Analysis of spatiotemporal evolution and driving factors of ecological environment quality in the Ili-Balkhash Lake Basin in Central Asia

Abstract

The Central Asian area is faced with unprecedented challenges due to the fragility of its ecological environment and under the double pressure of global warming and intensified human activities. As a typical transboundary watershed in Central Asia, the ecological system quality of the Ili-Balkhash Lake Basin (IBLB) is of vital importance to the sustainable development of the whole region. In this study, we constructed the Remote Sensing Ecological Index (RSEI) for the IBLB (2000–2020) using the MODIS dataset on the Google Earth Engine platform and applied Sen’s slope and Mann–Kendall (Sen-MK) trend analysis methods to assess RSEI trends. Additionally, geographic detectors and geographically weighted regression models were used to identify and analyze the global and local factors influencing RSEI. The outcome of the study suggests that: (1) The ecological quality of the basin has generally improved, and the proportion of high quality and good areas has risen, but spatial heterogeneity is obvious, and the area around the lake shows a trend of degradation; (2) LST and NDVI were the main natural drivers, and LUI and AWUE were the key human drivers, with significant interactions among the factors; (3) The GWR model further reveals that LST is the main limiting factor of RSEI, and the two are negatively correlated, while there is significant spatial variability in the effects of NDVI, LUI and AWUE on RSEI, with both positive and negative effects. This study lends a scientific basis for revealing the ecological evolution of watersheds in arid zones. The methodological system it constructed can provide a reference for the ecological monitoring and governance practice of transboundary watersheds around the world.

Introduction

Dramatic global temperature changes and intensified anthropological activities are profoundly affecting ecosystems¹. Central Asia has a typical arid continental climate and its water resources are distributed in a highly unbalanced manner². In recent years, the accelerated glacier melting³, soil erosion⁴, and frequent extreme weather events⁵ have made the ecological environment in Central Asia increasingly sensitive. In addition, Central Asia is rich in endemic species and complex ecosystems, and agricultural development is important in the region⁶. The IBLB, a transboundary basin in Central Asia, has garnered significant attention from scholars. Current research on the IBLB focuses on lake level changes⁷, surface water chemistry⁸, land use/cover in the Lake Balkhash Basin⁹, and changes in the Ili Delta wetlands¹⁰. There is still little research on ecological quality^11,12. Therefore, the development of the analysis of the spatial and temporal characteristics of the ecological environment of transboundary watersheds in Central Asia and the factors influencing them is of great scientific and practical significance for the protection of biodiversity, the promotion of the economic development of the countries and the sustainable development of the region.

Since the twentieth century, there has been a drastic development of remote sensing technology, and the evaluation of the ecological environment is also gradually changing from the traditional method of field testing to remote sensing¹³. However, a single indicator ignores the interconnectedness and complexity of ecosystems¹⁴. In order to better explore the quality of the ecosystem, scholars have proposed a variety of comprehensive evaluation indices, such as the Environmental Quality Index (EQI) of the U.S. Environmental Protection Agency¹⁵, the Provincial Ecological Quality Index (PEQI) of China¹⁶, the Environmental Performance Index (EPI) of Yale and Columbia Universities in the U.S., and the Ecological and Environmental Condition Index (EI) of China’s State Environmental Protection Administration¹⁷. While these methods are good for performing an assessment of ecosystem quality, they suffer from the problem of having more evaluation indices, which produce redundant effects¹⁸; it is difficult to analyze large scale and global phenomena¹⁹. On this basis, Xu²⁰ proposed the RSEI in 2013. The index is mainly composed of four remote sensing indicators: NDVI represents greenness, WET represents humidity, NDBSI represents dryness, and LST represents heat. These four indicators were standardized, followed by principal component analysis to calculate the covariance matrix and eigenvalues and eigenvectors of the standardized data, and to determine the weighting of each factor with respect to the ecological, in a manner to construct the RSEI²¹. In addition, principal component analysis can reduce the dimensionality of the data, simplify the model, while retaining most of the information and reducing the subjective influence of human experience, and can reflect the data information more objectively and effectively according to the nature of the data itself²². This research has already been used extensively by scholars to assess environmental quality at multiple scales²³, covering various geographical units such as cities²⁴, Mining districts²⁵ and river basins²⁶. Furthermore, the method has been applied in ecosystem studies in Central Asia: Maimaitituersun et al.²⁷ calculated the RSEI of the Urumqi River Basin in the arid zone on the PIE platform to study the ecological quality of the basin and the factors affecting it; Zhang et al.²⁸ quantitatively evaluated the ecological security of the oases in the Tarim River Basin in Central Asia by using the RSEI and NPP; Hou et al.²⁹ evaluated soil salinization and salinity in the Lake Bosten Basin based on the Soil Salinization Ecological Index to assess soil salinization in Bosten Lake Basin. Thus, the method is able to effectively and systematically evaluate the ecosystem quality at different spatial scales as well as in the relevant watersheds in Central Asia. It provides important technical support for scientific research and management practice in related fields.

