Introduction

Accounting for about 15% of the global land area, karst is a distinctive landscape generated from the interactions between water and soluble rocks (e.g., limestone and dolomite), forming highly diverse but vulnerable environments at the surface and underground [1, 2]. Covering more than 550,000 km2, the karst area in Southwest China represents the most typical and complex tropical-subtropical karst development, which is also the largest and most concentrated karst ecologically fragile area in the world [3, 4]. South China Karst is a serial World Heritage (WH) nomination submitted by the State Party China to the UNESCO World Heritage Committee [5]. Phase I was inscribed on the WH List in 2007 and comprises three internationally acclaimed areas: Shilin Karst, Libo Karst, and Wulong Karst. Phase II was inscribed in 2014 and includes Guilin Karst, Shibing Karst, Jinfoshan Karst, and Huanjiang Karst as an extension of Libo Karst. Due to the fragility, sensitivity, and low natural regeneration capacity of karst resources and environments [6], karst WH ecosystems are more vulnerable to degradation and less likely to recover once degraded than non-karst counterparts [7,8,9]. This poses a serious threat to the outstanding universal value, authenticity, and integrity of South China Karst [10].

While WH sites are prioritized for ecological conservation, they face serious threats from natural factors and human pressures, such as climate change, species invasions, tourism, and land use change [11,12,13]. For example, although the accreditation of WH is oriented to identify, recognize, and protect hotspots of outstanding universal value, it has increasingly been utilized as a marketing tool for tourism promotion due to the benefits of the “WH” brand [14,15,16]. Once an area is inscribed on the WH List, it usually becomes a well-known tourist destination, with an order of magnitude more tourist visits than unlisted destinations [17]. These threats inevitably alter the structure and processes of WH ecosystems, resulting in habitat deterioration, biodiversity loss, degenerated ecosystem functions, and ultimately declined ecosystem services [18]. Therefore, information on the spatiotemporal changes and driving factors of ecosystem services is needed to capture the ecological environment of WH sites. This can help to design timely and effective management strategies that aim to reconcile conflicts between ecological conservation priorities and economic development demands in WH sites.

Ecosystem services refer to all direct and indirect benefits that humans derive from the natural environment [19]. These services are influenced by multiple natural and anthropogenic factors, such as land use, topography, climate, vegetation, soil, and human activities [20,21,22]. Due to the outstanding universal value of biodiversity and the natural environment, WH sites can provide multiple regulating, supporting, and cultural services to humankind. However, different services may interact antagonistically or synergistically, resulting in trade-offs or synergies between ecosystem services [23]. Wang et al. [24] reported that carbon storage, habitat quality, aesthetic value, and recreational value of the Bogda WH site showed significant spatiotemporal variation, with higher values in central forests and high-coverage grasslands. Sui et al. [25] found that forests were crucial for maintaining habitat quality at the Jiuzhaigou WH site. They also found that ecological restoration projects effectively improved habitat quality after geohazards [25]. Wang et al. [26] showed that the ecosystem service value of the Bayanbulak WH site declined over time as a result of habitat degradation in the buffer zone. While these studies provide important information on the ecosystem services of non-karst WH sites, little is known about the spatiotemporal changes, trade-offs/synergies, and driving factors of the ecosystem services provided by karst WH sites [27]. These knowledge gaps may hinder our ability to balance the conservation and development of karst WH sites.

In this study, we used the InVEST tool to investigate spatiotemporal changes in four key ecosystem services —habitat quality (HQ), carbon storage (CS), soil retention (SR), and water conservation (WC)—of karst and non-karst WH sites in Southwest China. We also used the spatial overlay method to assess trade-offs/synergies among ecosystem services. Finally, we used geographical detector model (GDM) and structural equation modeling (SEM) to identify driving factors of variation in ecosystem services. The objectives of this study were (1) to assess the spatiotemporal variation in ecosystem services of karst and non-karst WH sites from 2000 to 2020, (2) to explore trade-offs and synergies among ecosystem services of karst and non-karst WH sites, and (3) to identify the key driving factors influencing ecosystem services. By comparing ecosystem services between karst and non-karst WH sites and revealing the key factors driving these differences, our study would advance our understanding of ecosystem services delivered by karst WH sites and provide constructive suggestions for optimal management and sustainable development of WH with different attributes.

