Introduction

Global warming has extensive consequences for worldwide sustainable development1. This phenomenon not only elevates the occurrence of extreme weather events but also disturbs ecosystem equilibrium and presents a significant challenge to global food security and water management2.The Chinese government has established a “dual-carbon” strategic objective to attain carbon peaking by 2030 and carbon neutrality by 2050 in response to this global challenge. Terrestrial ecosystems are crucial for CO2absorption and climate change regulation, owing to the numerous carbon sequestration units present on the Earth’s surface3. Augmenting CS in terrestrial ecosystems can significantly diminish CO2levels, representing one of the most economically viable and ecologically sustainable methods to alleviate the greenhouse effect. Precise and prompt evaluation of CS is crucial for diminishing greenhouse gas concentrations, managing regional microclimates, alleviating global climate change, and preserving ecological equilibrium and sustainable development4. Numerous studies have established that carbon sequestration in vegetation and soils varies by land use type (LUT), and that regional CS changes when LUT are changed5. This is attributable to the disparate carbon densities among various LUT6. The alteration of land use and land cover(LULC) is frequently affected by economic and ecological factors, and the development of a meticulously crafted LULC optimization strategy can improve the carbon sequestration potential of the ecosystem, thereby augmenting the carbon storage of ecosystems7. Consequently, a comprehensive investigation of the effects of regional land use change (LUCC) on terrestrial ecosystem CS, alongside an assessment of the carbon sequestration potential of various land use types and the characteristics of alterations in the spatial distribution of CS, is essential for enhancing future terrestrial ecosystems. CS will assist in developing a more strategic national land use plan, expedite the achievement of the “dual-carbon” strategic objective, and aid in the alleviation of global climate change8.

Evaluating LUCC and forecasting land-use development trends are crucial for monitoring the dynamics of CS and for future projections. Commonly used LULC forecasting models include CLUE-S12, CA-Markov13, FLUS14, and PLUS15. The Patch-generating Land Use Simulation (PLUS) model is the most distinguished due to its ability to accurately capture the intricate mechanisms of LUCC during the simulation process16. Incorporating spatial planning policies to direct land use and cover change (LUCC) in the modeling process can yield a more optimal model of regional development17. Wei et al18. and Wang et al19. formulated specific development scenarios aligned with the study area’s requirements, employing the PLUS model to forecast future LULC distribution patterns. This model has garnered extensive acclaim from various researchers. The examination of LUCC across various regions and scales requires consideration of both the ecological and environmental carrying capacity determined by natural factors and the impact of national policies and human development needs. In the delicate ecological environment of inland Northwest China, the analysis of CS must consider water resource utilization, agricultural land requirements, and economic development limitations imposed by the region’s geographical position. Consequently, a distinctive scenario model customized for local development must be created through comprehensive and multidimensional viewpoints. Numerous existing studies have solely forecasted short-term alterations in LULC patterns based on historical land use development trends, typically projecting changes over the next five to ten years18,20,21,22. However, they have not assessed the long-term sustainability of the designated LULC development patterns for the study area, nor have they evaluated whether the effects of changes in land use patterns on carbon storage are transient or enduring. What scenario is advantageous for the sustained development of the Bosten Lake Basin? The study will create a scenario model aligned with the region’s characteristics and examine scenarios that support sustainable development in both the short and longterm.

Since 1850, carbon losses due to LUCC have constituted approximately two-thirds of global greenhouse gas emissions23. has accounted for about two thirds of global greenhouse gas emissions23. LUCC is a critical determinant influencing the carbon cycle in terrestrial ecosystems. These alterations will affect carbon sources and sinks, significantly influencing the regional and global carbon balance24. In recent years, researchers both domestically and internationally have undertaken numerous studies regarding the effects of regional land use change on CS9. They have conducted comprehensive research on various scales of LUCC and its resultant impacts on CS, encompassing global, national, provincial, and municipal levels10. Zhu G et al11. identified grassland degradation as the primary factor contributing to the decline in carbon storage within arid zones; Fu Y et al25. determined that urbanization-induced erosion of arable land in the Nanchang urban agglomeration was the principal cause of carbon storage reduction; and Setiawan O et al26. discovered that the carbon storage of the Forest Management Unit (FMU) of Ampang Plampang in West Nusa Tenggara diminished primarily due to the conversion of forests to alternative land use types. These studies consistently demonstrated that LUCC directly influences CS in a specific region. The forecasting of changes in CS due to land use is a contemporary concept employed by numerous researchers both domestically and internationally in carbon sequestration studies27,28. Presently, three principal methodologies are employed for estimating ecosystem CS: field surveys29, remote sensing30, and ecosystem modeling31. Field surveys offer high precision; however, attaining consistent and accurate observations over extensive regions can be difficult. Remote sensing models facilitate large -scale estimations of CS for specific species32. Ecosystem model simulations allow for quantitative analyses of CS across various spatial and temporal scales. The Integrated Valuation of Ecosystem Services and Tradeoffs (InVEST) model is the most widely used approach33,34. Piyathilake et al35. and Babbar et al36. employed the InVEST model to assess ecosystem services in Uva Province, Sri Lanka, and Sariska Tiger Reserve, India, respectively. In comparison to other models, the InVEST model for evaluating ecosystem services is distinguished by its straightforward input parameters, exceptional adaptability, rapid execution, and precision. It effectively elucidates the correlation between LULC and CS, enabling the assessment of CS in a particular region at a specific time37. Currently, numerous scholars have integrated the PLUS model with the InVEST model to assess and predict ecosystem services at different regional scales. Li et al15., Du et al38., and Wang et al39. employed the PLUS-InVEST model to evaluate and forecast CS in the Yiluo River Basin, the Karst Region in Southern China, and the Chengdu-Chongqing Urban Agglomeration, respectively, using land use and land cover data. The PLUS-InVEST model demonstrates exceptional efficacy in land use and land cover change simulation and carbon sequestration assessment.