Exploring how natural factors and human activities affect the quality of the ecosystem is important for the scientific formulation of corresponding conservation policies and the realization of sustainable development. At present, common research methods on driving mechanisms include Ordinary Least Squares (OLS)³⁰, Geographical Detectors³¹, Random Forests (RF)³², Geographically Weighted Regression (GWR)³³, Structural Equation Modeling (SEM)³⁴ and Spatial Durbin Model³⁵. Among them, geographic detectors are widely used by scholars in related research on influencing factor detection as a method to measure spatial heterogeneity, detect explanatory factors, and analyze the interaction between variables^36,37. The significant advantage of Geographical Detector is that they can effectively quantitatively assess the relative contribution of influencing factors³⁸. Through the use of this method, the researcher is able to accurately determine the strength of the effect of each influencing factor, and accordingly, systematically rank the influence of each factor, so as to gain a deeper understanding of the extent to which different factors contribute to the formation of a particular geographic phenomenon. Compared with geographic detectors, the geographically weighted regression model allows regression analysis at a local scale, which can reveal the spatial variation patterns of driving factors and thus reveal the characteristics of spatial heterogeneity in a more detailed manner³⁹. At present, the research on ecological environment influencing factors mainly uses geographic detectors to explore the impact of single factors and interactive factors, while there are still few attempts to combine geographic detectors with geographically weighted regression models. Through this combination, we can not only grasp the overall impact pattern of influencing factors from a macro level, but also deeply analyze the local impact differences of each factor in space at a micro level. We can better reveal the multiple scales dynamic mechanisms affecting the environmental quality of the ecosystem and provide a more accurate basis of science for ecological management and policy making.

Therefore, relying on the GEE platform, we constructed the RSEI using relevant data from the MODIS dataset to quantitatively evaluate the ecosystem quality of the IBLB between 2000 and 2020. This paper analyzes in detail the dynamic trends of RSEI during the time period and applies Sen-MK trend analysis methods. On this basis, the Geographical Detector model as a comprehensive exploration of the main factors affecting the changes in the ecosystem quality of the study area was used. After identifying the key factors with high influence, this study further used GWR and conducted a regression analysis of the relationship between the roles of these key factors on a local scale. The results of the study will provide a scientifically sound basis for the development of ecological environmental protection and sustainable development policies in the IBLB and Central Asia. In addition, this study is of great scientific significance for the deep understanding of the evolution mechanism of ecosystem quality in arid and semiarid regions. It also provides a model methodology and experience for conducting ecological monitoring, assessment and driver analysis as well as for similar river basins around the world.

Materials and methods

Study area

The IBLB (72°24′ ~ 85°E, 42°11′15′ ~ 49°48′N) is part of the arid zone of Central Asia (Fig. 1). It is a transboundary river between the Ili Kazakh Autonomous Prefecture in Xinjiang, China and southeastern Kazakhstan. Its total area is 4.16 × 10⁵km², of which 0.57 × 10⁵km² is in China and 3.59 × 10⁵km² is in Kazakhstan, which is about 86% of the total area. With its unique geographic location and transboundary characteristics, the basin is a key region for the study of arid zones in Central Asia. It is not only an important part of the Tianshan Mountains and the Pamir Plateau, but also an ecological corridor connecting China with Central Asia. The topography of the watershed is complex, with significant elevation differences, showing a high southeast and low northwest trend. The southeastern part is dominated by the Tianshan Mountains, which are high and steep, and the average annual precipitation can reach more than 1000 mm, while the northwestern part is low and flat, and the average annual precipitation is only 300–400 mm. This spatial heterogeneity in precipitation, combined with the superimposed effects of temperate continental and arid climates (mean annual temperature of 17.5 °C and mean annual precipitation of 110 mm), makes the basin an ideal place to study climate change and water allocation. Strikingly, the core of the basin, Lake Balkhash, is a rare inland lake with a significant difference in salinity between the eastern and western halves of the lake. 80% of its lake water comes from the Ili River, which originates from the Tianshan Mountains, flows through the Ili Valley in China, crosses the border and then passes through the Kapchagai Reservoir in Kazakhstan before finally joining Lake Balkhash. The abundance of solar radiation and the vigorous evaporation in the IBLB, together with the complex topography and precipitation gradient, make the ecosystem of the region extremely fragile and sensitive to the response to external disturbances. The watershed has nurtured a unique natural environment and ecosystem that is home to numerous rare and endangered species. It is rich in water resources and constitutes an important lake ecosystem and water resource reserve in the arid zones of Central Asia.

Fig. 1

Overview map of the study area.

Data sources and preprocessing

Based on the GEE platform, this study utilizes MOD13A1, MOD09A1, and MOD11A2 products for 2000 to 2020 from MODIS data released by the United States Geological Survey (USGS). The RSEI was constructed through standardization processing and principal component analysis methods, with a spatial resolution of 500 m. Data sources are shown in Table 1. The data were cropped in ArcGIS 10.8 software after downloading and uniformly projected to the Universal Transverse Mercator (UTM) coordinate system at 43 degrees north latitude. The land use intensity (LUI) data were resampled to 500 m resolution from GLC-FCS30 land use classification data provided by the Institute of Aerospace Information and Innovation of the Chinese Academy of Sciences (https://data.casearth.cn/) through the nearest neighbor interpolation method. On this basis, various land use types are reassigned as follows: unused land is assigned a value of 1; construction land is assigned a value of 4; forest land, grassland, water area, and wetland are uniformly assigned a value of 2; and cultivated land is assigned a value of 3, thus obtaining the LUI data. Agricultural Water Use Efficiency (AWUE) data were calculated by ratios of Gross Primary Productivity (GPP) data from the GLASS dataset (http://www.glass.umd.edu/) to Evapotranspiration (ET) data from the MOD15A2 product in the MODIS dataset. The spatial resolution of both the RSEI and its 13 drivers was maintained at 500 m to ensure data consistency and validity of the analysis.

Table 1 Data source information.

