Introduction

Ecosystem services (Es) encompass the environmental conditions and benefits provided by ecosystems1, which are essential for human survival and development2,3. These services include both direct material resources and indirect intangible benefits4, reflecting the economic value and overall well-being that ecosystems deliver to humanity5,6,7,8. Land use change, driven by human interactions with and adaptations to ecosystems9,10,11, serves as a direct catalyst for alterations in ecosystem service value12,13,14. Such changes not only mirror the evolving demands that human societies place on the ecological environment but also highlight the resilience and adaptive capacity of ecosystems in meeting these demands15,16. The concept of the “Production-living-ecological” aggregates the primary functions of land, comprising production space, living space, and ecological space17,18,19. Production space primarily yields agricultural and industrial products20, living space satisfies human needs for housing21,22, consumption, and recreation, and ecological space provides critical ecological services and products23,24. Changes in land use impact the spatial patterns and functions of these three dimensions25,26, while the distinct characteristics of the “Production-living-ecological” facilitate the balanced development of production, habitation, and ecological conservation27,28. In contemporary interactions between human society and ecosystems29,30,31, the expansion of production and living spaces often compresses ecological space32, thereby disrupting ecosystem structure and function and diminishing the value of ecosystem services—a decline that has direct ramifications for human well-being33,34. Consequently, examining the relationship between land use change and ecosystem service value from the perspective of the “Production-living-ecological” is crucial for formulating scientifically grounded and rational land management policies that foster coordinated economic, social, and ecological development21,35.

Research on ESV has progressively evolved, establishing a mature theoretical and methodological framework36. In 1997, Daily2 systematically elaborated on ecosystem services and related concepts, laying the theoretical foundation for subsequent research. In the same year, Costanza et al.5 utilized various valuation methodologies to estimate the annual global value of ecosystem services. The introduction of the Millennium Ecosystem Assessment in 2003 provided a comprehensive classification system for ecosystem services that has since been widely adopted. Further refinements have been made by researchers such as Xie et al.15, who adjusted the equivalence factor coefficients to develop an assessment system tailored to China’s unique ecological characteristics. Building upon these frameworks, numerous studies have investigated the dynamics between land use change and ecosystem service value37,38. Methodologically, researchers typically employ land transfer matrices39,40,41, spatial analyses42,43,44, sensitivity coefficient analyses45, scenario simulations46,47, and valuation equivalence methods based on land use types to explore these relationships48,49,50. Moreover, research efforts have predominantly targeted provinces, watersheds, and ecologically significant areas, utilizing time-series and spatial analyses to examine the temporal and spatial variations in ESV51,52,53. Despite notable progress, further research is necessary to reinforce studies that align with regional land spatial planning policies.

The Anning River Basin, a pivotal region in China’s poverty alleviation initiatives, exemplifies the conflict between economic development and ecological conservation. Economic development in the basin inevitably impacts the ecological environment, underscoring the need for strategic land planning and utilization to safeguard ecosystems and promote a harmonious integration of economic and ecological interests. Located in the upper reaches of the Yangtze River, the Anning River Basin is recognized for its critical role in water conservation and biodiversity protection within the Hengduan Mountains. However, the basin’s ecological environment is exceedingly fragile, characterized by a complex and sensitive ecosystem structure that faces significant ecological challenges. As a typical dry-hot valley region, it is subject to arid and high-temperature conditions, leading to water scarcity, soil desiccation, and limited vegetation cover, all of which reduce the ecosystem’s regenerative capacity. Furthermore, human-induced degradation—through urban expansion and resource exploitation—exacerbates biodiversity loss and undermines the integrity and stability of ecological networks.In response to these challenges, this study adopts an integrated approach, drawing on the “Production-living-ecological” theory to examine ecological protection and regional development. The research employs Geo-information TuPu to analyze land use transformations in the Anning River Basin from 1985 to 2023 and quantifies ecosystem service value through revised equivalence factor tables. An improved cross-sensitivity coefficient is introduced to evaluate the impacts of various land use transformations on ecosystem service value, while geographic detectors are utilized to elucidate the effects of diverse influencing factors. Overall, this study aims to offer theoretical guidance for developing sustainable land management policies in the Anning River Basin, holding significant implications for enhancing ecological security and promoting sustainable regional development.

Materials and methods

Study area

The Anning River (Fig. 1) originates from Beipusagang in Mian County, Liangshan Yi Autonomous Prefecture, Sichuan Province. Flowing southward, it acquires the name Anning River upon merging with the Beijingsha River, which approaches from the northwest near Tuo Wu. As the largest tributary on the left bank of the lower Yalong River, the Anning River basin encompasses an area of approximately 11,053.49 km2, extends 326 km in length, and exhibits an average annual discharge of 231 m3·s⁻1. The river courses through Mian County, Xichang City, Dechang County, and Mi Yi County in Panzhihua City. The basin is characterized by a subtropical monsoon climate, with an average annual temperature exceeding 14 °C and an average annual precipitation of 1240 mm54. Serving as a critical ecological barrier in the upper reaches of the Yangtze River, the valley alluvial plain within the basin is recognized as the second largest plain in Sichuan Province after the Chengdu Plain. The region benefits from favorable geomorphological and climatic conditions, abundant mineral resources, and concentrated clean energy, all of which have significantly contributed to the rapid economic development of the Panxi urban agglomeration and established the area as a key driver of regional growth.