Materials and methods

Study area

We selected two typical and representative karst WH sites (Shibing Karst and Libo-Huanjiang Karst) as the study area, with two non-karst WH sites (Fanjingshan and Chishui Danxia) as the reference (Fig. 1). Shibing Karst and Libo-Huanjiang Karst are the components of South China Karst, which is the most spectacular example of humid tropical to subtropical karst landscapes in the world. Shibing Karst provides a world reference for dolomite karst in tropical and subtropical regions [28]. Libo-Huanjiang Karst is the best example of cone karst in the world [29]. Fanjingshan contains unique biological and ecological value, which is an outstanding representative of the evolution of subtropical mountain forests [30]. Chishui Danxia preserves a typical subtropical evergreen broad-leaved forest, which is an ideal place to study the ancient and modern vegetation on sandstone [31]. Overall, these four WH sites not only include outstanding universal value in terms of natural beauty, geological features, and biodiversity, but also provide a wide range of ecosystem services to humankind.

Fig. 1
figure 1

Distribution of the four World Heritage sites in Southwest China

Full size image

Data sources

We collected multi-source data to assess spatiotemporal changes and driving factors of ecosystem services provided by the four WH sites (Table 1). Specifically, we evaluated ecosystem services based on land use data, biophysical parameterization, meteorological, topographic, and soil data. We explored the driving factors of ecosystem services based on climate, topography, landscape, soil, vegetation, and human activity data. Land use data were divided into six classes: cropland, forest, shrubland, grassland, water body, and impervious land. Data in non-raster formats, such as climate data, were processed using the kriging spatial interpolation in ArcGIS 10.2 software for spatial mapping. All data were converted to a uniform projection coordinate system (WGS-1984) and were resampled to a spatial resolution of 30 m.

Table 1 Information on the data used to assess ecosystem services and identify the driving factors
Full size table

Ecosystem service assessment

Based on the environmental conditions and ecological functions of WH sites, we evaluated four key ecosystem services (i.e., HQ, CS, SR, and WC) using the InVEST model (version 3.12), which is one of the most widely used tools for ecosystem service assessment [21, 32, 33]. Specifically, we used the habitat quality module for HQ, the carbon storage module for CS, the sediment delivery ratio module for SR, and the water yield module for WC. The detailed parameters were provided in Tables S1–S3 [34,35,36,37,38,39,40,41,42,43] and the detailed equations for each service were as follows.

Habitat quality

HQ refers to the ability of ecosystems to provide the necessary conditions for the survival and breeding of species, which is key to the protection of biodiversity. The habitat quality module of the InVEST tool evaluates HQ using the habitat suitability, influencing distance and weight of threat sources, and the sensitivity of each habitat type to threat sources [32, 44]. HQ was calculated as:

$${HQ}_{i,j}={H}_{i}\times \left(1-\frac{{D}_{i,j}^{z}}{{D}_{i,j}^{z}+{k}^{z}}\right),$$
(1)

where \({HQ}_{i,j}\) is the HQ for land use type \(i\) in pixel \(j\), \({H}_{i}\) is the habitat suitability for land use type \(i\), \({D}_{i,j}\) is the threat level for land use type \(i\) in pixel \(j\), \(k\) is the half-saturation constant, and \(z\) is a default parameter.