RSEI construction

The RSEI is a technical index based on remote sensing technology, which is used to reflect the ecological status of the surface of the earth by analyzing the spectral characteristics of different bands in multispectral images. The index selects four indicators, namely greenness, humidity, heat and dryness, which can be intuitively judged by humans, to evaluate and study the surface ecological quality, so as to carry out effective ecological monitoring and management. The specific formulas for four of these indices are given in the text. After calculating and analyzing the above four indicators, Principal Component Analysis (PCA) is used to construct remote sensing RSEI. Principal component analysis simplifies the data representation by identifying the directions of maximum variance in the four indices and projects the four indices onto these new axes, which not only reduces the dimensionality of the data but also retains the most important features that affect the data distribution. Before conducting principal component analysis, the calculated NDVI, Wet, LST, and NDBSI also need to be regularized due to the inhomogeneity of their magnitudes, as a way to ensure that the weights of the indicators are not unbalanced, while abating the effects of temporal differences. Before conducting principal component analysis, the calculated NDVI, Wet, LST, and NDBSI also need to be standardized due to the inhomogeneity of their scales, as a way to ensure that the weights of the indicators are not unbalanced while abating the effects of time differences. The formula for standardizing each indicator is shown in Eq. (1):

$${\text{NI}}_{i}=\frac{\left({I}_{i}-{I}_{min}\right)}{\left({I}_{max}-{I}_{min}\right)}$$

(1)

here, NI_i is the standardized metric; I_i is referred to the value of the image element at image element i; I_max and I_min are the extreme values of the metric, respectively.

Subsequently, the first principal component analysis was performed on the four indicators to calculate the RSEI⁴⁰. The formula is shown in Eq. (2):

$$\text{RSEI}={\text{PCA}}_{1}\left[f\left(NDVI,WET,NDBSI,LST\right)\right]$$

(2)

Normalize the RSEI to values between 0 and 1.

Theil-Sen median and Mann–Kendall trend tests

The Theil-Sen Median is a nonparametric statistical technique for analyzing trends within time series data. The method is insensitive to outliers and measurement errors as well as being widely used in hydrological, meteorological and ecological applications^41,42. The formula for its calculation is:

$${\text{Z}}_{c}=\left\{\begin{array}{c}\frac{S-1}{\sqrt{Var(S)}},S>0\\ 0 , S=0\\ \frac{S+1}{\sqrt{Var(S)}},S<0\end{array}\right.$$

(3)

where

$$\text{S}=\sum_{i=1}^{n-1}\sum_{j=i+1}^{n}sgn\left({x}_{j}-{x}_{i}\right)$$

(4)

$$\text{sgn}\left(\theta \right)=\left\{\begin{array}{c}1 ,\theta >0\\ 0 , \theta =0\\ -1 ,\theta <0\end{array}\right.$$

(5)

$$\text{var}=\frac{1}{18}\left[n\left(n-1\right)\left(2n+5\right)-\sum_{t}{f}_{t}\left({f}_{t}-1\right)\left(2{f}_{t}+5\right)\right]$$

(6)

where x_j and x_i are the sequential date values, and n is the length of the date.

Geographic detector model

The Geographical Detector model as a type of analysis of spatial heterogeneity and reveals the spatial distribution pattern of the statistical method. It investigates the underlying mechanisms of various geographic phenomena mainly by probing the spatial relationships between independent and dependent variables. The geographic detector mainly includes four detectors⁴³. In this paper, we used divergence and factor detection and interaction detection to analyze the factors influencing the ecological index in the IBLB.

Differentiation and factor detection

In this study, we systematically assessed the mechanisms by which 13 categories of environmental drivers explain the spatial heterogeneity of the RSEI in the IBLB based on the spatial differentiation and factor detection methods of the model. By constructing a multilevel discretization model to implement spatial homogenization stratification for the independent variables, the variance decomposition algorithm was used to quantify the independent explanatory power (q statistic) of each factor to reveal the main controlling elements of spatial differentiation of ecological quality in the watershed and the intensity of their roles, the formulas are shown in Eqs. (7) and (8):

$$\text{q}=1-\frac{SSW}{SST}$$

(7)

$$\text{SSW}=\sum_{h=1}^{L}{N}_{h}{\sigma }_{h}^{2},SST=N{\sigma }^{2}$$

(8)

where the range of q values is [0,1]; h is the compartmentalization of the dependent or independent variable; σ² is the variance of the dependent variable; and SSW and SST are the total variance within the compartment and across the region, respectively.

Interaction detection

Detection of interactions is utilized to assess if two or more factors X, when acting together, enhance or diminish the explanatory power with respect to the dependent variable, the RSEI index Y, or if the effects of these factors on the RSEI index Y are independent of each other. In geographical research, interaction detection is very important because geographical phenomena are often affected by multiple factors, which may have mutually reinforcing or weakening relationships. The outcomes of the assessment can be put into five categories. See Supplementary Material 1 for a specific breakdown.

Geographically weighted regression model

GWR is one of the spatial profiling techniques that is used in a wide range of research involving spatial pattern analysis in fields such as geography, economics, environment, ecology and meteorology. The core strength of GWR is the construction of local regression equations for points on a spatial scale, which allows for deeper excavation of spatial variations and their drivers at specific scales. The method significantly improves the accuracy of the analysis by incorporating local effects of spatial objects⁴⁴.

Results and analysis

Results of PCA of the RSEI

By performing PCA on the four indicators (Table 2), we can find out the PC1 contributes 91.41% to RSEI on average, which exceeded 87%, while the sum of the other principal components (PC2, PC3, PC4) was less than 15%, and the changing trend was unstable, making it difficult to indicate ecological phenomena. Therefore, the first principal component can be used to create RSEI.

Table 2 Principal component analysis results.

