Next Article in Journal
Size of Sand Grains Controls Pore Structure and Water Dynamics: Implications for Water Retention and Hydraulic Conductivity
Previous Article in Journal
KG-LLM Synergy for Intelligent Soil and Water Conservation Standard Governance
Previous Article in Special Issue
Strategic Framework to Reinforce the Application for the UNESCO Global Geopark Label: The Case of Chefchaouen Geopark (NW Morocco)
 
 
Search for Articles:
Title / Keyword
Author / Affiliation / Email
Journal
Article Type
 
 
Advanced Search
 
Section
Special Issue
Volume
Issue
Number
Page
 
Journals
Land
Volume 15
Issue 5
10.3390/land15050863
land-logo
Submit to this Journal Review for this Journal Propose a Special Issue
► Article Menu

Article Menu

share Share announcement Help format_quote Cite
first_page
settings
Order Article Reprints
Font Type:
Arial Georgia Verdana
Font Size:
Aa Aa Aa
Line Spacing:
Column Width:
Background:
Article

Conservation Effectiveness and Spatial Drivers of Qianjiangyuan National Park: Causal Evidence from a Quasi-Experimental Framework

by
Chuqi Wang
1,
Yinglin Wang
1,
Jiwen Lu
1 and
Liang Li
1,2,3,*
1
School of Landscape Architecture, Beijing Forestry University, Beijing 100083, China
2
School of National Park, Beijing Forestry University, Beijing 100083, China
3
School of Ecology and Nature Conservation, Beijing Forestry University, Beijing 100083, China
*
Author to whom correspondence should be addressed.
Land 2026, 15(5), 863; https://doi.org/10.3390/land15050863 (registering DOI)
Submission received: 2 April 2026 / Revised: 7 May 2026 / Accepted: 12 May 2026 / Published: 17 May 2026
(This article belongs to the Special Issue National Parks and Natural Protected Area Systems)
Browse Figures
Versions Notes

Abstract

National parks are widely recognized as a key spatial conservation strategy for simultaneously safeguarding biodiversity and sustaining ecosystem services, yet comprehensive and causally robust evaluation frameworks are still needed to accurately assess their effectiveness and support evidence-based management. This study evaluates the conservation effectiveness of Qianjiangyuan National Park (QJYNP) from 2015 to 2024 using a multidimensional index, a PSM-DID quasi-experimental framework, and interpretable machine learning. The results show that the direct policy effect was significantly positive during 2015–2020, but shifted to a negative cumulative effect by 2024. The spillover effect in the buffer zone also turned significantly negative, potentially associated with tourism-related development shifting outward. In addition, slope, temperature, and population density were identified as key drivers of EEI heterogeneity with nonlinear threshold effects, while road-related impacts intensified over time. These findings indicate that quasi-experimental approaches better capture phased policy effects than conventional descriptive comparisons, and suggest that simple boundary controls are insufficient; instead, buffer zones should be incorporated into integrated management frameworks to mitigate external development pressures.