The Bosten Lake Basin, positioned in the arid region of northwestern China, functions as a vital water conservation area for the “ecological water transfer” to the lower reaches of the Tarim River. It embraces various types of land use landscapes and possesses abundant ecological functions.Nonethe-less, owing to the vulnerability of its ecosystem, the water resources and oasis regions adjacent to the lake necessitate rigorous conservation and enhancement initiatives. The Fourteenth Five-Year Plan of the Bayingol Mongolian Autonomous Prefecture in Xinjiang, China mandates strict adherence to the “three red lines”: ecological protection, permanent basic farmland, and urban development boundaries (http://xjbz.gov.cn/xjbz/c101430/202312/6ac640c896dc450886325f099fa6f198.shtml). This study identifies spatial equilibrium among ecology, agriculture, and urban areas by establishing three developmental scenarios: Natural Development Scenario (NDS), Cropland Protection Scenario (CPS), and Ecological Protection Scenario (EPS). Designating the watershed as a constrained development zone and alleviating limitations on urban expansion. Forecasts of future LULC patterns for both short-term and long-term development scenarios are highly conducive to the sustainable development of the region and directly influence the future CS of the watershed40. Therefore, this paper utilizes the PLUS-InVEST model and LULC data from 1990 to 2020 to estimate and predict the CS and its spatial pattern characteristics in the Bosten Lake Basin for the years 1990–2020, 2030, and 2050. This study aims to accomplish the objectives outlined below: (1) Employ the PLUS model to simulate the spatio-temporal distribution patterns of LULC in the Bosten Lake Basin for short-term (2030) and long-term (2050) development across three scenarios: NDS, CPS, and EPS. (2) Employ the InVEST model to assess the spatio-temporal distribution characteristics of CS in the Bosten Lake Basin ecosystem under three different scenarios from 1990 to 2020, 2030, and 2050. (3) Analyze the spatio-temporal characteristics of the center of gravity shift in CS and the correlation within the spatial distribution in the Bosten Lake Basin. The study’s findings offer a scientific and rational foundation for land use planning to foster the sustainable development of the Bosten Lake Basin from a long-term perspective.

Study area and data sources

Study area

Located in the Bayingol Mongolian Autonomous Prefecture of Xinjiang, China, holds a distinctive position within the enclosed Yanqi Basin. An overview map of the study area, accompanied by actual photographs of the main land use types, is presented in Fig. 1. These photographs were captured by our team during their field investigations. It is bordered to the south by the Taklamakan Desert, to the north by the northern Tianshan Mountains, and to the east by the Kuruk Desert. The basin, due to its inland position and remoteness from the ocean, exhibits a temperate continental climate. The basin, especially the flat region, receives minimal rainfall and scant precipitation. The elevation of this watershed varies from 1037 to 4796 m, displaying a topography that is higher in the west and lower in the east. The upper sections of the watershed encompass the renowned Bayinbuluk Wetland Grassland, the second-largest grassland in China. The lower basin is abundant in water resources, with Bosten Lake being the largest inland freshwater lake in China. Its flat topography facilitates extensive agricultural development on the western shore and accelerates urban construction growth. The basin’s flora displays a spatial distribution pattern progressing from “desert-oasis-mountain” from east to west. The primary administrative region of the Bosten Lake Basin includes the four administrative units of Hejing, Bohu, Yanqi, and Heshuo counties, situated within the Bayingol Mongolian Autonomous Prefecture. This region functions as a crucial center for the economic advancement of Bayingol Mongolian Autonomous Prefecture and is instrumental in overseeing the ecological environment.

Fig. 1
figure 1

Schematic diagram of the study area and actual photographs of the main landscapes.

Full size image

Data sources

The LULC product data for the years 1990, 2000, 2010, and 2020 were sourced from the 30 m × 30 m CLCD dataset published by YANG et al41. from Wuhan University, China (https://zenodo.org/records/5816591). This dataset utilizes the Google Earth Engine (GEE) cloud computing platform to process Landsat imagery, develop spatio-temporal features, and employ a random forest classifier, resulting in an overall classification accuracy of 80%42,43. The Bosten Lake watershed boundary was determined using the hydrologic analysis tool of ArcGIS 10.8 software (https://www.esri.com/en-us/arcgis/products/arcgis-desktop/overview) and extracted from DEM data (http://www.gscloud.cn). The land use data was cropped using the extracted boundaries to obtain the land use data for the study area. The LUT were subsequently reclassified into seven categories according to LULC classification standards and the specific conditions of the study area: cropland, forest, grassland, water, unused land, built-up land, and wetland. A field visit was conducted to collect verification points for land use types, as depicted in Fig. 1, in order to validate the accuracy of the LULC product data classification.

Numerous factors influence LUCC43. According to related studies44, natural, economic, and social factors were selected as predictors of LULC. The natural factors include nine categories: digital elevation model (DEM), slope, slope orientation, mean annual temperature, mean annual precipitation, normalized difference vegetation index (NDVI), soil organic carbon, soil moisture content, and soil pH. The economic factors encompass three categories: population density, gross domestic product (GDP), and nocturnal illumination data. The social factors include three categories: proximity to rivers, proximity to railways, and proximity to roads. Data on rivers, railways, and roads from OpenStreetMap. Distances to rivers, roads, and railroads, derived from Euclidean distance analysis, were utilized as variables in forecasting future land use patterns. All data were resampled to a resolution of 30 m for computations. Table 1 summarizes the data sources for these drivers.

Table 1 Data sources of the drivers.
Full size table

Research methods

This study employed the PLUS-InVEST model to simulate future carbon sequestration and its alterations in terrestrial ecosystems within the Bosten Lake Basin. The PLUS model assessed LUCC across three distinct development scenarios, followed by the InVEST model predicting changes in CS under the same three scenarios for the years 2030 and 2050. Standard deviation ellipses (SDEs) were employed to examine the characteristics of center of gravity shifts in CS. Ultimately, spatial autocorrelation was employed to evaluate the spatial correlation of CS in the Bosten Lake Basin, while the PLUS-InVEST model integrates the benefits of simulating large-scale LULC distribution patterns and directly modeling CS in relation to with the LUCC, thereby offering a more holistic approach for forecasting CS variations in the basins of the Northwest Arid Region under prospective LULC scenarios. The comprehensive structure of the study is shown in Fig. 2.

Fig. 2
figure 2

Research framework.

Full size image

Analysis of LULC dynamics

The dynamics of a single land use type captures the rate of change in the area of a LUT in the study area over a certain time frame, focusing on the analysis of changes in individual LUT45. The formula is:

$$K = \frac{U_i - U_j}{U_i} \times \frac{1}{t}\times 100 \%$$
(1)

where K is the dynamics of a single land use type in the study area; Ui is the area of a land type at the beginning of the study; Uj is the area of a land type at the end of the study; t is the study period length.