Analysis of spatiotemporal variation of ecological quality in the IBLB

To better quantify the study area’s ecological quality, the hierarchy of ecological indices after principal component analysis was classified into five types. Where values of 0–0.2 are poor and indicated in red; values of 0.2–0.4 are range and indicated in orange; values of 0.4–0.6 are intermediate and indicated in yellow; values of 0.6–0.8 are good and indicated in light green; and values of 0.8–1 are outstanding and indicated in green. According to the classification, a RSEI change map of the IBLB from 2000 to 2020 was produced (Fig. 2). The results of the categorization of the different types of ecological indices for each year (Fig. 3) can be seen by counting and calculating the different types of ecological indices: from 2000 to 2020, the ecological environmental quality of the IBLB was mainly good and moderate, with the sum of the areas with good and moderate ecological environments in 2001, 2004, 2006, 2007, 2009, 2011, 2012 and 2020 accounting for more than 90% of the total area. The next highest percentage of poorer area was 47.1 and 32.7% in 2015 and 2005. The proportions of excellent and poor area are small, with the sum of both not exceeding 10% of the total area in most years. In terms of trends, there is an overall increasing trend in the proportion of excellent and good areas. The growth trend is particularly pronounced between 2015 and 2020, with the sum of the two areas accounting for 84.9% of the total in 2020 at its maximum. While the proportion of moderate, poor and poor areas showed an overall decreasing trend. The decline in the proportion of poor area is particularly pronounced, with its area below 5% from 2011 to 2020. The range of the mean value of RSEI from 2000 to 2020 is between 0.55 and 0.75, while the minimum value is 0.59 at the maximum value of 0.73. The overall data fluctuates greatly. A linear fitting analysis of the 21year data shows that the slope is 0.00161. It indicates that the overall RSEI in the area showed a slight upward trend during the study period and the quality of the ecosystem increased slightly. The above results show that during the period 2000–2020, there is an overall trend of improvement in the overall ecosystem quality of the IBLB in a favorable direction.

Fig. 2

RSEI map of IBLB, 2000–2020.

Fig. 3

Plot of RSEI area proportion change and RSEI mean value trend of Ili Balkhash Lake Basin from 2000 to 2020.

Analysis of temporal and spatial trends of ecological quality in the IBLB

Sen-Mk trend analysis of RSEI in the IBLB from 2000 to 2020 to derive the results of variations in ecosystem quality in the region (Fig. 4). Through analysis, it can be seen that the overall changes in the IBLB are mainly concentrated in slight improvements, accounting for 47.34%; the second largest change proportion is slight degradation, accounting for 38.81%; the area with no obvious changes accounts for 13.61%; the areas of obvious improvement and severe degradation account for a small proportion, which are 0.23 and 0.01% respectively. In summary, 47.57% of the watershed area has an improving ecological trend, while 38.82% of the area has a degrading ecological trend. Therefore, the trend of improvement was higher than the trend of degradation in the IBLB during the study period. The spatial distribution of RSEI trends in the study area is very fragmented, especially in the northern region of the basin, where ecological environment quality improvement and degradation trends are interlaced. In addition, the degraded areas are mainly concentrated around Lake Balkhash, in contrast to the areas suffering from improved ecological quality, which are mainly around the Kapchagia Reservoir and the areas of the Ili River Valley in the middle reaches of the Ili River and the upper reaches of the Ili River.

Fig. 4

Development trend of RSEI from 2000 to 2020.

Analysis of driving factors

For the purpose of exploring the factors affecting the change of RSEI in the IBLB during 2000–2020, with reference to the selection of factors affecting the RSEI by previous authors^45,46, and taking into account the actual situation of the study area and the accessibility of the data, the following 13 factors, all with a resolution of 500 m, were mainly selected for analysis in this study. Among them, the natural elements are: normalized difference vegetation index (NDVI), land surface temperature (LST), wetness index (WET), normalized difference built up and soil index (NDBSI), digital elevation model (DEM), slope gradient, aspect, net primary productivity (NPP), and evapotranspiration (ET); and the human elements are: land use intensity (LUI), population density (PD), nighttime light index (NLI), and agricultural water use efficiency (AWUE). ArcGIS 10.8 software was utilized to create a 5 km*5 km fishing net for sampling with 16,635 sampling points. Subsequently, in order to obtain the best detection results, the 13 factors were discretized and reclassified using the natural breakpoint method, and later imported into the Geodetector for analysis.

Single factor detection results analysis

The impact factors were detected respectively for the years 2000, 2005, 2010, 2015 and 2020 (Fig. 5). Higher q values indicate a greater degree of explanatory power of the driver to the causal variable, and all of the selected factors exerted a significant influence on the spatial differentiation of RSEI. Among these 13 impact factors, the mean explanatory power of RSEI is ranked in descending order: LST > NDVI > WET > NDBSI > DEM > ET > NPP > AWUE > slope gradient > LUI > aspect > NLI > PD. See Supplementary Material 3 for a specific breakdown.

Fig. 5

Explanatory power (q value) of each factor on RSEI.

Among the natural factors, LST has the greatest influence in the five years selected, with its q value ranging from 0.5 to 0.75. The year that had the greatest interpretive power was in 2010 which had a q value of 0.74501, and the least interpretive power was in 2020 which had a q value of 0.51565. After that, the factors with greater explanatory power are NDVI and WET, with the average q values of the five years being 0.31 and 0.25 respectively. Factors slope gradient and aspect have the smallest explanatory power. The average q values of these two factors over the past five years are 0.04 and 0.11 respectively. Except for the q value of slope gradient in 2010 which is 0.12, the other q values of the two factors are all less than 0.07. It indicates that in terms of natural elements, the RSEI in the IBLB is more strongly influenced by LST, NDVI, and WET, and less so by slope gradient and aspect. Among the human factors, LUI has demonstrated the strongest powers of explanation in both the years 2000 and 2015, in terms of q values 0.056 and 0.059, respectively; the mean value of q values for AWUE is the largest, at 0.045. In addition, the factor has the strongest explanatory power for the years 2005, 2010 and 2020 at q values of 0.0197, 0.0879 and 0.0442, respectively. While PD and NLI have less explanatory power with q value means: 0.0058 and 0.0044 respectively. It indicates that LUI and AWUE are more influential on RSEI in terms of human elements, while PD and NLI are less influential on RSEI.