Carbon storage

Terrestrial ecosystems absorb CO2 from the atmosphere through photosynthesis and store it in vegetation, soil, and litter, thereby providing a climate regulating service. The carbon pool includes aboveground biomass, belowground biomass, soil, and dead organic matter [32, 45]. The carbon module of the InVEST tool evaluates CS using the land use data and the carbon density of each land use type. CS was calculated as:

$${C}_{total}= {C}_{above}+{C}_{below}+{C}_{soil}+{C}_{dead},$$
(2)

where \({C}_{total}\) is the total CS, \({C}_{above}\), \({C}_{below}\), \({C}_{soil}\), \({C}_{dead}\) are the CS of each land use type in aboveground biomass, belowground biomass, soil, and dead organic matter, respectively.

Soil retention

SR primarily includes erosion control and overland sediment retention, which is particularly important in mountainous regions. The sediment delivery ratio module of the InVEST tool calculates SR by subtracting actual soil erosion from potential soil erosion based on the revised universal soil loss equation [32, 45]. SR was calculated as:

$$SR=R\times K\times LS\times \left(1-C\times P\right),$$
(3)

where \(R\) is the rainfall erosivity factor, \(K\) is the soil erodibility factor, \(LS\) is the topographic factor, \(C\) is the vegetation practice factor, and \(P\) is the management factor.

Water conservation

WC is the ability of ecosystems to retain and store precipitation through their unique structure and water interactions, and to regulate water flows and cycles through evapotranspiration. To evaluate WC, we first used the water yield module of the InVEST tool to calculate water yield by subtracting evapotranspiration from the precipitation based on the regional water balance [32, 45]. Water yield was calculated as:

$$WY=\left(1-\frac{AET}{P}\right)\times P,$$
(4)

where \(WY\) is the water yield, \(AET\) is the actual annual evapotranspiration, \(P\) is the annual precipitation. We then integrated the topographic index, the soil saturation hydraulic conductivity, and the flow rate coefficient to calculate WC as [46]:

$$WC = \,min\left( {1,\frac{{249}}{{Velocity}}} \right) \times min\left( {1,\frac{{0.9 \times TI}}{3}i} \right) \times min\left( {1,\frac{{Ksat}}{{300}}} \right) \times WY,$$
(5)

where \(Velocity\) is the flow rate coefficient, \(TI\) is the topographic index, \(Ksat\) is the soil saturation hydraulic conductivity, and \(WY\) is the water yield.

Combined ecosystem service

To assess the spatiotemporal changes in ecosystem dimensions, we standardized the above four ecosystem services by unifying their values between 0 and 1, and then summed them to obtain the combined ecosystem service (CES) [32, 47]:

$$CES={\sum }_{i=1}^{4}\frac{{ES}_{i,j}-{ES}_{i,min}}{{ES}_{i,max}-{ES}_{i,min}},$$
(6)

where \({ES}_{i,j}\) is the value of the ecosystem service \(i\) in pixel \(j\), and \({ES}_{i,max}\) and \({ES}_{i,min}\) are the maximum and minimum values of the ecosystem service \(i\), respectively.

We calculated the mean value and standard deviation of ecosystem services for each WH site in each year using the R package ‘raster’ (version 3.6-20) [48]. We then used one-way ANOVA to test the difference in the mean value and standard deviation of ecosystem services between karst and non-karst WH sites. Data were checked for normality using the Shapiro–Wilk normality test and log-transformed where necessary.

Trade-off/synergy assessment of ecosystem services

Different from traditional correlation analysis that could only assess trade-offs/synergies between two types of services [33, 49], we used the ranking method to analyze trade-offs/synergies among the four ecosystem services simultaneously [50], which is also widely used in other studies [51,52,53,54]. First, we normalized the values using a scale from 0 to 1 by dividing all values by the maximum value for each service. Second, we used the natural break classification method to produce three provision levels (low, medium, and high) for each of the four services (Table S4), which were coded as 1, 2, and 3, respectively. Finally, we determined the interaction codes (IC) by combining the provision levels of individual services into an “ecosystem service bundle” using the following formula:

$$IC=HQ\times 1000+CS\times 100+SR\times 10+WC.$$
(7)

The output codes were IC numbers, which varied between 1111 and 3333. According to the IC numbers, we classified the relationships among ecosystem services into strong and weak trade-offs/synergies (Table S5). A strong trade-off occurs when the capacity of one service is high while the capacity of other services is medium or low. A weak trade-off occurs when the capacity of two or three services is high while other services exhibit a low capacity. A high synergy occurs when the capacity of all services is high or medium. A low synergy occurs when all services exhibit a low or medium capacity. We also used one-way ANOVA to test the difference in the proportion of strong and weak trade-offs/synergies among ecosystem services between karst and non-karst WH sites.