The integrated land use type dynamics describes the overall rate of change in LULC across the region and can be employed to study regional difference in LULC dynamics. The formula is:

$$L=\frac{\sum\limits^{n}_{i=1} | U_{i}-U_{j} |}{\sum\limits^{n}_{i=1}U_{i}} \times \frac{1}{t} \times 100 \%$$
(2)

where L is the integrated land use type dynamics of LULC transformation in the study area during a study period; n is the number of land types.

Simulation of LULC with PLUS model

The PLUS was introduced by the HPSCIL@CUG laboratory team at China University of Geosciences (Wuhan) in 202017. This model utilizes a Cellular Automata (CA) methodology based on raster data. The PLUS model integrates a rule mining methodology derived from the Land Expansion Analysis Strategy (LEAS) with a Cellular Automata model founded on Random Seed (CARS). The LEAS module employs a random forest classification algorithm to investigate the correlation between the growth of each ___location category and various determinants. This analysis results in the estimation of the development probability for each place class and the contribution of drivers to the expansion of each place class during that timeframe. The future LUCC is simulated and predicted using the CARS module, which incorporates image element values and ___domain weights for specific LUT. This model provides enhanced simulation accuracy and landscape pattern indicators that closely correspond with actual landscapes, in comparison to other CA models. Additionally, the model integrates Markov chains and linear regression models to fulfill the requirement for predicting LULC quantities. This study employs these models to forecast land use and land cover quantities for specified years.

Neighborhood weight setting

Neighborhood weights indicate the probability of expansion for each LUT37.

$$\Omega^{t}_{i,k} = \frac{con (c^{t-1}_{i} = k)}{n \times n -1} \times w_{k}$$
(3)

where wk denotes the neighborhood weight parameter; n × n denotes the cellular unit;\(\:\:\text{c}\text{o}\text{n}({\text{c}}_{\text{i}}^{\text{t}-1}=\text{k})\) denotes the sum of grid cells occupied by the ground class at the end of the metacell iteration; \(\:{{\Omega\:}}_{\text{i},\text{k}}^{\text{t}}\)denotes the weight of the ___domain of the ground class in the spatial cell at time t. The neighborhood weights take values between [0, 1]. Its larger value indicates a higher expansion capacity40. According to the LUT transfer rule of Bosten Lake Basin, the neighbor weight parameter is obtained from the references, which are shown in Table 2.

Table 2 Neighborhood factors and parameters.
Full size table

Design of LULC scenarios

“14th Five-Year Plan” of the Bayingol Mongolian Autonomous Prefecture in Xinjiang, China, and by precisely identifying the spatial equilibrium among ecology, agriculture, and urban areas, alongside the research conducted by Chen J46, Yaermaimaiti A47, and others regarding the landscape configuration and environmental attributes of the Bosten Lake Basin, three categories of land-use transfer matrices have been formulated: the NDS, CPS, and EPS. The land type transfer cost matrix in the PLUS model indicates the feasibility of land use type transitions. A value of 0 denotes that switching is prohibited, whereas 1 indicates that switching is permitted. This research has delineated specific transfer cost matrices for three scenarios (Table 3).

Scenario 1: Natural Development Scenario (NDS). This scenario fails to consider the binding effects of planning policies on LUCC and growth. Rather, it adheres to historical evolutionary trends, perpetuating the developmental trajectory noted between 1990 and 2020.

Scenario 2: Cropland Preservation Scenario (CPS). This scenario seeks to reduce the likelihood of cropland transitioning to alternative land categories by 50%, while permitting the natural development of other land categories.

Scenario 3: Ecological Protection Scenario (EPS). This situation emphasizes the safeguarding of the ecological environment. It considers the significance of Bayinbuluk National Nature Reserve and Bosten Lake National Wetland Park as primary objectives for ecological resource conservation. In this scenario, the Markov model enhances the conversion probability of unused land to grassland, forest, and wetland by 30% each. Conversely, it diminishes the likelihood of grassland and wetland converting to water, developed land, and barren land by 40% each. The likelihood of converting cropland to built-up land and unused land is diminished by 30% each.

Table 3 Conversion cost matrix of land use type for each scenario.
Full size table

Notes: C, F, G, W, U, B and S denote cropland, forest, grassland, water, unused land, built-up land and wetland, respectively.

Accuracy verification

Land use simulation requires high accuracy and typically entails the categorization of various land use types. The Kappa coefficient, a precise metric for evaluating classification accuracy, satisfies the stringent accuracy requirements of the PLUS model for land use simulation and takes into account the classification of all categories, thereby enabling a more thorough assessment of the model’s performance across various land use types48. Therefore, the Kappa coefficient was employed to evaluate the precision of the model simulation as follows:

$$Kappa = \frac{P_{1}-P_{2}}{P_{3}-P_{2}}$$
(4)

where P1 denotes the percentage of simulated correct rasters; P3 denotes the percentage of simulated correct in the ideal state; and P2denotes the percentage of simulated correct in the random state. The Kappa value was between 0 and 1. The criteria for its determination were obtained from relevant studies49 (Table 4).

Table 4 Kappa coefficient simulation effect correspondence.
Full size table

The PLUS model was utilized to simulate the spatial distribution of LULC in the Bosten Lake Basin for the year 2020, and this simulation was compared with the actual LULC data from 2020. The Kappa statistic is measured at 0.86, and the overall accuracy is 92%. The PLUS model can more precisely represent the LUT in the Bosten Lake Basin. This study employed the PLUS model to simulate the LULC pattern of Bosten Lake Basin under three scenarios: NDS, CPS, and EPS, for both short-term development (2030) and long-term development (2050).

Carbon storage estimation with the InVEST model

Carbon storage accounting

The InVEST model was developed by Stanford University, the Nature Conservancy, the World Wildlife Fund, and various other organizations. It is a model for decision-making within ecosystems50. The InVEST model can be utilized for various factors including water retention, habitat quality, and carbon sequestration. The CS module aids in evaluating the CS of ecosystems. The carbon storage module comprises four fundamental carbon reservoirs: aboveground carbon storage (AGC), belowground carbon storage (BGC), soil organic carbon storage(SOC), and dead organic matter carbon storage(DOC)51. The calculation formula is:

$$C=C_{above}+C_{below}+C_{soil}+C_{dead}$$
(5)

where C denotes total CS; Cabove denotes above-ground CS in vegetation; Cbelow denotes below-ground CS in vegetation; Csoil denotes soil CS; and Cdead denotes dead organic matter CS.