Factor interaction detection analysis

Interaction detection analysis of the imaging factors in the study area leads to multifactorial effects on RSEI in the region (Fig. 6). As can be seen from the analysis, the two factor interactions all have greater interpretative power than the one factor interactions, and the interactions of the factors all exhibit two factor enhancement and nonlinear enhancement. Among them, 2010 and 2015 were mainly double factor enhancements, and 2000, 2005 and 2020 were mainly nonlinear enhancements, which indicates that the interaction of surface temperature as well as other factors had a significant effect on RSEI in the structure of the study area. During these five years, the interaction q values between LST and several other factors in the natural factors were large, especially the interaction q values of this factor in 2010 were all greater than 0.74, indicating that the interaction of surface temperature and other factors had a significant effect on the RSEI in the study area. In addition, the interaction q values between NDVI and the three factors DEM, NPP, and ET are also large, indicating that the interaction between NDVI and DEM, NPP, and ET has a relatively large impact on RSEI. Among the humanistic factors, the influence of AWUE and LUI interacting with other factors is slightly higher than the influence of NLI and PD interacting with other factors.

Fig. 6

RSEI interactive detection results of each influence factor from 2000 to 2020.

Analysis of drivers of localized spatial and temporal changes in RSEI affecting the ILBL

The geographic detector can well explore the heterogeneity of geographic space and identify the main factors affecting spatial distribution. However, the geographic detector cannot explain the spatial relationship between RSEI and various influencing factors. Therefore, in order to further explore the spatial local regression of each influence factor on RSEI, the GWR model was chosen to further analyze and discuss the strength and direction of the influence factors. On the basis of exploring the natural and human factors affecting the spatial and temporal changes of RSEI in the IBLB with the geographic detector, the top two influential factors among the natural factors: surface temperature and normalized vegetation index, and the top two influential factors among the human factors: land use intensity and agricultural water use efficiency, were selected for further analysis in a geographically weighted regression model. which were further analyzed in GWR model. First, in trying to avoid bias in the estimation results due to interactions, the covariance test of each influence factor was performed in ArcGIS 10.8 using the ordinary least squares (OLS) method. The results show that the Variance inflation factor (VIF) for all the influential factors which are all less than 7.5, which indicates that there is no covariance among the five factors and that meets the requirements of the study. See Supplementary Material 2 for a specific breakdown. Subsequently, the results of the OLS and GWR models are compared in (Table): the R² in the GWR model from 2000 to 2020 are all greater as compared to the R² in the OLS model, the R² Adjusted in the GWR model are all greater as compared to the R² Adjusted in the OLS model, meanwhile the value of AICc in the GWR model are all smaller than the OLS model, this result indicates that the goodness of fit of the GWR model are all than the OLS model. The analysis of the results of the local R² of GWR (Fig. 7) shows that the local R² ranges from 0.550809 to 0.917345 for the period 2000–2020, and the R² of all units is above 0.55, which is a high level of feasibility and confidence. The above results indicate that the GWR model is stronger in terms of the power of explaining the ecosystem quality and the four driving factors, and has a better explanatory effect, so the GWR model was selected for the analysis.

Fig. 7

Spatial distribution of local R² of the GWR model from 2000 to 2020.

The positive and negative signs in the regression factor represent the positive and negative effects of the influencing factor on RSEI, and the larger the absolute value for the regression coefficient, the greater the influence. The spatial distribution of out regression coefficients of four factor regression coefficients reclassified in ArcGIS 10.8 using the natural breakpoint method for 2000, 2005, 2010, 2015, and 2020. By analyzing the results of GWR, only LST has all negative effects on RSEI from 2000 to 2020, and the three factors NDVI, LUI, and AWUE have both positive and negative effects on RSEI. The specific impacts are as follows:

Analysis of local natural influences

The spatial variability in the impact of LST on RSEI in the IBLB is also high for natural factors. The regression coefficients are also mainly concentrated between −0.76 and −1.42, indicating that LST has a mainly negative effect in the Lake Balkhash basin (Fig. 8). In terms of spatial distribution, in 2000, the area of high LST impact on RSEI in the basin was mainly in the West side for Lake Balkhash, with the impact decreasing in a stepwise manner towards the eastern part of the basin. In 2005, LST has a large regional impact primarily on the city of Balkhash and the Lake Balkhash inlet in the Ili Delta, and gradually decreased outward from this center, while the impact on the Ili River Valley was relatively small. In 2010 and 2015, LST had a relatively large impact on the Ili River Valley and tapered off toward the northwest and northeast. In 2020, the LST has a greater impact on the northwest region of the basin and gradually decreases towards the southeast. The basin is located in a middle latitude arid area. According to the regression coefficient results, the LST of the basin is negatively correlated with the RSEI. The higher the surface temperature, the worse the RSEI environment will become. Thus there is a greater impact on the low elevation areas of the northwest region of the watershed and a lesser impact on the high elevation areas of the Ili Delta.

Fig. 8

The GWR model calculated the regression coefficients of LST from 2000 to 2020.