Driving factor analysis of ecosystem services

To identify the key drivers influencing the variation in ecosystem services, we selected a total of 14 potential factors from the domains of climate, topography, landscape, soil, vegetation, and human activities (Table 2). Climate factors included mean annual temperature (MAT) and mean annual precipitation (MAP). Topographic factors included digital elevation model (DEM) and slope. Landscape was indicated by the landscape division index, which has been widely used to describe the degree of landscape fragmentation [47, 55] and was calculated from the land use data using FRAGSTATS software (version 4.2). Soil factors included soil thickness, pH, soil organic carbon (SOC), and bulk density. Vegetation was indicated by the normalized difference vegetation index (NDVI). Human factors included population density, distance from road, nighttime light, and gross domestic product (GDP).

Table 2 Information on the driving factors of ecosystem services
Full size table

Geographical detector model

GDM is a powerful spatial statistical technique based on the analysis of spatial variation in the geographical layers of variables, which includes factor detection, risk detection, ecological detection, and interactive detection [32, 45, 56, 57]. We used factor detection to quantitatively determine the individual contribution (i.e., explanatory power q-value) of the 14 driving factors to the variation in CES. We also used interactive detection to determine the degree of interaction between any two drivers, as measured by the q-value. The q-value was expressed as follows:

$$q=1- \frac{{\sum }_{h=1}^{L}{N}_{h}{\upsigma }_{h}^{2}}{N{\upsigma }^{2}},$$
(8)

where h = 1, 2, 3, …, L is the classification or stratification of the driving factor; Nh and \({\upsigma }_{h}^{2}\) are the sample number and variance of the layer h, respectively; N and \({\upsigma }^{2}\) are the total sample number and variance, respectively; and q is the explanatory power of the driving factor and ranges from 0 to 1. The larger the q value, the greater the contribution of the driving factor to the variation in CES. Interactions between any two factors (e.g., X1 and X2) were classified into five categories based on whether they enhance or diminish the explanatory power when acting together, or whether they affect each other independently (Table 3). We conducted the GDM using the R package ‘GD’ (version 1.10), which is an optimal parameters-based GD tool where the optimal parameters are explored as the best combination of spatial data discretization method, break number of spatial strata, and spatial scale parameter [56].

Table 3 Classification of factor interaction
Full size table

Structural equation modeling

SEM is a widely used tool to identify causal relationships between variables in ecology and environmental science [58,59,60]. To explore the direct and indirect effects of climate, topography, landscape, soil, vegetation, and human factors on CES, we constructed a SEM based on a priori conceptual model. We first constructed composite variables for climate, topography, soil, and human factors using their corresponding measured variables, and then included these composite variables in the SEM. We used the following common indicators to check the fitness of the model: (1) Chi-square (χ2) test, with a p > 0.05 indicating that the model has a good fit; (2) root mean square error of approximation (RMSEA), which can overcome the drawback of the overall different value influenced by the estimated parameters in model validation, with a value < 0.05 indicating that the fitness of the model is good; (3) Bentler’s comparative fit index (CFI) and Tucker-Lewis index (TLI), which are between 0 and 1, with a value > 0.95 indicating that the model fits the data very good. We conducted SEM using the maximum-likelihood estimation method with the R package ‘lavaan’ (version 0.6-15) [61].