Carbon density revising

The carbon density data of each LUT in the study area are mainly from existing related studies11,52. The carbon density data were chosen on the basis of local measured data in Xinjiang and integrated with relevant studies from Northwest China to derive preliminary carbon density data for the study area; however, the measurement data may contain some inaccuracies. Biomass and soil organic carbon density exhibited a negative correlation with precipitation and a positive correlation with temperature53. Techniques for rectifying carbon density data utilizing temperature and precipitation information are extensively employed in arid zones54. Therefore, this correction method was employed in this study to acquire adjusted localized carbon density data (Table 5).

Table 5 Modified carbon density of different land use types in the Bosten Lake Basin./(t·hm-2).
Full size table

Standard deviation ellipse (SDE)

The SDE was initially proposed by Lefever in 192655,57. The application of SDE facilitates a quantitative elucidation of centrality, discreteness, directionality, and spatial patterns, as well as other comprehensive characteristics of the spatial distribution of economic phenomena and geographic elements from a global spatial perspective56, utilizing parameters such as the range and center of the ellipse, the long axis, the short axis, and the azimuthal angle57. This study employed the SDE to examine the spatial centrality and directional distribution of CS within the study area. It primarily comprises four elements: the coordinates of the ellipse’s center of gravity, the lengths of the major and minor axes, and the angle of rotation. The calculation formula is58:

$$M(\bar{X},\bar{Y}) = \left| \frac{ \sum\limits^{n}_{i=1} w_{i}x_{i}}{\sum\limits^{n}_{i=1} w_{i}} ,\frac{ \sum\limits^{n}_{i=1} w_{i}y_{i}}{\sum\limits^{n}_{i=1} w_{i}} \right |$$
(6)
$$tan\ \theta = \left[ (\sum\limits^{n}_{i=1} w^{2}_{i}\tilde{x}^{2}_{i} - \sum\limits^{n}_{i=1} w^{2}_{i}\tilde{y}^{2}_{i} ) + \sqrt{ (\sum\limits^{n}_{i=1} w^{2}_{i}\tilde{x}^{2}_{i} - \sum\limits^{n}_{i=1} w^{2}_{i}\tilde{y}_{i} ^{2})^{2} + 4 ( \sum\limits^{n}_{i=1} w^{2}_{i}\tilde{x}^{2}_{i}\tilde{y}^{2}_{i} ) } \right] / 2 \sum\limits^{n}_{i=1} w^{2}_{i}\tilde{x}_{i}\tilde{y}_{i}$$
(7)
$$\sigma_{x} = \sqrt{ \frac{\sum\limits^{n}_{i=1}(w_{i}\tilde{x} \text{cos}\,\theta - w_{i}\tilde{y} \text{sin}\,\theta)^{2}} {\sum\limits^{n}_{i=1} w^{2}_{i}}}$$
(8)
$$\sigma_{y} = \sqrt{ \frac{\sum\limits^{n}_{i=1}(w_{i}\tilde{x} \text{sin}\,\theta - w_{i}\tilde{y} \text{cos}\,\theta)^{2}} {\sum\limits^{n}_{i=1} w^{2}_{i}}}$$
(9)

where M(\(\overline{x}\),\(\overline{y}\)) is the center of gravity coordinate; ɵ is the angle of rotation, the angle between the table due north and the long axis of clockwise rotation; xi, yi are the coordinates of the point; wi is the weight; \(\widetilde{x_{i}}\) ,\(\widetilde{y_{i}}\) are in turn the coordinate deviations of the points from the center of gravity; \(\sigma_{x}\) is the standard deviations along the X, and \(\sigma_{y}\) is the standard deviations along the Y.

Spatial autocorrelation analysis

Spatial autocorrelation is a collection of methods and techniques for spatial analysis. It is employed to assess the significant correlation between a phenomenon and the phenomena of adjacent units, elucidate the mechanisms of spatial interactions among the subjects of study, and quantify the extent of aggregation or disaggregation among the characteristics of spatial elements59. Spatial autocorrelation encompasses global spatial autocorrelation (GMI) and local spatial autocorrelation (LMI). GMI assesses the existence or nonexistence of spatial autocorrelation of a variable over the entire space60. The calculation formula is:

$$GMI = \frac{\sum\limits^{n}_{i=1} \sum\limits^{n}_{j=1}w_{ij}(x_{i}-\bar{x})(x_{j}-\bar{x})} {S^{2} \sum\limits^{n}_{i=1} \sum\limits^{n}_{j=1}w_{ij}}$$
(10)

where GMI takes the value in the range (-1, 1). When GMI > 0, it indicates positive spatial autocorrelation, and the closer it is to 1, the more closely related the cells are; when GMI < 0, it indicates that the region is significantly different from its neighbors; when GMI is close to 0, it indicates that the observations are independent of each other and randomly distributed.

LMI was designed to further explore the spatial connectivity patterns that distinguish local regions from their neighbors61. The calculation formula is:

$$LMI = \frac{ (x_{i}-\bar{x})} {S^{2}}\sum\limits^{n}_{j=1}w_{ij}(x_{j}-\bar{x})$$
(11)

where n is the total number of samples; wij are the spatial weights between i and j, respectively. LMI > 0 indicates clustering of similar values (high or low) around the regional unit; LMI < 0 indicates clustering of non-similar values in the region; LMI close to 0 indicates no spatial correlation between the region and neighboring regions.

Results

Analysis of LULC dynamics and multi-scenario forecasting

As shown in Spatial Distribution Fig. 3, grassland is the predominant land use type in the Bosten Lake Basin. It constitutes more than 50% of the total area of the region and is concentrated in the central and western sections of the basin. The second largest area consists of unused land, encompassing approximately 30% of the region, predominantly located around Lake Bosten. The percentage of water (approximately 6%) and cropland (around 5%) is relatively minimal, with water primarily sourced from Bosten Lake and glacial meltwater in the western section of the basin, and cropland predominantly located in the oasis along the western shore of Bosten Lake. The oasis region on the west bank of Lake Bosten is low-lying, abundant in water resources, and characterized by extensive agricultural and developed land. Developed land constitutes the smallest proportion of the watershed’s area, yet it exhibits the most rapid growth rate, rising from 0.01% in 1990 to 0.24% in 2020. Wetlands are concentrated in two regions: one in the upper watershed and the other in the lower watershed. The upper watershed predominantly features Bayinbruk wetlands, whereas the lower region is primarily characterized by wetlands along the western shore of Bosten Lake, comprising roughly 2% of the total area.