Fig. 1
Fig. 1
Full size image

The Basin of Anning River. Figure was generated using ArcGIS version 10.7 (https://desktop.arcgis.com/zh-cn/desktop/index.html), and the map review number is GS (2024) 0650 (https://cloudcenter.tianditu.gov.cn/administrativeDivision/).

Data sources

The Lu data employed in this study are derived from the annual land cover products developed by the research team led by Yang Jie and Huang Xin at Wuhan University. This dataset spans from 1985 to 2023 and features a spatial resolution of 30 meters (https://doi.org/10.5281/zenodo.12779975). These data—including nighttime light imagery, gross domestic product, population density, road network density, Digital Elevation Model (DEM), slope, annual average precipitation, and soil erosion levels—were obtained from the Resource and Environmental Science Data Center of the Chinese Academy of Sciences (https://www.resdc.cn/). The land use dataset, with a resolution of 30 m × 30 m, served as the reference for resampling, and all other datasets were adjusted to match this spatial resolution. Statistical information was obtained from the Sichuan Statistical Yearbook (https://tjj.sc.gov.cn/scstjj/c105855/nj.shtml), and grain prices were sourced from the National Agricultural Product Revenue Compilation. The annual land cover products encompass nine land types: cultivated land, shrub land, forest land, grassland, water bodies, glaciers and permanent snow, impermeable surfaces, bare land, and wetlands. In accordance with the LUCC Classification System established by the Chinese Academy of Sciences, these nine land types were reclassified into six categories: cultivated land, forest land, water bodies, construction land, grassland, and unused land11. Consistent with the principles of the PLES theory and employing a "bottom-up, functional grading" approach23, cultivated land is designated as P–E land; forest land and grassland are classified as E–P land; water bodies and unused land are categorized as E land; and construction land is identified as living production land (P–L land)11.

Research methods

Analysis of land use dynamics

The rate of change in Lu exhibited regional variability. The dynamic degree model of Lu was used to calculate the changes in the PLES land types in the Anning River Basin from 1985 to 2023, as follows55:

\(\text{K}=\frac{{\text{U}}_{2}-{\text{U}}_{1}}{{\text{U}}_{1}}\times \frac{1}{{\text{T}}_{2}-{\text{T}}_{1}}\times 100\text{\%}\) (1)

Where K represented the dynamic degree of a single PLES land type; T2-T1 was the time interval; U1 was the initial area of a specific PLES land type in the study area, U2 was the final area of the same land type at time T2; and U2-U1 was the change in area of that land type.

Land use transfer and fluctuation mapping analysis

Geo-information TuPu theory was widely applied in the field of LuT, visually displaying information on LuT56,57. This paper constructed a transfer map of the PLES land types to reflect their change characteristics. The mapping principle formula was:

$$\text{W}=10\times \text{A}+\text{B}$$
(2)

where W was the raster-calculated map code, A was the initial coding of the PLES land types at the beginning of the study, and B was the coding of the PLES land types at the end of the study. Based on the overlay analysis of the initial and final data of the study period, the corresponding fluctuation maps were constructed according to the new and shrinking characteristics of the PLES land types during the transformation process.

Ecosystem service value assessment

This study adopts the unit-area ESV table proposed by Xie et al.15 and modifies it to accurately reflect the land use conditions within the research area. By establishing correspondences between arable land, forest land, grassland, water bodies, and uncultivated land and the secondary classifications of cultivated fields, forests, grasslands, water bodies, and barren land in the ESV framework, a preliminary unit-area equivalent factor table was developed for the study area. Subsequently, further calculations and statistical analyses were conducted in accordance with the PLES land use classification, yielding a unit-area equivalent factor table that aligns with the PLES criteria for the study area. Utilizing data from 1985 to 2023—including unit prices, yields per hectare, and sowing areas of key national grain crops (wheat, soybeans, and corn)—the economic value of the average grain yield was calculated to be 8,648.41 yuan per hectare. Based on the principle that the economic value of a standard equivalent ESV factor corresponds to one-seventh of the economic value of the average grain yield, the equivalent coefficient of the standard ecosystem service value factor for the study area was determined to be 1,235.49 yuan per hectare. The formula is:

$${E}_{j}=\frac{1}{7}\sum_{i=1}^{n}\frac{{m}_{i}{p}_{i}{q}_{i}}{M}$$
(3)

where Ej represented the economic value provided by the cultivated land system per unit area; i was the type of grain crop; mi was the total sown area of the i-th grain crop; pi was the average price of the i-th grain crop; qi was the average yield per unit area of the i-th grain crop; and M was the total sown area of all grain crops.

Through revised calculations, the unit area ESV table for the PLES land types in the Anning River Basin was derived (Table 1). This study calculated the ESV of the natural environment, excluding P-L land from the evaluation.