Results

Spatiotemporal differences in ecosystem services of the WH sites

Overall, there were high spatiotemporal differences in ecosystem services of the four WH sites (Figs. 2, 3, and S1–S7). In terms of spatial variation, the values of ecosystem services in the property zone were much higher than in the buffer zone across the four WH sites (Figs. 2, S1–S4). In terms of temporal variation, irrespective of WH sites, the mean value of HQ gradually decreased from 2000 to 2020 (Fig. 3a), while that of SR increased (Fig. 3c). There was little temporal change in the mean value of CS, except for a decreasing trend over time in Shibing Karst (Fig. 3b). For WC, the mean value decreased from 2000 to 2010 and then increased in 2020 in Chishui Danxia and the two karst WH sites, while the trend was reversed in Fanjingshan (Fig. 3d). Irrespective of WH sites, the mean value of CES gradually decreased from 2000 to 2020 (Fig. 3e). There were similar temporal patterns of ecosystem services between the property zone and the buffer zone of the four WH sites (Figs. S3, S4). Except for WC, the mean values of HQ, CS, and SR were significantly higher in non-karst WH sites than in karst WH sites, resulting in the same pattern for the mean value of CES (Figs. 3, S3, and S4).

Fig. 2
figure 2

Spatial differences in ecosystem services of the four World Heritage sites in Southwest China in 2020. HQ: habitat quality; CS: carbon storage; SR: soil retention; WC: water conservation; CES: combined ecosystem service

Full size image
Fig. 3
figure 3

Temporal variation in the mean value of habitat quality (HQ, a), carbon storage (CS, b), soil retention (SR, c), water conservation (WC, d), and combined ecosystem service (CES, e) of the four World Heritage sites in Southwest China. One-way ANOVA was used to test the difference in the mean value of ecosystem services between karst and non-karst World Heritage sites

Full size image

There was also temporal variation in the spatial heterogeneity of ecosystem services of the four WH sites (Fig. S5). Specifically, the standard deviation of all ecosystem services increased from 2000 to 2020, except for the standard deviation of CS and WC in Libo-Huanjiang Karst (Fig. S5). These temporal changes in the spatial heterogeneity of ecosystem services were largely consistent between the property zone and the buffer zone of the four WH sites (Figs. S6, S7). In terms of spatial variation, except for SR, the standard deviation of ecosystem services was greater in the buffer zone than in the property zone of the four WH sites (Figs. S6, S7). Furthermore, compared to non-karst sites, karst WH sites had a significantly higher standard deviation of CES, which was particularly evident in the standard deviation of CS and WC (Figs. S5–S7).

Trade-offs/synergies among ecosystem services of the WH sites

Trade-offs and synergies among ecosystem services varied considerably in space and time (Figs. 4, S8, and S9). In terms of spatial variation, strong trade-offs and weak synergies were mainly distributed in the buffer zone of the four WH sites, whereas weak trade-offs and strong synergies were distributed in both the property and buffer zones (Figs. 4, S8, and S9). In terms of temporal variation, there was little temporal change in the proportion of strong synergies across the four WH sites (Fig. 5a), whereas the proportion of weak synergies increased over time (Fig. 5b). However, the proportion of strong trade-offs decreased from 2000 to 2010 and then increased in 2020 in Shibing Karst and the two non-karst WH sites, while the trend was reversed in Libo-Huanjiang Karst (Fig. 5c). The proportion of weak trade-offs showed an inverse temporal trend compared to that of strong trade-offs (Fig. 5d).

Fig. 4
figure 4

Spatial differences in trade-offs and synergies among ecosystem services of the four World Heritage sites in Southwest China in 2020

Full size image
Fig. 5
figure 5

Temporal variation in the proportion of strong synergies (a), weak synergies (b), strong trade-offs (c), and weak trade-offs (d) among ecosystem services of the four World Heritage sites in Southwest China. One-way ANOVA was used to test the difference in the proportion of trade-offs/synergies between karst and non-karst World Heritage sites

Full size image

Weak trade-offs among ecosystem services dominated the four WH sites, with a mean proportion of 68.6%, followed by strong trade-offs (15.8%), strong synergies (9.8%), and weak synergies (5.8%) (Fig. 5). Compared to non-karst sites, karst WH sites had a significantly lower proportion of strong synergies and a significantly higher proportion of weak synergies (Fig. 5a, b).