Fig. 3
figure 3

LULC pattern of Bosten Lake Basin from 2000 to 2020.

Full size image

In order to examine in detail how fast or slow each land use type is changing, land use dynamics are used as a measure, as shown in Table 6. Between 1990 and 2020, the cumulative dynamics of land use types in the Bosten Lake Basin peaked at 0.288% during the period from 2000 to 2010. The integrated dynamics of land use change over the past three decades is 0.201%. Analysis of the dynamics of land use changes across these three time periods reveals that built-up land consistently exhibits the most substantial area expansion, peaking at a growth rate of 50.64% during 2000–2010. This can be ascribed to the execution of the national “small towns, big strategy” post-2000, resulting in the persistent expansion of urban regions and the swift advancement of small-town development. Grassland and aquatic regions exhibit a declining trend in their transformation into wetlands, while the Bayinbruk wetland, situated in the upper basin, is progressively expanding annually. Additionally, the lakeshore wetland comprising reed marshes on the western shore of Lake Bosten is also experiencing growth.

Table 6 Attitude of LULC dynamics (%) in the Bosten Lake Basin by time period from 1990 to 2020.
Full size table

The LULC transfer matrix was employed to illustrate the conversion relationships among distinct land use categories. Figure 4 shows the transfer Sankey diagram. This diagram provided a lucid depiction of the inflows and outflows of each LUT across various time intervals. Over time, cropland, forest, built-up land, and wetlands have increased, whereas grassland and water have diminished. Unused land, conversely, remains largely unaltered. During the three specified intervals—1990 to 2000, 2000 to 2010, and 2010 to 2020—the cumulative area of LULC conversion was 3091.17 km², 2783.797 km², and 3174.289 km², respectively. The increase in cropland, forest, and wetland was mainly attributable to the conversion of grassland into these categories over a 30-year span. The converted areas were 462.128 km², 219.343 km², and 362.337 km², respectively. The built-up land primarily originated from cropland (3.470 km²) during 1990 to 2000 and transformed from grassland (30.491 km²) during 2000 to 2010 and again from grassland (22.261 km²) during 2010 to 2020. Water was predominantly transformed into unused land, leading to a reduction in its size. Simultaneously, unused land predominantly converted into grassland.

Fig. 4
figure 4

The Sankey map of land use type transfer in Bosten Lake Basin:1990–2020.

Full size image

The PLUS model has demonstrated an overall accuracy of 0.92 and a Kappa coefficient of 0.86, making it suitable for predicting the spatial distribution of LULC in the region (Fig. 5). Three distinct development scenarios were formulated: NDS, CPS, and EPS, with the results illustrated in Fig. 5. Additionally, a LULC change transfer matrix (Fig. 6) was developed to depict the transitions within each LULC category across scenarios from 2020 to 2030 and from 2020 to 2050. The spatial distribution reveals that grassland constitutes the predominant land use type in the Bosten Lake Basin, succeeded by unused land and water. Conversely, forest and built-up land represent the least significant proportion. Significant variations in LUT alterations are evident across different scenarios during the transition process.

Under NDS, there is a trend of wetland encroaching upon grassland, construction land encroaching upon cropland, an increase in forest land area, and a rapid expansion of built-up land, with a maximum increase of 25.512%. Under CPS, the cropland area is safeguarded, restricting the proliferation of built-up land, while there is a discernible trend of encroachment into wetlands, with cropland experiencing the most significant increase of 17.621%. Under EPS, forest land expanded significantly by 54.820%, primarily converted from grassland. Cropland on the western bank of Bosten Lake is fragmented and encircled by fallow land, constrained by suboptimal soil conditions. The expansion of cropland is most significant under CPS, yet only reaches 21.274%. To substantially increase cropland area while ensuring the sustainable development of urban areas, it is imperative to enhance the soil conditions of the surrounding unused land adjacent to the oasis on the western bank of Bosten Lake, thereby facilitating more effective cropland development.

Fig. 5
figure 5

LULC patterns in the Bosten Lake Basin under different future scenarios for 2030 and 2050.

Full size image
Fig. 6
figure 6

LUCC transfer matrix for different future scenarios in the Bosten Lake Basin in 2030 and 2050.

Full size image

Carbon storage modeling and dynamic change analysis

Based on the results of the land use projections, the carbon module of the InVEST 3.13 software (https://naturalcapitalproject.stanford.edu/software/invest) was employed to compute the CS values for the Bosten Lake Basin in the years 1990, 2000, 2010, and 2020, with the findings illustrated in Fig. 7a-d. The spatial distribution indicates that regions with elevated CS values predominantly consist of wetlands, followed by grasslands and croplands. In contrast, low CS values are predominantly located in urban areas and aquatic environments, including glaciers in the upper sections of the watershed. The spatial variations in CS are clearly linked to LULC patterns. The CS exhibited an upward trajectory from 1990 to 2020, commencing at a minimum of 290.307 Tg in 1990 and culminating at a maximum of 299.936 Tg in 2020, reflecting an increase of 9.629 Tg. A gradual increase is noted, accompanied by a deceleration in the growth rate over time. The highest increase in CS transpired between 1990 and 2000, totaling 5.162 Tg, whereas the lowest increase occurred between 2010 and 2020, amounting to 2.051 Tg. The distribution pattern of CS in the Bosten Lake Basin has exhibited relative stability over the past thirty years.

To elucidate the spatial variations in CS within the Bosten Lake Basin over the past thirty years, a raster subtraction operation was performed on the CS distribution maps from 1990 to 2020. Based on the research by Liu et al29., the spatial change values of CS were classified into three distinct categories. In this context, a change exceeding 5% was classified as an increase, a change less than − 5% was deemed a decrease, and a change within ± 5% was regarded as essentially unchanged CS. The resultant map (Fig. 7e) provides a thorough overview of CS alterations within the Bosten Lake Basin. Significantly, the majority of the basin’s CS has exhibited relative stability over the past thirty years. The stable CS region encompasses an area of 43,092.305 km², representing 97.84% of the total basin area.The region exhibiting CS increases constitutes 1.61% of the total basin area, predominantly concentrated in the Bayinbuluk wetland in the upper basin and the wetlands on the western bank of Bosten Lake. In contrast, regions of CS reduction comprise 0.55% of the overall basin area. The reduction zones are predominantly located in the glacial water region in the upper reaches and in downstream areas undergoing urban expansion.

Fig. 7
figure 7

Spatial distribution of carbon storage in Bosten Lake Basin:1990–2020.