Table 1 ESV Coefficient Table for the Anning River Basin.
Full size table

Spatial analysis

Semi-variogram function

The Anning River Basin was partitioned into grid units measuring 1 km × 1 km and 500 m × 500 m, with the ESV of each unit considered as a regional random variable. A quantitative analysis across these scales was conducted to elucidate variations in ecosystem service values among different grid units, thereby assessing their spatial heterogeneity., as detailed below58,59:

$$\gamma (y)=\frac{1}{2N(y)}\sum_{i=1}^{N(y)}{\left[z({x}_{i})-z({x}_{i}+y)\right]}^{2}$$
(4)

In the given equation, γ(y) was the semi-variogram value of ESV, indicating the variability between grids at distance; N(y) is the number of point pairs at distance y; Z(xi) and Z(xi+y) were the ESV at spatial locations xi and xi+y respectively. Important parameters of the semi-variogram function include: C0 nugget value, C0+C sill value, C0/(C0+C) nugget-to-sill ratio, and A range, where the nugget value reflected the random fluctuations and uncertainties of ESV at a microscopic scale; the sill value reveals the overall variability of ESV at a macroscopic scale; the nugget-to-sill ratio quantified the randomness of spatial distribution and reflected the correlation of ESV between regions. A low value (<0.25) indicated strong spatial correlation of ESV in adjacent areas, meaning similar ESV in neighboring regions, while a high value (0.75) indicated highly random spatial distribution of ESV, with significant regional differences. The range indicated the distance over which spatial correlation of ESV is effective.

Hot spot analysis

The Gi* index for hot spot analysis reflected the distribution pattern of cold and hot spots of ESV and spatial aggregation areas, revealing the spatial heterogeneity of Es capacity60. The cold and hot spot detection tool analyzed the ESV of grid units in the Anning River Basin during different periods, generating a spatial distribution map of cold and hot spots.

Ecosystem service sensitivity analysis

Traditional sensitivity analysis

The sensitivity coefficient CS was employed to evaluate the reliability of ESV calculations, thereby minimizing the uncertainty of the estimated results. In this study, the sensitivity index was determined by increasing or decreasing the total ESV and the value coefficients of various PLES land types by 50%61. The formula was:

$$CS=\left|\frac{({ESV}_{i}-{ESV}_{j})/{ESV}_{j}}{({VC}_{ik}-{VC}_{jk})/{VC}_{jk}}\right|$$
(5)

Here, ESV represented the total ESV, VC represented the unit area ESV coefficient of PLES land types, i and j denoted the values prior to and following the adjustment, and k represented different types of Es, CS was the sensitivity of various Es coefficients within the study area of this paper. If CS is less than 1, it indicates that ESV is inelastic with respect to value change, implying that the findings are reliable. Conversely, if CS exceeds 1, indicating that the results may lack credibility.

Coefficient of improved cross-sensitivity

CICS measured the sensitivity of the net conversion between two different Lu types to changes in ESV62,63,64. When the CICS value was greater than 0, it indicated that the net conversion of the two land categories would help enhance Es functions; conversely, it would suppress Es functions. As a sensitivity analysis indicator, the absolute value of CICS indicated the sensitivity of Es to LuT, with larger absolute values indicating higher sensitivity and vice versa. The formula was:

$${CICS}_{kl}=\frac{({V}_{ck}-{V}_{cl}){\Delta A}_{kl}}{{\Delta P}_{ESV}}$$
(6)

In the equation, Vck represented the adjusted ESV equivalent factor for land type k; Vcl denoted the adjusted ESV equivalent factor for land type l; ∆Akl indicated the net transformation area between land types k and l for the n-th year and the n-1 year; and ∆PESV signified the change in ESV for the n-th year and the n-1 year.

Optimal parameter geographic detector

At the core of the geographic detector methodology is an approach designed to identify and exploit spatial heterogeneity. Specifically, the optimal-parameter geographic detector65,66 employs the “GD” package in R (version 4.2.1) to discretize continuous variables and compute a g-value for each factor using various segmentation techniques—namely, equal interval, natural interval, quantile, geometric, and standard deviation classifications. The discretization scheme that produces the highest g-value is then selected as the optimal parameterization for the corresponding independent variable. Additionally, the factor detector is used to examine spatial variations in the dependent variable and to evaluate the influence of factors (X) through the q-value, which quantifies how effectively these factors explain the spatial heterogeneity of Y. The q-value, ranging from 0 to 1, indicates that higher values are associated with a stronger explanatory power regarding ecological and environmental quality.

Results

Dynamic changes and transformation analysis of the PLES land

Based on Table 2, the overall dynamic degree of P-E land in the Anning River Basin is approximately 1.23%, indicating an upward trend. A notable decrease in this rate was observed between 1995 and 2005, likely attributable to the implementation of the "Returning Farmland to Forest" pilot policy initiated in Sichuan Province in 1999. In contrast, the overall dynamic degree of E-P land is -0.44%, reflecting a gradual decline; however, an increase was recorded between 1995 and 2005, corresponding to the reduction in P-E land. Furthermore, the overall dynamic degree of E land stands at 2.54%, demonstrating a stable upward trend. Notably, the overall dynamic degree of P-L land is as high as 29.05%, signaling a substantial increase in its area over the past 38 years. This pattern clearly reflects elevated levels of human activity, an accelerated process of urbanization, and a significant impact on the ecological environment.