There was significant spatial variation in the effect of NDVI on RSEI, with regression coefficients ranging from −1.36 to 2.69 (Fig. 9). The spatial distribution of NDVI impacts on RSEI is mainly in the east west direction with a stepwise distribution. Except for the year 2000 when NDVI had mainly negative effects on the entire basin, in other years NDVI had a relatively large positive effect on the RSEI of the entire basin. The more significant impacts are concentrated in the Ili River Valley in 2000 and 2020. In 2000 and 2020, the NDVI in this region mainly showed a negative effect on RSEI. Positive effects were predominant in 2005, 2010 and 2015. According to the analysis of the development trend of RSEI in the basin, we know that the overall ecological environment of RSEI in the Ili River Valley is better, and the development trend is also in the direction of improvement. In conclusion, NDVI in the Ili River Valley can reflect the local ecology well, and the influence of NDVI on RSEI is large.

Fig. 9

The GWR model calculated the regression coefficients of NDVI from 2000 to 2020.

Analysis of local humanities influences

In terms of human factors, the impact of LUI on the RSEI of the watershed is also very significant, and the regression coefficients are mainly concentrated between −0.15 and 0.09 (Fig. 10). Only in 2005 did the impact of LUI on RSEI show all negative effects. In 2000, 2010, 2015, and 2020, the impact of LUI on RSEI was both positive and negative. The areas with the greatest impact of LUI in 2000 were in the west and east of the basin, and there is a negative impact in the west and a positive impact in the east. The impact gradually decreased from the east and west of the basin to the middle area. T Relatively large impact of LUI on RSEI in the Ili River Valley and Balkhash City region in 2005, but had a relatively small impact on other areas. The areas of high LUI impact on RSEI in 2010 were mainly in the Ili Valley and the northeastern portion of the basin. The areas where LUI had a large impact on RSEI in 2015 were mainly in the Ili River Valley, producing mainly negative impacts and positive impacts on the northwestern part of the basin. The area of greatest LUI impact on RSEI in 2020 is in the northern part of the watershed and decreases the further south you go. In summary, the areas where LUI has the greatest impact on RSEI are concentrated around Lake Balkhash and the Ili River Valley where population activity is more intense.

Fig. 10

The GWR model calculated the regression coefficients of LUI from 2000 to 2020.

The spatial variability in the effect of AWUE on RSEI in the IBLB is also more pronounced. The regression coefficients are mainly centered between −0.05 and 0.04, and the effect on RSEI is also both positive and negative (Fig. 11). In 2000, AWUE had a greater impact on RSEI in the Ili River Valley and the northeastern part of the basin. Of these, the RSEI is positive for the Ili Valley and negative for the northeastern portion of the watershed. In 2005, AWUE had a greater impact on the northern Ili Valley and northern Lake Balkhash with negative and positive impacts, respectively. In 2010, AWUE had a significant impact on the RSEI in the northwest and northeast of Lake Balkhash, with positive and negative impacts. In 2015, AWUE had a greater impact on the northwest of Lake Balkhash and the Ili River Valley, with negative and positive effects, respectively. AWUE in 2020 more negatively affects the northwestern region of Lake Balkhash Overall, the regions where AWUE has a large impact on RSEI in the IBLB over the period 2000 to 2020 are mainly in the Ili River Valley and the northwestern region of Lake Balkhash. In addition, AWUE has many positive impacts on the basin.

Fig. 11

The GWR model calculated the regression coefficients of AWUE from 2000 to 2020.

We reveal the spatial smoothing of the RSEI driver of IBLB based on GWR. Of the four factors screened by the covariance test, the GWR model showed stronger explanatory power compared to OLS. Among the natural factors, LST had a negative effect on RSEI across the domain, and the response was more significant in the low elevation northwest. While the spatial and temporal differentiation of NDVI action direction is obvious, the Ili River Valley is the core sensitive area. Among the human factors, LUI is negatively dominant in anthropogenic intensive areas, while AWUE shows a bidirectional effect. It was found that natural factors act with greater intensity, but human factors have higher spatial complexity, and the Ili River Valley, which has excellent ecological quality, is driven by both high intensity natural and human factors. The results not only make up for the limitations of geographic probes, but also provide a key scientific basis for differentiated ecological management in watersheds.

Discussion

Characteristics of RSEI spatial and temporal variations in the IBLB

This study is based on the GEE platform and integrates MODIS remote sensing data products to construct a RSEI for the IBLB. The spatial and temporal evolution of ecological quality in the study area from 2000 to 2020 was systematically revealed. Quantitative analysis showed that the ecological environment quality in the study area mainly showed moderate to good grades. This spatial distribution feature is consistent with the findings of Xia et al.¹¹ in the assessment of ecosystem service functions in arid zones. Notably, the Ili River Valley showed significant ecological quality improvement, which corroborates with the findings of Liu et al.⁴⁷ based on vegetation cover assessment. The area around Kapchagia Reservoir also shows a continuous improvement trend, which is closely related to the vegetation recovery driven by accelerated glacier ablation and increased runoff from the Ili River in the context of global warming⁴⁸. In terms of yearby year changes, the RSEI in the study area fluctuates greatly from year to year and does not show significant regularity, a feature that is similar to the ecological evolution pattern of typical arid zones such as the Urumqi River Basin²⁷. In this paper, by fusing multisource remote sensing big data to construct annual RSEI series and analyze the trend, the essential characteristics of its nonlinear changes are effectively revealed. The large interannual fluctuations in the RSEI may be due to the offsetting effect of physical geographic heterogeneity. Significant differences in ecosystem types in the upper, middle and lower reaches of the basin have led to spatial and temporal offsetting of ecological quality changes in different regions. Moreover, it is exacerbated by the double game of human activities and climate change. For example, the construction of the Kapchagai Reservoir has contributed to a certain extent to the restoration of the regional ecosystem. However, the downstream Ili River delta on the south side of Lake Balkhash has suffered irreversible damage to its ecosystem as a result of reservoir regulation⁴⁹, partially offsetting the positive ecological effects of upstream reservoir construction.