Full size image

The InVEST model was employed to forecast carbon sequestration for three separate development scenarios in 2030 and 2050, with the projections illustrated in Fig. 8. The figure indicates that the pattern of CS in 2030 and 2050 is largely consistent with the distribution observed from 1990 to 2020. Grassland constitutes the principal carbon reservoir in the Bosten Lake Basin, accounting for 64–71% of the total CS. When analyzed alongside Fig. 11 h, it is apparent that irrespective of the development scenario model, the total CS within the watershed generally rises. Regarding short-term development (up to 2030) and long-term development (up to 2050), CS are projected to rise significantly under the EPS scenario, reaching 1.449 Tg and 5.833 Tg, respectively. Conversely, CS are anticipated to increase least under the CPS scenario, with increases of 0.302 Tg and 2.920 Tg, respectively. By 2030, CS in built-up land, water, and wetland are expected to rise significantly under the NDS scenario. In the CPS scenario, cropland, forest, and unused land are projected to experience the most significant increases in CS, while grassland is expected to see the largest increase in the EPS scenario.

Fig. 8
figure 8

Carbon storage projections for three scenarios in 2030 and 2050.

Full size image

The fluctuations in CS within the primary carbon pool of the Bosten Lake Basin from 1990 to 2020 are illustrated in Fig. 9. By 2030, under the NDS, CPS, and EPS scenarios, carbon stocks are projected to attain 301.053 Tg, 300.238 Tg, and 301.385 Tg, respectively. In 2050, CS are projected to be 305.345 Tg, 302.856 Tg, and 305.769 Tg, respectively. Significantly, CS remains consistently elevated under the EPS scenario throughout all years and scenarios. The NDS and CPS scenarios are presented subsequently. These indicate that grassland, the primary carbon reservoir in the watershed, has experienced the least reduction in area compared to the other two scenarios due to the implementation of ecological protection policies. Moreover, land development is constrained by the NDS, resulting in expansive regions of vegetative cover, including cropland and wetland. This leads to an increased overall CS. The carbon pool is primarily composed of soil organic carbon (SOC), which constitutes more than 93% of the total CS. BGC accounts for approximately 4.4%, whereas DOC possesses the minimal share of under 1%. AGC, SOC, and DOC demonstrate an increasing trend in all three scenarios, with the exception of BGC. DOC undergoes the most substantial increase owing to the extensive vegetative cover throughout the Bosten Lake Basin, which constitutes roughly 65% of the area’s total expanse. Given that the SOC of this vegetation surpasses that of AGC, BGC, and DOC, alterations in vegetative cover result in significant changes in SOC.

Fig. 9
figure 9

Carbon storage in the basic carbon pool of Bosten Lake Basin under different scenarios in 2030 and 2050.

Full size image

To visualize the spatial transformation of CS in the Bosten Lake Basin under different scenarios for 2030 and 2050, we utilized the raster phase subtraction method to examine the spatial patterns of CS alterations from 2020 to 2030 and from 2020 to 2050 (Fig. 10). The predominant portion of the study area displays negligible CS variations. In both the short-term (2020–2030) and long-term (2020–2050) outlooks, over 97% of the region sustains stable CS levels across all three scenarios. In the context of short-term development, it is apparent that by 2030, more than 98% of the CS remains stable across all scenarios. The greatest increase in CS is noted in an area of 307.310 km² under the EPS scenario, whereas the most significant decrease in CS is recorded over an area of 225.970 km² under the CPS scenario. By 2050, over 97% of the CS is projected to remain stable across all scenarios regarding long-term development. The maximum increase in CS is 736.744 km² under the EPS scenario, whereas the maximum decrease in CS is 398.658 km² under the CPS scenario. The findings indicate that although the majority of areas show negligible changes in CS, certain regions may be undergoing more pronounced increases or decreases in CS, contingent upon the development scenario.

Fig. 10
figure 10

Spatial changes in carbon storage under three scenarios in 2030 and 2050.

Full size image

From Fig. 11, the trend of CS changes in each land use type can be visualized. The total CS of cropland, forest land, built-up land, and wetland exhibits an upward trend, while the CS of grassland and watershed demonstrates a downward trend. The CS of unused land initially declines before rising, mirroring the area changes of each land use type in the study area. This indicates a strong correlation between the variations in CS across different land use types and the alterations in land use area. Forest and grassland areas experienced the most significant growth under EPS, predominantly located in the central and western regions of the basin, which favor the execution of ecological protection policies and enhance regional CS. Cropland saw the greatest increase under CPS, indicating that the protection of cropland on the west bank of Lake Bosten positively influenced CS in agricultural areas. Water, built-up land, and wetlands expanded the most under NDS, primarily situated in the eastern part of the basin, demonstrating that variations in CS across different land use types were closely linked to alterations in land use area. The changes in the total CS in the region are summarized in the following way: the overall CS in the region increases the most under the EPS.

Fig. 11
figure 11

Carbon storage for different land use types in different years (Tg).

Full size image

Carbon storage center of gravity shift and SDE shift analysis

From Fig. 12, it can be seen that the center of mass of CS in the Bosten Lake Basin relocate’d 35.5° to the west-northwest from 1990 to 2000, covering a distance of 3.297 km. Between 2000 and 2010, the center of mass was displaced 4.5° to the west-northwest, covering a distance of 4.954 km. The center of mass of CS shifted 37.3° east-southwest over a distance of 7.846 km from 2010 to 2020.The overall movement suggests that the center of mass of CS shifted westward alongside the citys rapid development, indicating that carbon sources in the study area relocated away from the city as it expanded. The SDE of CS in the Bosten Lake Basin exhibited a slight movement towards the northwest-southeast from 1990 to 2000, a slight shift towards the southwest-northeast from 2000 to 2010, a more pronounced movement towards the northwest-southeast from 2010 to 2020, and an overall increase in the northwest-southeast direction from 1990 to 2020.

Fig. 12
figure 12

Distribution of carbon storage center of gravity shifts and SDE shift directions:1990–2020.

Full size image

As Fig. 13 reveals, in both 2030 and 2050, the projected scenarios exhibit a uniform shift towards the east, signifying a consistent directional offset. This persistent trend is ascribed to the gradual diminution of grassland area in the upper section of the basin, leading to a concomitant decline in grassland carbon sequestration. In contrast, the expansion of cropland and wetland areas in the lower section of the basin leads to an increase in CS. Notwithstanding the expansion of built-up land, the carbon gain surpasses the carbon loss, resulting in a displacement of the center of mass of the CS towards the east. The diminished offset evident in the CPS and EPS scenarios indicates that if the principal sources of CS in the study area are adequately conserved, the overall CS will maintain equilibrium.