Table 2 Dynamic Degree of PLES Land in the Anning River Basin.
Full size table

Using the geo-information TuPu method, we analyzed the transformation areas of PLES land during four periods: 1985–1995, 1995–2005, 2005–2015, and 2015–2023 (see Fig. 2 and Table 3). Specifically, the areas transformed during these periods were 48,836.88 hectares, 71,284.05 hectares, 68,305.32 hectares, and 67,467.96 hectares, respectively. From 1985 to 1995, there was a notable expansion of E-P land. The transitions from "P-E land to E-P land" and "P-L land to E-P land" accounted for 62.49% and 0.82% of the total transformed area, respectively, while the growth of P-E land—which represented 33.65% of the transformed area—was primarily attributed to the conversion of E-P land. Additionally, E land expanded by encroaching upon E-P land, and P-L land predominantly emerged from the transformation of E land. In the subsequent period (1995–2005), the growth of P-E land became more pronounced, largely driven by conversions from E-P land (58.95%), E land (2.20%), and P-L land (1.96%). During this period, P-E land was primarily converted into E-P land, whereas P-L land was mainly transformed into both P-E land and E-P land. Between 2005 and 2015, E-P land increased by 42,211.89 hectares, constituting 61.80% of the total transformed area. Concurrently, 21,571.29 hectares of E-P land shifted to P-E land, reflecting a reduction of 20,451.51 hectares compared to the previous period (1995–2005). In the final period (2015–2023), the area converted to E-P land was significantly greater than in earlier periods. Notably, with the exception of the 1995–2005 period—during which the conversion from E-P land to P-E land exceeded that from P-E land to E-P land—all other periods exhibited the reverse trend.

Fig. 2
Fig. 2
Full size image

Transfer Map of Different PLES Land.

Table 3 Transfer Areas of Various PLES Land Types.
Full size table

By reclassifying the “inflow” and “outflow” of PLES Land, fluctuation maps (see Figs. 3 and 4) were generated. In conjunction with the transfer area data provided in Table 3, the analysis indicates that between 1985 and 1995, the newly added area of E-P land reached its maximum, while concurrently, the area of shrinking P-E land also peaked. A significant transformation relationship between these two was observed, with the inflow to E-P land notably exceeding that of P-E land. In the subsequent period from 1995 to 2005, the newly added area of P-E land peaked, particularly in the central and northern regions, while the shrinking area of E-P land also reached its highest point, resulting in a net differential change of 18,148.5 hectares. This phenomenon may be closely associated with the increasing demand for arable land driven by local economic development. Between 2005 and 2015, the newly added area of E-P land was evenly distributed throughout the basin, whereas the increase in P-E land was concentrated in the central region; this was in contrast to the shrinkage patterns observed for both land types. This trend may have been significantly influenced by the "Returning Farmland to Forest" policy, which led to a concentrated increase in P-L land in the central basin. From 2015 to 2023, the area of E-P land further expanded, particularly in the central basin, whereas P-E land exhibited an opposing trend. Throughout the entire study period from 1985 to 2023, the newly added and shrinking areas of both E land and P-L land were relatively minor and did not prominently feature in the fluctuation maps.

Fig. 3
Fig. 3
Full size image

Fluctuation Map of Newly Added Land. Figure was generated using ArcGIS version 10.7 (https://desktop.arcgis.com/zh-cn/desktop/index.html), and the map review number is GS (2024) 0650 (https://cloudcenter.tianditu.gov.cn/administrativeDivision/).

Fig. 4
Fig. 4
Full size image

Fluctuation Map of Shrinking Land. Figure was generated using ArcGIS version 10.7 (https://desktop.arcgis.com/zh-cn/desktop/index.html), and the map review number is GS (2024) 0650 (https://cloudcenter.tianditu.gov.cn/administrativeDivision/).

Temporal and spatial variation characteristics of ESV

Temporal variation characteristics of ESV

Figure 5 illustrates the trajectory of ESV in the Anning River Basin from 1985 to 2023. Notably, the calculation of ESV excludes P-L land; therefore, the ESV data presented in the chart does not incorporate any components associated with P-L land. During the study period, E-P land has served as the primary contributor to ESV, accounting for more than 90% of the total annually. In contrast, P-E land contributes approximately 6% to 7% of the total ESV each year, while E land contributes around 2%. An analysis of the total ESV contributions by the different types of PLES Land (see Fig. 6) reveals that both P-E land and E land significantly enhance ESV. After an initial decline observed in 2005, the ESV of P-E land exhibited a continuous increase, accumulating an overall rise of 148 million yuan. Meanwhile, the ESV of E land experienced fluctuations with an upward trend, amounting to an accumulated increase of 7.8 million yuan between 1985 and 2023. In contrast, the ESV of E-P land demonstrated a cumulative decrease of 766 million yuan during this period.

Fig. 5
Fig. 5
Full size image

Changes in ESV in the Anning River Basin.

Fig. 6
Fig. 6
Full size image

ESV of Different Types of PLES Land.

Figure 7 depicts the variations in individual ecosystem services values (ESV) for different types of PLES Land from 1985 to 2023. Analysis of the impacts of PLES Land changes on nine categories of ecosystem services reveals that the primary services provided by P-E land are hydrological regulation (HR) and food production (FP). In contrast, E-P land predominantly offers climate regulation (CR) and hydrological regulation (HR), while HR is also the principal service associated with E land. Regulating services constitute a significant portion of the overall ESV, contributing approximately 67% of the total value when evaluated on an individual basis (see Figs. 8, 9). However, these services have exhibited considerable fluctuations, marked by a sharp increase in 2005 followed by a continuous decline, culminating in a cumulative reduction of 326 million yuan. Meanwhile, supply services contribute the least, accounting for about 5.52% of the total ESV, and their trend has remained relatively stable (see Fig. 8). Comparable declining trends are observed in both supporting and cultural services (see Figs. 10, 11). Notably, aside from food production, which increased by 31 million yuan, the remaining ecosystem services generally followed a decreasing trend; climate regulation and biodiversity experienced the most substantial reductions, amounting to 196 million yuan and 74 million yuan, respectively.