Discussion on the influencing factors of RSEI about the IBLB

Located in the eastern part of Kazakhstan, a region of the IBLB is characterized by an upward trend in the overall ecosystem quality. This is the same as the results of previous studies¹². As a typical transboundary watershed in the arid zone of Central Asia, it is remarkable to explore the influencing factors of RSEI in this region. After an deep analysis of the evolutionary trends of the RSEI within the IBLB from 2000 to 2020, this study found that the degradation of the basin’s ecological quality mainly occurred in the area around Lake Balkhash. Specifically, significant ecological changes have occurred in the Ili Delta area south of Lake Balkhash, manifested as a transformation from bare land to sparse vegetation. At the same time, the dominance of vegetation continues to decrease, leading to an intensification of forestland fragmentation in the region⁵⁰. In addition, the construction of the Kapchagia Reservoir and the increasing paddy field reclamation activities in the coastal areas, coupled with the excavation of the Almaty Canal, have jointly led to a reduction in the minimum ecological runoff and suitable ecological runoff in the lower reaches of the Ili River, thus having a significant negative impact on the ecological environment of Lake Balkhash⁴⁹. Improvements in the environmental quality are mainly limited in the area around the Kapchagia Reservoir, the middle reaches of the Ili River, and the Ili River Valley. The main driving factors for this phenomenon include the melting of ice and snow caused by climate change and the water storage effect of the Kapchagia Reservoir. These factors have jointly promoted the significant improvement of vegetation conditions in the aforesaid areas, thereby improving the overall ecological environment quality⁵¹. The increase in meltwater from ice and snow provides the necessary moisture for vegetation growth, while the water storage function of the reservoir ensures the stability of the hydrological cycle. The two work synergistically to effectively promote the recovery and improvement of the ecosystem.

We can tell from the results of the one way probing, that LST was identified as the most critical factor affecting RSEI, a finding that agrees with previous studies found in the arid zones of Central Asia¹¹. In further exploring the interactions between the factors, it was found that the interaction between LST and DEM had a greater relative influence over several years than the interactions of the other factors. The analysis results of the geographically weighted regression model also show that there is a significant negative correlation between LST and RSEI within the basin. That is, an elevated surface temperature causes a decrease in RSEI values, reflecting a deterioration in ecological quality. This is particularly significant for the low elevation areas in the northwestern part of the watershed, while it has relatively little impact on the high altitude Ili Delta area. This difference is mainly attributed to the close connection between the changes in RSEI and NDVI. NDVI as an effective indicator of regional ecological quality, and areas with higher values usually correspond to better RSEI values. The status of vegetation growth depends strongly on climatic factors. Whereas the study area is located in an arid zone with scarce precipitation, the impact of humidity on RSEI is not significant, while the role of LST on vegetation growth is crucial⁵². In the interaction factor detection analysis, the interactions between LST and DEM, and between DEM and NDVI showed higher influence. This phenomenon can be explained by the fact that the DEM has a significant effect on both LST and NDVI: as elevation increases, the temperature decreases, which in turn may affect the distribution and amount of vegetation, thus indirectly affecting the results of the RSEI assessment.

In terms of the influence of human factors, LUI and AWUE are important factors affecting RSEI in the IBLB¹². LUI is a direct indicator of the utilization of land resources by human activities, and there is a close correlation between LUI and ecological quality. In the IBLB, the acceleration of urbanization and industrialization, especially in areas with concentrated urban distribution within the basin, has led to a significant increase in land use intensity. This high intensity human activity has changed the natural landform and vegetation coverage pattern of the basin, causing increased soil erosion, reduced vegetation coverage, and degradation of ecosystem service functions, which in turn has had a profound impact on the Ecological environment. The pollution emissions and resource consumption brought about by industrialization have also put additional pressure on the ecological environment of the basin^53,54. In terms of agricultural development, agricultural activities in the basin are relatively developed. Due to the scarcity and uneven distribution of water resources in Central Asia, the actual AWUE becomes a key factor affecting ecological quality. The level of AWUE is directly related to the sustainability of agricultural production and the health of the ecosystem. Inefficient use of water resources will not only lead to the waste of water resources, but may also cause soil salinization, lowering of groundwater levels and other problems^55,56, which will impact the watershed RSEI.

In the IBLB, the influence of natural factors on RSEI was significantly stronger than that of anthropogenic factors. This finding is significantly different from the findings of Yangtze River Basin⁵⁷ and Dongting Lake Basin⁴⁶. The development of this phenomenon can be attributed to the unique characteristics of the natural environment of the basin and its relatively concentrated pattern of human activities. Much of the region is in arid or semiarid climatic zones, where precipitation is scarce and evaporation is intense, making moisture a key limiting factor for ecosystems. Climate change trends act directly and broadly on the ecological quality of the region as a whole. As the main water source in the watershed, the Ili River’s runoff distribution and seasonal changes are strongly constrained by precipitation, snowmelt, and evaporation. The total amount of surface water and groundwater and their spatial and temporal distribution are decisive for factors such as NDVI, LST and WET. Even if human activities modify local hydrological processes to some extent, the hydrological pattern of the whole region is still dominated by natural recharge and evaporation processes. In addition, Lake Balkhash, as a huge natural geographic unit, has an impact on the climate and ecology of the neighboring region through phenomena such as water balance, changes in water quality, and lake land winds. In contrast, human activities in the basin are relatively clustered, mainly in areas with favorable natural conditions such as river valleys, oases and water sources. While these areas have large localized ecological impacts, the extent of their impacts is still small compared to the vastness of the entire watershed. Compared with some highly economically developed regions such as the city of Wuhan⁵⁸, where the intensity of human activities is extremely high, the watershed may have a relatively low intensity of human activities as a whole, or the direct impacts of its activity patterns on large scale ecosystems are smaller than the fluctuations of natural factors. In the vast deserts, grasslands and mountainous regions, which are less affected by human activities, the ecological situation is still largely determined by natural factors.