Driving factors of ecosystem services of the WH sites

GDM showed that the 14 driving factors all had significant effects on the variation in CES across the four WH sites (p < 0.05), but with different contributions (Fig. 6a). Landscape division index had the largest effect (q = 0.63), followed by NDVI (q = 0.52), distance from road (q = 0.39), slope (q = 0.38), population density (q = 0.33), SOC (q = 0.28), DEM (q = 0.25), soil thickness (q = 0.22), bulk density (q = 0.19), MAP (q = 0.14), GDP (q = 0.13), MAT (q = 0.11), pH (q = 0.11), and nighttime light (q = 0.05). The interactive detection results indicated that the interaction between any two factors had a stronger effect than their individual influence, showing either a two-factor enhancement or a nonlinear enhancement (Fig. 6b). Furthermore, higher q-values were primarily found in the interactions with landscape division index, NDVI, distance from road, and population density. The most significant interactions were between landscape division index and NDVI (q = 0.76) and between landscape division index and distance from road (q = 0.74).

Fig. 6
figure 6

Results of factor detection (a) and interactive detection (b) for the driving factors influencing the combined ecosystem service of World Heritage sites in Southwest China. ‘ + ’ denotes a two-factor enhancement, whereas ‘*’ denotes a nonlinear enhancement

Full size image

SEM further revealed the causality and direction of different driving factors affecting CES (Fig. 7). Specifically, the landscape division index had a negative effect on CES, both directly and indirectly through affecting vegetation and soil factors. Topographic factors had a positive effect on CES, both directly and indirectly through affecting vegetation and soil factors. Climate factors had a direct positive effect on CES, but an indirect negative effect through affecting vegetation. Both vegetation and soil factors had a direct positive effect on CES. Human factors, both directly and indirectly through affecting vegetation, had a positive effect on CES, which was primarily driven by the positive effect of distance from road.

Fig. 7
figure 7

Structural equation modelling showing the direct and indirect effects of various driving factors on the combined ecosystem service (CES) of World Heritage sites in Southwest China. The model fit indexes show that the model has a very good fitting. Numbers adjacent to arrows are standardized path coefficients, all of which are significant (p < 0.05). Arrow width is proportional to the standardized coefficient. Blue and red arrows indicate positive and negative effects, respectively

Full size image

Discussion

General patterns of ecosystem services of the WH sites

Our study showed that the values of ecosystem services were generally higher in the property zone than in the buffer zone of the four WH sites in Southwest China. Given that anthropogenic factors (e.g., distance from road and population density) played an important role in affecting ecosystem services, this result could be attributed to the fact that the property zone is strictly protected, while the buffer zone is often subjected to anthropogenic threats such as agricultural and tourism activities [15]. Across the four WH sites and three periods, the mean values of HQ, CS, SR, and WC were 0.80, 118.06 t hm−2, 923.78 t hm−2, and 693.07 mm, respectively, which were generally higher than the mean values of these ecosystem services in non-WH areas [21, 62,63,64,65]. This result highlights that WH sites are irreplaceable resources for biodiversity and the basis for ecosystem services [11, 13, 66].