Fig. 13
figure 13

Distribution of future carbon storage center of gravity shift and SDE shift direction.

Full size image

Autocorrelation Analysis of carbon storage space

The GMI indicates whether the spatial data exhibit an aggregated or discrete trend, as well as the strength and significance of that trend. GMI analysis of the spatial distribution of CS in the Bosten Lake Basin under three distinct future development scenarios for 2030 and 2050 utilizing Geoda software. Figure 14 illustrates that the majority of the point data is concentrated in the first and third quadrants, while a minimal portion is located in the second and fourth quadrants. The Moran’s I values for the three scenarios in 2030 are 0.806, 0.802, and 0.808, while for the three scenarios in 2050, they are 0.827, 0.816, and 0.829.The aforementioned Moran’s I values exceed 0.8, indicating a robust positive spatial correlation. This signifies that the distribution pattern of CS in the study area is distinctly clustered and positively correlated, lacking a spatially random distribution. The spatial correlation is defined by the phenomenon where adjacent regions to high CS areas exhibit high CS, while adjacent regions to low CS areas display low CS. In these regions, zones of high CS engage with one another, while zones of low CS similarly interact among themselves. In terms of short-term and long-term development, the spatial correlation is strongest under EPS, while the CS spatial correlation gets stronger over time.

Fig. 14
figure 14

Moran scatters plot of global spatial autocorrelation analysis of carbon storage in Bosten Lake Basin.

Full size image

LMI analysis of CS in the Bosten Lake Basin was conducted under three distinct future development scenarios for 2030 and 2050, utilizing GeoDa 1.20 software (https://geodacenter.github.io/) (Fig. 15). The autocorrelation of the study area was classified into High-High Cluster, Low-Low Cluster, High-Low Cluster, Low-High Cluster, and Not Significant. Comparing the LULC classification map with the LMI distribution map reveals that the spatial distribution of CS values across the three scenarios for 2030 and 2050 is analogous: regions with elevated CS values predominantly occur in wetlands and forests, while regions with diminished CS values are primarily located in unused land and water bodies. The distribution of areas with high and low aggregation is minimal, primarily concentrated around water bodies. Consequently, the quantity of CS in the Bosten Lake Basin is primarily associated with the LUT.

Fig. 15
figure 15

Local spatial autocorrelation analysis of carbon storage in Bosten Lake Basin in 2030 and 2050.

Full size image

Discussion

Analysis of Carbon storage response to LUCC

The distribution pattern of LULC has a direct impact on the regional CS, rendering the precision of LULC classification a critical determinant in the reliability of CS estimation outcomes. Acquiring long-term, high-resolution remote sensing images free from cloud cover and consistent with the study’s temporal parameters is difficult, hindering our ability to perform independent categorization. Consequently, we depended on the CLCD product data for CS predictions. To ensure result accuracy, data from 92 field survey points were utilized in this study to validate the 2020 CLCD data, as illustrated in Fig. 1. The verification points represent the distribution of the field survey locations, while the actual map depicts the current land use status in the Bosten Lake Basin, achieving a 98% classification accuracy. He et al43., Li et al15., and Wang et al62. have all conducted pertinent studies utilizing this LULC classification product, yielding highly reliable research results. This product data is therefore highly reliable and appropriate as the foundational data for this study.

The carbon sequestration capacity of various land use types (LUT) within the study area is not exclusively dictated by their carbon storage potential; it is also affected by the dimensions of each individual LUT. The carbon sequestration capacity differs among various land use types. Tang et al2., Beroho et al13., and Chen et al63. have all concluded that wetlands possess the highest carbon sequestration capacity among land use types in the study area. This study, utilizing the carbon density table, also established that wetland is the land use type with the most substantial carbon sequestration capacity. Nonetheless, the limited distribution of wetlands and their relatively low total carbon stock render them the least significant carbon reservoir in the region. Despite grasslands lacking the highest CS capacity in the region, they represent the largest carbon reservoir, comprising over half of the region’s area and thus possessing the highest total sequestered carbon.

The alteration in the center of gravity of CS is not exclusively determined by the expansion of urbanized land; it is also associated with the increase or decrease in the extent of the region’s principal carbon reservoirs and the high carbon density LUT. In this study area, the center of mass of CS does not approach the built-up region; rather, it recedes from it, despite the significant expansion of the built-up area projected for both 2030 and 2050. This phenomenon occurs because carbon depletion is inferior to carbon gain, attributed to the limited area of developed land and the expanding wetlands, which sequester the most carbon and are situated near the developed areas. The expanse of grassland, the predominant carbon reservoir in the western region of the watershed and distant from urban development, is diminishing. The alteration in land use and land cover area directly influences the change in CS, ultimately leading to a shift in the center of gravity of CS to the east. In both short-term and long-term development scenarios, the increase in CS is most pronounced under the EPS scenario. This indicates that the EPS is an appropriate land use scenario model for the sustainable development of the Bosten Lake Basin.

Alterations in land use will affect regional CS; therefore, pertinent initiatives can be implemented to regulate and modify land use within the watershed: (1) Rationalize the arrangement of oasis cropland within the watershed to prevent over-cultivation and degradation, while enhancing soil CS capacity through the improvement of cropland quality and the implementation of a rotational fallow system. Embracing the principle of green development can significantly enhance CS, mitigate global climate change, and foster the sustainable development of the Bosten Lake Basin. (2) The ecological environment in the arid zone is highly fragile, characterized by a scarcity of ecological land, including wetlands and forests. Wetlands possess the highest carbon sequestration capacity within the watershed; therefore, the conservation of the Bayinbuluk Wetland Landscape should be prioritized. The Bayinbuluk Wetland should be designated as wetlands of international and national importance (http://www.xjbz.gov.cn/xjbz/jrbz/202409/dfce6fddbcda498085e42700cbb22a2d.shtml). (3) It is found that the construction land in the oasis on the western shore of Lake Bosten primarily built-up land from the conversion of grassland and cropland, resulting in a decrease in the region’s total CS. Nevertheless, owing to the demands of urban development in the Yanqi Basin (https://www.xjyq.gov.cn/), it is imperative to moderately develop construction land, prioritize land conservation and intensification, enhance land use efficiency, and enforce rigorous management of building land.