Fig. 7
Fig. 7
Full size image

Changes in Individual ESV in the Anning River Basin.

Fig. 8
Fig. 8
Full size image

Changes in ESV of Provisioning Services.

Fig. 9
Fig. 9
Full size image

Changes in ESV of Regulating Services.

Fig. 10
Fig. 10
Full size image

Changes in ESV of Supporting Services.

Fig. 11
Fig. 11
Full size image

Changes in ESV of Cultural Services.

Spatial variation characteristics of ESV

Based on Table 4, between 1985 and 2023 the exponential model provided the best fit for the semivariogram at both the 1 km × 1 km and 500 m × 500 m scales, with the coefficient of determination (R2) remaining consistently similar across the years. Notably, at the 1 km × 1 km scale, the optimal model fit was achieved in 2015. In that same scale, the nugget (C₀) reached its maximum in 1995, suggesting that spatial heterogeneity in ecosystem service values (ESV) during that year was strongly influenced by random factors. Conversely, the nugget values in 2005 and 2015 were the lowest, indicating a reduced impact of random factors on spatial heterogeneity in those years. Over the study period, the sill (C₀ + C) demonstrated a gradual increase, implying a modest enhancement in the degree of spatial variation in ESV; however, the overall magnitude of change was slight, which indicates that the fundamental spatial structure remained relatively stable. Moreover, the nugget coefficient (C₀/(C₀ + C)) consistently remained below 0.25 throughout the period, underscoring that the spatial heterogeneity of ESV within the watershed was primarily determined by structural factors rather than random influences, thereby exhibiting strong spatial correlation. It is noteworthy that the overall declining trend in the nugget coefficient suggests that the influence of random factors has been gradually increasing. Additionally, in 2005 the range parameter reached its maximum, indicating an expanded spatial correlation in ESV for that year.At the 500 m × 500 m scale, apart from a relatively small nugget value in 2005, the nugget values across other years were similar, and the trends observed in both the sill (C₀ + C) and the nugget coefficient were consistent with those at the 1 km × 1 km scale. However, the range parameter at the 500 m × 500 m scale differed from that at the 1 km × 1 km scale, likely due to the finer resolution more clearly reflecting the topographic characteristics of the study area, where variations in terrain influenced the range.

Table 4 Fitting Parameters of the Semi-Variogram for ESV.
Full size table

Figures 12 and 13 illustrate the spatial distribution of ESV hotspots and cold spots at both scales. Overall, from 1985 to 2023 the spatial distribution of hotspots at the 1 km × 1 km scale remained relatively constant; hotspots were primarily concentrated in high-altitude regions extensively covered by forests and grasslands, which are known for their significant climate regulation effects. In contrast, cold spots were predominantly observed in plain regions, particularly in areas experiencing frequent human activities and expansive construction land use, which directly contributed to the development of ESV cold spots. Notably, after 2005 the hotspot distribution reached its greatest extent, before later diminishing. At the 500 m × 500 m scale, the spatial patterns of hotspots and cold spots were analogous to those at the 1 km × 1 km scale, yet the finer scale provided more detailed observations. Between 1985 and 2005, there was a noticeable contraction in the area of cold spots accompanied by an expansion in hotspot areas; after 2005, this pattern reversed, with an increase in cold spots and a decline in hotspots, mirroring the trend observed at the larger scale. The increase in hotspot areas may be attributed to reforestation policies (i.e., the conversion of farmland to forest), which expanded forest distribution and consequently enhanced ESV. Concurrently, the post-2005 shifts in hotspot and cold spot distributions suggest that intensified human activities have had a clear impact on ecosystem service values.

Fig. 12
Fig. 12
Full size image

1 × 1 km Distribution of Cold and Hot Spots of ESV. Figure was generated using ArcGIS version 10.7 (https://desktop.arcgis.com/zh-cn/desktop/index.html), and the map review number is GS (2024) 0650 (https://cloudcenter.tianditu.gov.cn/administrativeDivision/).

Fig. 13
Fig. 13
Full size image

500 × 500 m Distribution of Cold and Hot Spots of ESV. Figure was generated using ArcGIS version 10.7 (https://desktop.arcgis.com/zh-cn/desktop/index.html), and the map review number is GS (2024) 0650 (https://cloudcenter.tianditu.gov.cn/administrativeDivision/).

Sensitivity analysis of ESV

By increasing the ESV coefficient by 50%, a sensitivity analysis of the Anning River Basin’s ESV was conducted (see Table 5). The results indicate that the sensitivity coefficient for E–P land is the highest, primarily due to the substantial proportion of forest and grassland in the basin. In contrast, the sensitivity coefficients for the three types of Lu remain below 1, suggesting that the ESV coefficient exhibits resilience and that the assessment results are robust. The CICS model was employed to calculate the cross-sensitivity coefficients for the periods 1985–1995, 1995–2005, 2005–2015, and 2015–2023 (Fig. 14). It is evident that the transformation of E–P land and P–E land to other land-use types shows higher sensitivity to ESV, with the conversion from E–P land to P–E land being the most sensitive. With the exception of a sensitivity coefficient of 0.69 for E land during 1995–2005, the sensitivities for other periods and conversions to the remaining land-use types are relatively low, with CICS values generally below 0.2, indicating limited sensitivity of ESV to these transformations. Additional key findings are as follows:

  1. (1)

    Cross-Sensitivity of P–E Land to Other Land-Use Types: During the study period, apart from the positive CICS values for the conversion from P–E land to E land in 1995–2005 and 2005–2015, all other values were negative, indicating that these conversions tend to suppress ESV. The absolute CICS value between P–E land and E–P land remained relatively stable, whereas the absolute CICS value between P–E land and E land fluctuated from 0.28 to 0.54, indicating increased sensitivity of ecosystem services (ES) to this conversion.

