in the pursuit of economic growth.This aligns with prior findings by scholarsalthough in some regions, the relationship between poverty and ecology has reached a relatively harmonious state, numerous areas still experience a vicious cycle between poverty and ecological degradation (Li et al 2019).This could be attributed to variations in industrial structure and development stages across different regions.Some areas may heavily rely on high-polluting and high-energy-consuming industries, tightly linking economic growth with environmental burdens.Hence, future research, especially when analyzing poverty alleviation measures, should consider the perspective of industrial structure.
Moreover, poverty in Guangdong Province is gradually showing signs of improvement, with a reduction in spatial agglomeration effects.The northern region is experiencing a phenomenon of concentrated poverty, the central region exhibits relatively higher poverty reduction efficiency, while the eastern region has not achieved decoupling between ecology and the economy.It is evident that the poverty reduction situation varies among different impoverished counties in Guangdong Province, and this varies from other global regions as well (Park 2019).Zhou and Xiong (2018) concluded through their study that the distribution of poverty is primarily associated with topographical factors such as altitude, slope, and elevation.Therefore, in future studies examining the factors affecting poverty reduction in impoverished counties, one should not only consider the region's level of economic development but also emphasize natural factors and government actions.
In light of this, the paper suggests that the northern region's government should fortify ecological safeguards, foster the decoupling of ecology and economy, and tailor resource sharing to the unique attributes of individual counties and districts to prevent contiguous poverty agglomeration.In the central region, targeted poverty reduction efforts should be directed towards Lianshan and Liannan, bolstering economic development through the strengths of Qingxin District and other counties to achieve equilibrium.In the eastern region, despite moderate poverty reduction pressures, the effectiveness of poverty alleviation remains relatively modest, prompting regional governments to recalibrate poverty reduction strategies for each county, and intensify environmental pollution oversight and sanctions.Subsequent research can delve more profoundly into the intricate relationship between ecological and economic factors, particularly in striking a balance between economic growth and ecological conservation within the context of sustainable development in impoverished regions.Furthermore, examining the influence of diverse industries on the ecological environment and exploring potential outcomes under varied policy frameworks can be specifically scrutinized via the establishment of more meticulous models.

Figure 2 .
Figure 2. Scatter plot of the composite poverty reduction index.

Figure 3 .
Figure 3. Spatial distribution of poverty reduction pressure in poor counties in Guangdong Province in 2013.

Figure 4 .
Figure 4. Spatial distribution of poverty reduction pressure in poor counties in Guangdong Province in 2018.

Figure 5 .
Figure 5. Spatial distribution of poverty reduction pressure in poor counties in Guangdong Province in 2021.

Figure 6 .
Figure 6.The starting point of the poverty reduction pressure level threshold.

Table 1 .
Indicator system based on the DSR model.

Table 2 .
Results of Pearson correlation analysis.Data source: calculated by the authors using SPSS 20.0.*** indicates a significance level of 0.01 for the correlation coefficient, ** indicates a significance level of 0.05 for the correlation coefficient; * indicates a significance level of 0.1 for the correlation coefficient.

Table 3 .
Evaluation criteria of BP network for each index.

Table 4 .
Decoupling index for poor counties in Guangdong Province.

Table 5 .
Composite poverty reduction index for poor counties in Guangdong Province.

Table 6 .
Moran's I values for the CPDI for poor counties in Guangdong Province.