However, our study found that the mean values of HQ and CES decreased from 2000 to 2020 across the four WH sites. Such disappointing reports on WH do not only exist in China but also abound throughout the world in the context of rising threats and declining support [67]. These results suggest that while WH sites are hotspots of biodiversity and ecosystem service provision, they are experiencing a rapid decline in ecosystem services, probably as a result of intensified human activities. For example, we found that cropland area increased between 2000 and 2020 across the four WH sites (Table S6), particularly in Shibing Karst (from 2438.46 to 4187.97 hm2) and Chishui Danxia (from 1626.03 to 3425.94 hm2). Indeed, there is ample evidence that land use conversion from forest to cropland can degenerate ecosystem structure and functions, resulting in declined ecosystem services [21, 32, 68, 69]. On the other hand, the increasing intensity of tourism activities (e.g., a 50-fold increase in the number of tourists in Shibing from 2000 to 2020), and associated construction projects, will inevitably influence the structure and functions of WH ecosystems, thereby decreasing the provision of ecosystem services [14, 16].

Trade-offs and synergies among ecosystem services of the WH sites

In a limited ecosystem, the enhancement of the supply capacity of one ecosystem service is often at the expense of other services [46, 70]. This notion is supported by our result that the four WH sites were dominated by trade-offs among ecosystem services, although they represent the most intact and natural ecosystems with the highest biodiversity on Earth. This finding is also consistent with previous studies reporting that trade-offs among ecosystem services are common [53, 71, 72]. In terms of temporal variation, we found that the proportion of weak synergies generally increased from 2000 to 2020 across the four WH sites. According to our methods, weak synergies meant a low level of at least one service and a lack of services at a high level. Therefore, an increasing proportion of weak synergies indicates the deterioration of ecosystems. This may be because the degeneration of one service can decrease the ability of ecosystems to support other services, resulting in a synergistic decline in ecosystem services [73, 74]. In terms of spatial variation, we found that strong trade-offs and weak synergies were mainly distributed in the buffer zone of the four WH sites. This pattern may result from the intensive human activities that affect ecosystem services and their relationships. As areas with weak synergies were mostly distributed in the buffer zone and increased over time, our study suggests that managers should take timely and effective action in these areas to generate strong synergies among ecosystem services, thereby protecting the outstanding universal value, authenticity, and integrity of WH sites.

Differences in ecosystem services between the karst and non-karst WH sites

Our study showed that non-karst WH sites provided higher levels of HQ, CS, SR, and CES than karst WH sites in Southwest China. This result aligns with previous studies reporting that non-karst areas generally provide higher levels of ecosystem services than karst areas [21, 75, 76]. Furthermore, the spatial heterogeneity of ecosystem services was significantly higher in karst than in non-karst WH sites, which may result from the high landscape fragmentation and heterogeneity that characterize karst ecosystems [77, 78]. Our driving factor analyses also suggested that the landscape division index, an indicator of landscape fragmentation, was an important factor negatively affecting the provision of ecosystem services. Indeed, compared to non-karst landscapes, karst landscapes are more separate and fragmented due to complex water–rock interactions, which can reduce ecosystem connection and functions, thereby limiting the provision of ecosystem services [33, 47, 79]. Furthermore, we found that the interaction between landscape division index and distance from road had the largest effect on ecosystem services. This suggests that intensive human activities could exacerbate the negative impact of landscape fragmentation on ecosystem services of karst WH sites. Therefore, controlling human activities should be emphasized to prevent the decline of ecosystem services in karst WH sites with high landscape fragmentation. While not directly explored in our study, Zhang et al. [21] reported that non-karst areas have higher biodiversity than karst areas. This may also contribute to the higher levels of ecosystem services of non-karst WH sites because high biodiversity can increase the provision of ecosystem services through niche complementarity and sampling effects [80, 81]. Nevertheless, this explanation can also be supported by our results because non-karst WH sites had higher values of HQ, which is widely utilized as a reliable proxy for biodiversity [82,83,84].

Management implications for WH sites

A clear understanding of spatiotemporal changes and trade-offs/synergies in ecosystem services can have important implications for developing management efforts that balance ecological conservation and socio-economic development in WH sites. Our study showed that the values of ecosystem services were much higher in the property zone where primary forests exist than in the buffer zone where human activities are intensive. This finding highlights that ecological protection should always be given top priority in the development of WH sites. This is particularly important for the management of karst WH sites because karst ecosystems are more vulnerable to degradation and less likely to recover once degraded due to their special geomorphological and hydrogeological characteristics [8, 9].