  2. (2)

    Cross-Sensitivity of E–P Land to Other Land-Use Types: The absolute CICS value for the conversion between E–P land and P–E land increased from 3.15 to 3.19, demonstrating heightened sensitivity of ES to this conversion, which is classified as highly sensitive. Notably, the CICS shifted from − 3.15 during 1985–1995 to 2.67 during 1995–2005, reflecting a change from a suppressive to a promoting effect on ESV; subsequently, the CICS reverted to negative values, indicating that this conversion continues to exert a suppressive effect. In contrast, the sensitivity of ES to conversions between E–P land and E land remained stable with minimal variation.

  3. (3)

    Cross-Sensitivity of E Land to Other Land-Use Types: The CICS for the conversion between E land and P–E land shifted from − 0.06 to − 0.16, suggesting an increased sensitivity of the ecosystem service system to this conversion while maintaining its suppressive effect on ESV. Additionally, the CICS for the period 1995–2005 was 0.69, implying a temporary transition from suppression to promotion of ESV. Meanwhile, the absolute CICS values for the conversion between E land and E–P land were both below 0.1, indicating minimal sensitivity of the ES system to this type of conversion.

Table 5 Sensitivity Analysis of ESV.
Full size table
Fig. 14
Fig. 14
Full size image

Cross-sensitivity coefficients of ESV in the Anning River Basin.

Analyzes the driving factors

Based on 500 × 500 m resolution data from the Anning River Basin, we calculated the total ESV as the dependent variable and selected nine potential driving factors as independent variables: nighttime light data, GDP, population density, land use type(LuT), road network density, DEM, slope, annual average precipitation, and soil erosion level(SER). A multicollinearity test revealed that the tolerance values for all nine factors exceeded 0.1 and their variance inflation factors were all below 3, indicating an absence of multicollinearity and confirming their suitability for subsequent analyses. The data were classified and discretized using the natural breaks method, and an optimized parameter geographic detector was employed to assess the influence of each driving factor on ESV as well as the interactions between these factors. The p-values for all nine factors were nearly 0, underscoring the study’s strong statistical significance. In the single-factor analysis (Fig. 15), population density exhibited the highest q-value at 0.602, followed by land use type (0.489), DEM (0.484), and road network density (0.333), while the other factors displayed relatively low q-values. Among the factors significantly affecting ESV, socioeconomic factors were predominant, particularly the influence of population density.

Fig. 15
Fig. 15
Full size image

Single-factor detection results of ESV driving factors.

Furthermore, as illustrated in Fig. 16, dual-factor interactions exhibited a markedly stronger driving effect compared to single factors. For instance, the interaction effects of population density with elevation and land use had q-values of 0.687 and 0.683, respectively, thereby enhancing the explanatory power of individual factors in accounting for spatial variations in ESV. Additionally, q-values for the interaction effects involving population density with other factors all exceeded 0.6, indicating a pronounced strengthening of dual-factor interactions. The q-values for interactions involving land use, DEM, and other factors were also above 0.5, and the interactions between GDP with DEM and land use similarly exceeded 0.5. These findings suggest a significant interplay between socioeconomic and natural environmental factors, implying that shifts in socioeconomic conditions may impact natural ecosystems and consequently alter ecosystem service values.

Fig. 16
Fig. 16
Full size image

Interaction detection results of ESV driving factors.

Discussion

Analysis of dynamic changes and conversions

This study presents a detailed examination of the dynamic changes and land-use conversions among P-E land, E land, E-P land, and P-L land within the Anning River Basin. The analysis reveals an overall dynamic degree of 1.23% for PLES land, suggesting positive shifts in regional land use that are likely attributable to recent policy implementations, urbanization processes, and increased human activities67,68. Notably, between 1995 and 2005, there was a significant reduction in the dynamic degree of P-E land, a change that appears directly linked to Sichuan Province’s 1999 “Returning Farmland to Forest” policy69. This initiative, which aimed to curtail the exploitation of cultivated land and restore the ecological environment, markedly altered the region’s land-use structure during that period.

Furthermore, the investigation of the transition from ecological to E-P shows a dynamic degree of -0.44%, indicative of a gradual reduction in this category. Despite an increase in the absolute area of E-P land between 1995 and 2005, this trend is likely a consequence of the ongoing decline in P-E land. The study also highlights that E land exhibits a dynamic degree of 2.54%, reflecting a consistent upward trend that underscores a heightened local emphasis on ecological preservation. In contrast, P-L land demonstrated a remarkably high dynamic degree of 29.05% over the 38-year study period, indicating significant expansion driven primarily by accelerated urbanization and intensified human activities70.