Analysis of factors influencing changes in carbon storage

In this study, the InVEST model was utilized to assess the overall carbon sequestration capacity of the Bosten Lake Basin and the variations in CS resulting from different LUT from 1990 to 2050. The results demonstrate that the carbon sequestration capacity of the Bosten Lake Basin is projected to rise over the next 30 years, with the greatest sequestration potential noted under the EPS scenario. The augmentation in carbon sequestration is chiefly ascribed to wetland conservation, forest rehabilitation, and the recession of glaciers into unused land. The transformation of grasslands into unused land is the principal factor contributing to the decline in CS (Fig. 16). Consequently, it is essential to enact and advocate for ecological conservation initiatives. The execution and advocacy of significant forest and grassland conservation initiatives, including the “Three North Protective Forests” project and the Natural Forest Resources Protection Project, have led to the ongoing enlargement of forested regions, progressive enhancements in tree quality, and substantial rises in the carbon density of both planted and natural forests’ carbon density64. Grassland degradation is chiefly ascribed to the synergistic impacts of climate change and anthropogenic activities. Despite the augmented precipitation in the grassland regions of the upper Bosten Lake Basin during the growing season, heightened evapotranspiration resulting from elevated temperatures contributes to intensified drought conditions in mountainous regions65. During the growing season, the incidence of drought surpasses 40%, which has a direct impact on the healthy development of grasslands66. Grazing activities substantially contribute to the degradation of grasslands in Xinjiang. In the long term, these grazing practices have resulted in a decrease in CS. Increased grazing intensity results in varying degrees of grassland degradation, consistently causing a decline in aboveground carbon stock, soil carbon stock, and total CS67.

To enhance and stabilize carbon sinks in the Bosten Lake basin, it is essential to fortify the protection and restoration of the wetlands adjacent to Bosten Lake and the Bayinbruk wetlands, expand the wetland area, augment the coverage of wetland vegetation, and improve wetland management. Secondly, it is imperative to advance the development of ecological agriculture and forestry, enhance ecological agriculture in the Bosten Lake oasis, and actively implement afforestation and forest conservation in the central and western regions of the basin to augment the carbon sink capacity of the ecosystem. Simultaneously, it is imperative to reinforce the management of grazing practices in extensive grasslands within the basin, implement sustainable livestock husbandry, maintain a balance in livestock populations, achieve equilibrium between grass and livestock, and augment the capacity of grasslands to sequester carbon.

Fig. 16
figure 16

Carbon storage enhancement measures.

Full size image

Limitations and applicability analysis

This study utilized previous research and official statistics to produce preliminary carbon density data. Employing a correction formula established by Zhou et al54., the initial data were subsequently modified to incorporate the climate factor, a crucial determinant of CS, yielding region-specific carbon density data that have been validated for scientific accuracy and reliability53. Subsequently, the InVEST model was employed to assess the CS and its spatial-temporal patterns of evolution in the Bosten Lake Basin between 1990 and 2050. To maintain the relevance of the corrected carbon density in the Bosten Lake Basin and improve the reliability of the assessment results, it is imperative to perform an accuracy verification analysis of the adjusted carbon density findings through pertinent studies. Zhang et al68. assessed soil carbon density and carbon stock in western China utilizing the first and second soil census data, revealing minimal discrepancies with the corrected carbon density presented in this study (refer to Table 7). This suggests that the corrected carbon density possesses a high level of reliability and accurately represents the carbon stock conditions in the study area. This study’s results can establish a scientific foundation for improving carbon sink functionality and refining management strategies for terrestrial ecosystems in arid watersheds. It is essential to acknowledge that the InVEST model solely evaluates the effect of LUT on CS, neglecting additional influencing factors such as photosynthetic rate, soil microbial activity, vegetation type, and anthropogenic activities. Moreover, the oversimplified depiction of the carbon cycle may constrain the precision of CS estimations, resulting in ambiguities in the spatial distribution of carbon sequestration patterns. Nonetheless, the application of the InVEST model provides a pragmatic method for assessing regional CS levels.

Table 7 Carbon density comparison.
Full size table

Conclusions

This study established three scenario development models and their corresponding parameters based on the regional characteristics of the Bosten Lake Basin and the requirements for regional development, namely NDS, CPS, and EPS. The PLUS model was employed to forecast LULC patterns for 2030 and 2050, while the InVEST model was utilized to predict and evaluate the spatial distribution and alterations of CS within the Bosten Lake Basin across various timeframes. The research additionally examined alterations in the position of the center of gravity of the CS. Spatial autocorrelation analysis was ultimately performed utilizing Geoda. Following these thorough evaluations, the subsequent conclusions were drawn:

(1) In the Bosten Lake Basin, grassland has been the predominant land use type from 1990 to 2020, covering more than 50% of the basin area. This is succeeded by unused land and water. During the study period, cropland, forest, built-up land, and wetland have exhibited an upward trend, whereas grassland, water, and unused land have shown a declining trend. In future developmental trends, grassland will continue to be a significant land use type in the Bosten Lake Basin. The most rapid growth of built-up land occurs under the NDS, the expansion of cropland is expedited under the CPS, and the conservation of water and forest is most significant under the EPS.

(2) The CS in the Bosten Lake Basin has demonstrated an annual upward trend from 1990 to 2020, resulting in a total increase of 9.629 Tg. Elevated CS values are predominantly found in wetland regions, while diminished CS values are primarily located in aquatic areas. The region with rising CS surpasses that with declining CS. Grassland constitutes the largest CS in the Bosten Lake Basin, and the reduction in grassland area is concomitantly diminishing its carbon sequestration capacity. Nonetheless, under the EPS condition, the decline of grassland was effectively regulated. The CS will rise in the future across three scenarios, with the most significant increase occurring in EPS.

(3) The CS center of gravity for carbon storage in the Bosten Lake Basin shifted westward from 1990 to 2020 and then eastward from 2020 to 2050. This alteration was influenced by changes in the area of the largest carbon pool and the extent of high carbon density land use types, as well as urban development in the region. The distribution of CS in the study area exhibited distinct clustering and a positive correlation, with elevated CS values predominantly located in wetland and forest regions, whereas diminished CS values were concentrated in barren land and aquatic zones. The distribution of CS in the study area exhibited distinct clustering and a positive correlation, with elevated CS values predominantly found in wetland and forest regions, whereas diminished CS values were concentrated in barren land and aquatic zones.