Sustainable development recommendations

Based on the research in this paper, we propose strategies and recommendations for the protection of ecosystems and the environment in the region and Central Asia. Within the IBLB, it is proposed to implement a number of ecological repair and conserved measures, including vegetation restoration projects along the periphery of Lake Balkhash and the protection of areas of improved ecological quality such as the area around the Kapchagia Reservoir, the intermediate reaches of the Ili River, and the Ili River Valley, resulting in a sustained improvement in the quality of their ecosystem. At the same time, water resources management should be optimized, an integrated management strategy should be implemented, upstream and downstream water resources should be rationally allocated, and paddy field reclamation and large scale water conservancy projects should be strictly controlled to reduce the negative impact on the environment. Promotes sustainable agricultural models and controls overexploitation to improve the environmental friendliness and resource utilization efficiency of agricultural production. In Central Asia, cross border cooperation can be strengthened and a transboundary ecological and environmental monitoring and management mechanism can be established to optimize water resources management and implement integrated management strategies. To cope with climate change, adaptive management strategies should be formulated to enhance the carbon sequestration capacity of river basins, and land use intensity should be rationally planned to control the negative environmental impacts of urbanization and industrialization. At the same time, the efficiency of agricultural water use is being improved, promoting water saving irrigation techniques and sustainable agricultural models in order to ensure the sustainability of agricultural production and the health of the ecosystem. Through the establishment of a long term RSEI monitoring system and the strengthening of domestic and international scientific research cooperation, it is expected to promote the environmental protection and continuous development of the ecological environment in the IBLB and Central Asia, ensuring both regional ecological health and human welfare.

Shortcomings and prospects

We constructed RSEI on the GEE platform using relevant data from the MODIS dataset to comprehensively evaluate the ecological and environmental quality of the IBLB and deeply analyze its main influencing factors. The goal of this study is providing the scientific basis and strategic recommendations for ecological protection of the watershed and sustainable development of the Central Asian region. In the selection of influence factors, this paper ensures that the raster data resolution of the selected factors is the same as that of RSEI, which is 500m*500m, to enhance the accuracy of the detection results. Although we synthesized 13 factors for both natural and human aspects, some had to be excluded due to resolution consistency requirements. At the same time, the impact of policy factors and human activities is difficult to quantify, which to some extent limits the comprehensiveness of the analysis. In the process of RSEI construction, in order to reduce the influence of water bodies on the evaluation results, we used the normalized water body index to mask the water bodies and glaciers. However, considering the importance of water bodies in the ecosystems of the Central Asian drylands, this treatment may miss the dramatic effects of water bodies on the quality of ecosystems. In view of this, future research should be devoted to improving the existing or developing new systems for evaluating the quality of the environment in order to achieve a more comprehensive and objective evaluation of the environmental environment. Combining the time enhanced RSEI with the Luminous Index allows for a deeper and more comprehensive analysis of the ecological environment. The innovation and application of this method provide a valuable reference for establishing a more comprehensive ecological environment evaluation standard⁵⁹. Therefore, it is encouraged to integrate RSEI and night light and other multisource data in ecological environment evaluation research to promote further development in this field.

Conclusion

Under the premise of estimating ecological quality based on RSEI, this study used the geographic detector model and GWR models to explore the spatiotemporal change trends and influencing factors of RSEI in the IBLB from 2000 to 2020, and screened out the factors with greater influence for spatial impact analysis. The following are the main findings of the study:

• (1) The overall ecological environment quality is good and has a trend of continuous improvement, but the spatial heterogeneity is significant. From 2000 to 2020, the ecological environmental quality of the IBLB remained generally good and moderate, with a trend of significant improvement, and the proportion of high quality and good quality areas increased year by year. However, changes in ecological quality are characterized by spatial fragmentation. Improved areas are mainly concentrated in the Kapchagai reservoir and the middle and upper reaches of the Ili River. Degraded areas, on the other hand, are mainly distributed around Lake Balkhash, suggesting that the ecosystem around the lake may be more fragile.

• (2) Natural and human factors together drive ecological change with significant interactions. Natural and human factors together drive ecological changes with significant interactions. Geographic Detector Model analysis showed that LST and NDVI were the main natural factors affecting RSEI. LUI and AWUE were the key human factors. There were significant two way enhancement and nonlinear enhancement interactions among the factors, in which the interaction between LST and DEM was particularly prominent in the natural factor. The interaction between LUI and AWUE and other factors is more significant in the humanistic factor.

• (3) LST was the main limiting factor of RSEI, while other factors showed spatially differentiated effects. The GWR model further reveals that LST is significantly negatively correlated with RSEI and is the main limiting factor of RSEI, and its negative effect is spatially significantly differentiated. On the other hand, the effects of NDVI, LUI and AWUE on RSEI showed significant spatial variability, and the direction and intensity of the effects changed with time, with both positive and negative effects.

In summary, the overall ecological environment quality of the IBLB was good from 2000 to 2020, and showed a trend of continuous improvement. However, against the risk of possible future deterioration of the ecosystem, the protection of vegetation should be strengthened and water resources, especially agricultural water resources, should be rationally utilized in order to mitigate the adverse effects of human activities on the ecosystem.