Utilizing the geographic TuPu method, the conversion processes of PLES land were analyzed across four distinct periods from 1985 to 2023. The findings indicate that between 1985 and 1995, a substantial expansion of E-P land occurred concurrently with a reduction in P-E land. During this interval, the conversion of P-L land to other uses was particularly noteworthy, with evidence suggesting that much of the P-L land originated from E land. The observed conversion patterns throughout the study period further elucidate the interconnected dynamics of urbanization and economic development71. Specifically, from 1995 to 2005, the increase in P-E land was predominantly driven by the conversion of both E-P land and E land, indirectly highlighting the influence of economic growth on the demand for cultivated land. Between 2005 and 2015, the expansion of E-P land was particularly pronounced, accompanied by a reduction in conversions from E-P to P-E, possibly due to subtle adjustments in government policies72. The persistence of the “Returning Farmland to Forest” policy during this period also contributed to a substantial increase in P-L land in the central region.

Temporal-spatial characteristics of ESV

Temporal characteristics

This study rigorously examines the temporal evolution of ESV in the Anning River Basin from 1985 to 2023. As illustrated in Fig. 5, the E-P land type has consistently contributed over 90% of the annual ESV, affirming its central role. In contrast, the P-E land type has consistently contributed between 6% and 7%, whereas the E land type accounted for only approximately 2%. This structural imbalance underscores the diverse functionalities and contribution levels across different land types in delivering ecosystem services73. Notably, the ESV of the P-E land experienced a slight decline in 2005; however, it subsequently increased, culminating in a cumulative growth of 1.48 million yuan. Similarly, while the E land demonstrated an upward fluctuating trend over the period, its cumulative contribution increased by only 78,000 yuan. In stark contrast, the E-P land experienced a cumulative reduction of 7.66 million yuan, reflecting a significant decline in ecosystem service value. This trend is likely attributable to the tension between regional economic development and ecological conservation, particularly under the pressures of rapid industrialization and urbanization, which curtailed the utilization of E-P land and, consequently, its associated ecosystem services74.

An analysis of changes across different ecosystem service types (see Fig. 7) reveals that the P-E land principally supported hydrological regulation (HR) and food production (FP) services, while the E-P land excelled in climate regulation (CR) and hydrological regulation (HR). The E land also contributed to hydrological regulation. Overall, regulatory services constituted approximately 67% of the total ESV, underscoring their critical role in maintaining ecological balance and sustaining ecosystem functions. However, these regulatory services demonstrated significant fluctuations over time, particularly after 2005, when their combined value declined by 3.26 million yuan75. Such fluctuations not only reflect alterations in ecosystem conditions but also highlight its vulnerability, thereby signaling the need for increased efforts in the maintenance and restoration of ecosystem services. Although food production services increased by a total of 310,000 yuan during this period, other services, including climate regulation and biodiversity enhancement, experienced declines of 1.96 million yuan and 740,000 yuan40, respectively. The pronounced reduction in climate regulation services suggests that amid global climate change and escalating regional environmental pressures, the ecosystem’s resilience and recovery capacity are crucial. These findings offer valuable insights for regional planning, underscoring the importance of integrating robust ecological protection measures with economic development initiatives to preserve the stability of ecosystem services.

Spatial characteristics

The spatial analysis, as summarized in Table 4, indicates that between 1985 and 2023, the exponential model provided the best fit for the semivariogram at both the 1 km × 1 km and 500 m × 500 m scales. The consistency of the determination coefficients (R2) across both scales suggests similar patterns of ESV variation regardless of spatial resolution. At the 1 km × 1 km scale, the model achieved its best fit in 2015, while the nugget (C₀) reached its maximum in 1995, implying that random factors predominantly influenced spatial heterogeneity in that year. Conversely, the nugget values for 2005 and 2015 were substantially lower, indicating a reduced influence of random factors during these periods. Over the entirety of the study period, the range (C₀ + C) exhibited a gradual increase, signifying an intensification of spatial variability in ESV; however, the overall magnitude of change was modest, suggesting a relatively stable spatial configuration within the basin. Furthermore, the nugget coefficient (C₀/(C₀ + C)) consistently remained below 0.25, indicating that the spatial heterogeneity of ESV across the basin was primarily driven by structural factors rather than random variability, thereby demonstrating strong spatial correlation76. Notably, the gradual decrease in the nugget coefficient over the study period implies that the impact of random factors on spatial heterogeneity may be increasing, potentially associated with the high frequency of human activities in the region.

Figures 12 and 13 illustrate the spatial distribution of ESV hotspots and cold spots. At a 1 × 1 km scale, hotspots concentrate primarily in high-altitude regions dominated by forests and grasslands, which play a significant role in climate regulation. In contrast, cold spots predominantly occur in plains characterized by frequent human activities77, particularly in areas undergoing urban expansion and extensive land development. This spatial pattern indicates that anthropogenic activities have directly reduced ESV and created regional imbalances. After 2005, the extent of hotspots reached its maximum—an observation possibly linked to policy adjustments and the implementation of restoration measures during that period. However, in subsequent years, the number of cold spots increased while hotspots declined, underscoring the growing impact of human activities on ESV. Analysis at a finer scale of 500 m × 500 m further corroborates this trend; the granular changes in hotspots and cold spots highlight the significant influence of topographic features on ESV distribution78.