Our study also showed that ecosystem services of the WH sites, especially HQ, have continued to decline and areas with weak synergies have continued to increase over the last two decades. These results may suggest that ecological protection and conservation have not been well implemented in the WH sites studied here, which is also indicated by the increased area of cropland (Table S6). Therefore, local governments should set up ecological red lines and persistently implement ecological restoration projects such as Natural Forest Protection Projects, the Grain for Green Program, and the Karst Rocky Desertification Restoration Project to effectively rewild the damaged habitats in WH sites [44]. While prohibiting agricultural activities, managers can also implement community-based conservation policies [85], such as establishing ecological compensation mechanisms and offering positions (e.g., tour guides and law enforcement rangers) to promote the job transition for indigenous residents in WH sites. These solutions could effectively address conflicts between WH management agencies and local communities [44, 66].

Moreover, as increasing tourism activities could be a major threat to the originality and integrity of WH ecosystems, industry transition from traditional tourism to ecotourism is a necessary choice to strike a balance between conservation and development in WH sites [12]. On the one hand, the number of tourists should be controlled so as not to exceed the carrying capacity of WH ecosystems [15, 18]. In this case, strategies that rest attractions and divert tourists can be employed to protect the originality and integrity of WH ecosystems. On the other hand, new projects and infrastructures should be properly planned and assessed to minimize their negative impacts on the outstanding universal value, authenticity, and integrity of WH. Due to the complexity and uniqueness of WH ecosystems, coupled with the protective properties of WH, it is necessary to consider various factors and involve multiple stakeholders to establish a comprehensive and optimal ecotourism strategy for WH sites [16].

Limitations and future directions

Although our study has identified the spatiotemporal patterns and driving factors of ecosystem services between karst and non-karst WH sites in Southwest China, several uncertainties and limitations need to be addressed in future studies. First, due to the special dual (surface and underground) hydrological structure of karst ecosystems, soil, nutrients, and surface water can leak and migrate to other areas along rock fissures, funnels, and sinkholes [2]. However, the InVEST model does not involve surface loss parameters, which may result in higher simulated ecosystem services of karst WH sites than the actual value. Therefore, future studies should incorporate the loss parameters to adjust the model for a more precise simulation. Second, despite the popularity of the InVEST model, it often estimates the annual mean values of ecosystem services without considering seasonal variability and feedback. Furthermore, we only investigated ecosystem services of the four WH sites at three time points. Due to the rapid climate change, ecosystem services may vary considerably between years [86, 87]. Therefore, future studies should evaluate ecosystem services at a higher time resolution to reduce the uncertainty in the temporal pattern. Finally, although the WH sites explored in our study have highly diverse forest types, we estimated ecosystem services based on a coarse land use classification. This may lead to uncertainties in the results because different forest types have different properties (e.g., carbon density, evapotranspiration, and runoff). Therefore, future studies should carefully differentiate vegetation types to improve the accuracy of results for WH sites.

Conclusions

In this study, we investigated the spatiotemporal differences, trade-offs/synergies, and driving factors of ecosystem services between karst and non-karst WH sites in Southwest China and provided relevant suggestions for WH ecosystem management. Importantly, our study showed that ecosystem services varied considerably in space and time, showing an overall higher value in the property zone than in the buffer zone of the WH sites and a decreasing trend over time. These variations in ecosystem services of the WH sites likely resulted from the increased cropland area and intensified tourism activities. Compared to non-karst sites, karst WH sites had significantly lower values of ecosystem services, which were mainly associated with their higher degree of landscape fragmentation and heterogeneity. Taken together, our findings could provide scientific support for decision-making aimed at reconciling conflicts between ecological conservation priorities and economic development demands, and then achieving optimal management and sustainable development of WH sites with different attributes.