Sensitivity analysis of ecosystem services

The sensitivity analysis of ESV reveals that the sensitivity coefficient for E-P land is the highest, primarily due to the abundance of forests and grasslands in the region79. Notably, when comparing the sensitivity of P-E land with that of E land, the latter demonstrates greater stability, with coefficients remaining below 1; this suggests a degree of resilience in ESV. These results provide a theoretical foundation for future ecosystem management by emphasizing the need to consider interrelationships among different land uses and to adopt appropriate conservation measures. Additionally, combined with the CICS model analysis, the cross-sensitivity coefficients over various periods indicate that the transition between E-P land and P-E land exhibits high sensitivity. This finding suggests that in policy formulation and land management, the dynamic relationships among different land types must be recognized to enhance overall ecological benefits effectively. Notably, from 1985 to 1995, the nature of the transformation of E-P land shifted from suppressive to promotive, only to revert again to a suppressive state thereafter64. This evolution underscores the necessity of accounting for dynamic policy-driven effects when evaluating ecosystem services.

Driver analysis

The single-factor analysis identifies population density as having the highest q-value (0.602)80, indicating its significant impact on ESV. This finding clearly demonstrates the direct impact of population concentration on the demand for ecosystem services. In regions with higher population densities, the heightened requirement for arable land and ecosystem services exacerbates the challenges associated with ecological protection and sustainable land management. However, this observation contrasts with the study by Xue et al.81, which determined that natural factors exert a more significant influence than social factors, such as population, on the demand for ecosystem services in the Yellow River Basin. This discrepancy may be attributed to variations in the selected factors and regional characteristics. Following population density, land use type (q-value of 0.489) and digital elevation model (DEM; q-value of 0.484) also have substantial impacts on ESV, suggesting that variations in topography and land use practices significantly influence the capacity and efficiency of ecosystem service provision. Consequently, policy formulation should emphasize rational land use planning to maximize the composite benefits of ecosystem services. Furthermore, the analysis reveals that the interaction effects of dual factors are generally stronger than those of individual factors82. Notably, the interactions between population density and DEM and between population density and land use types exhibit strong explanatory power, with q-values of 0.687 and 0.683, respectively. These results imply that effective policy formulation requires the integration of multiple driving factors to attain optimal ecological protection while meeting economic development objectives. Particularly in rapidly urbanizing areas, the pronounced impact of urban expansion on the ecological environment calls for policies that coordinate financing, infrastructure development, and ecological protection to generate synergistic benefits.

In summary, the relationship between dynamic changes in PLES land in the Anning River Basin and ESV underscores the complex interactions among human activities, policy interventions, and environmental conservation. As the region pursues economic development, further optimization of land use and the enhancement of ecosystem protection and restoration initiatives are imperative for ensuring sustainable development. Future research should explore how ESV under varying socio-economic contexts influences policy choices and long-term outcomes, thereby providing a robust foundation for the formulation of scientifically sound and effective ecological protection policies.

Future prospects and research limitations

As human activities continue to intensify and the effects of climate change become increasingly pronounced, the tension between ecological conservation and economic development demands has grown. This study provides a critical theoretical framework for elucidating the dynamic changes in land use and ecosystem service value within the Anning River Basin; however, several limitations warrant further investigation. First, although the study period spans from 1985 to 2023, it does not fully capture the entire trajectory of ecosystem service dynamics. Future research should consider extending the temporal scope to comprehensively examine the long-term impacts of policy interventions on land use changes and ecosystem service value, while integrating remote sensing technology with GIS methodologies to enhance data accuracy and timeliness. Second, despite the inclusion of various potential driving factors, the analysis of the complex interplay between socioeconomic and environmental influences remains inadequate. Subsequent investigations should focus on elucidating the specific roles of sociocultural factors and community participation in ecological protection. Third, the analytical models and methods employed may overlook subtle differences in ecosystem responses; thus, future studies could incorporate advanced techniques, such as machine learning, to more precisely reveal the underlying interconnections within ecosystem feedback mechanisms. Finally, the influence of geographic and spatial factors on ecosystem service value requires deeper exploration. In the context of rapid urbanization, a systematic analysis of the effects of varying land use types, quality, and spatial distribution on ecosystem service value is urgently needed, with quantitative studies based on high-resolution satellite data offering a promising avenue for obtaining more detailed insights. Therefore, future research should broaden both methodological approaches and analytical perspectives to assist policymakers in making more scientifically robust ecological management decisions, thereby advancing sustainable development and the construction of an ecological civilization.

Conclusions

  1. (1)

    Significant Lu changes occurred in the Anning River Basin. From 1985 to 1995, E-P land made up 62.49% of changes. After the 2005 Grain for Green policy, P-L land increased by 29.05%, highlighting urbanization pressures on the ecosystem.

  2. (2)

    E-P land provided over 90% of the ESV, while P-E and E lands contributed about 6–7% and 2%, respectively. ESV increased by 148 M yuan for P-E and 78 M yuan for E lands but decreased by 766 M yuan for E-P. Regulating services made up roughly 67% of the total ESV yet dropped by 326 M yuan since 2005.

  3. (3)

    Marked spatial heterogeneity exists in basin ESV. In 1995, random factors dominated, but by 2005 and 2015, structural factors prevailed. ESV variation has increased over time, with high-altitude hotspots and plains cold spots. Since 2005, both have contracted due to ecological restoration and sustainable development.

  4. (4)

    Ecosystem services are highly sensitive to E-P conversion, while P-E changes suppress ESV. From 1985 to 2005, net effects between these conversions shifted from suppression to promotion, and E conversions have minimal impact.

  5. (5)

    Based on 500 m resolution data for the Anning River Basin, population density is the key driver of ecosystem service value (q = 0.602). When combined with elevation and land use, its explanatory power increased, highlighting interactions between socioeconomic and natural factors.