1. Introduction

Biodiversity conservation is a core issue in global sustainable development [1,2]. The Global Biodiversity Framework proposes achieving the “30 × 30” target by 2030, which aims to protect 30% of the world’s land and marine areas [3]. National parks are legally designated protected areas (PAs) primarily aimed at conserving large-scale natural ecosystems of national significance. The establishment of national parks is regarded as a key spatial policy tool for coordinating regional ecological conservation and curbing biodiversity loss [4]. To achieve this ambitious goal, China launched its first 10 national park system pilot projects between 2015 and 2019. However, simply demarcating national park boundaries does not necessarily guarantee ecosystem conservation; achieving conservation outcomes is largely constrained by protection gaps [5,6,7], the impacts of external urbanization and internal development [8,9], and the extent of natural resource exploitation [10]. These challenges necessitate rigorous assessments to verify whether national parks effectively fulfill their intended conservation objectives.
National parks serve dual objectives of conserving biodiversity and providing ecosystem services [10,11,12]. Accordingly, increasing attention has been paid to evaluating conservation outcomes based on these two objectives. For biodiversity conservation, some studies have assessed conservation effectiveness using species-based indicators, such as habitat suitability or connectivity [13,14], but overlook the impact of human pressure [15]. In addition, ecosystem integrity is regarded as a fundamental basis for biodiversity conservation. It refers to the capacity of an ecosystem to maintain its composition, structure, and function within natural ranges of variation relative to a baseline state [16,17]. High-quality ecosystems are usually characterized by optimized landscape patterns and vigor [18]. Therefore, related studies have introduced the concept of ecosystem health [19] and used the vigor–organization–resilience paradigm to quantify the complex dimensions of ecosystem integrity [20,21]. Landscape metrics are frequently employed to describe ecosystem organization [22], while net primary productivity (NPP) and aboveground biomass (AGB) are used to describe ecosystem vigor [23,24]. In terms of ecosystem service assessment, many studies have used the InVEST model to evaluate habitat quality, water conservation, carbon sequestration, and other ecosystem service capacities of national parks [25,26,27], while others have examined the impacts of national park establishment on socio-economic development outcomes [28]. Biodiversity conservation and ecosystem services are closely linked; however, existing studies have often focused on a single conservation objective, and relatively few have incorporated ecosystem health and ecosystem services into a unified framework for evaluating national park conservation effectiveness [29,30].
In terms of comparative approaches, most studies identifying the policy effects of national parks have relied on traditional “before-and-after” or “inside-and-outside” comparisons [30]. However, such comparisons may produce biased results because protected areas are usually not randomly established, but are often located in areas with naturally favorable ecological conditions. Therefore, conservation-effectiveness evaluation has increasingly shifted toward more robust quasi-experimental designs [31,32]. Establishing comparable unprotected control areas as counterfactual references to identify the net effects of conservation policies has become the mainstream approach in conservation project evaluation [33,34]. A seminal study demonstrated that the actual contribution of protected areas in Costa Rica was a reduction of approximately 10% in potential deforestation, far below the baseline level of 44% [33]. Some studies have used propensity score-matching models to examine the effectiveness of China’s national parks in providing ecosystem services and forest conservation [35,36,37]. Meanwhile, national park establishment may generate spillover effects beyond park boundaries. Existing studies have mainly focused on spillover effects related to ecosystem service values or regional economic development [38,39], whereas relatively few have used matching-based approaches to evaluate the spillover effects of national park establishment on ecological conservation outcomes.
The quasi-experimental methods mainly answer whether national park establishment has generated significant net policy effects, but they provide limited explanation of why such effects emerge and which natural or anthropogenic factors constrain the sustained improvement of conservation outcomes. For national park management, identifying only the direction and magnitude of policy effects is insufficient; it is also necessary to reveal the driving mechanisms underlying the formation and change in conservation outcomes. In densely populated national parks, community livelihoods, tourism development, resource use, and surrounding land-use transitions can redistribute human pressures and thereby affect ecological integrity across park and buffer zones [40]. The outcomes of national parks are driven by interactions among conservation institutions, management interventions, human activities, and baseline environmental conditions [11]. The effects of these factors on ecological outcomes are often nonlinear and spatially heterogeneous. Compared with conventional statistical methods, machine-learning models have advantages in interpreting complex nonlinear mechanisms, and interpretable machine-learning approaches have increasingly been used to identify the driving mechanisms of ecosystem health and ecological quality [41]. However, functional zones within and around national parks often differ substantially in conservation intensity, human activity pressure, and land-use patterns. Few studies have further compared whether the driving mechanisms of conservation outcomes differ among different zones of national parks.
To address the identified research gaps, this study advances beyond single-dimension assessments and conventional before-after comparisons by constructing a multi-dimensional EEI, applying a PSM-DID framework to identify both direct and spillover effects, and employing interpretable machine learning to quantify spatial driving mechanisms. Specifically, this study aims to answer the following three questions: (1) What are the spatiotemporal dynamics of ecological conservation outcomes in QJYNP and its surrounding areas from 2015 to 2024, before and after the establishment of the national park? (2) What are the direct and spillover effects of QJYNP’s establishment on ecological conservation effectiveness within the park and in the buffer zone? (3) Which natural and anthropogenic factors dominate the spatial heterogeneity of EEI in the study area, and do their driving mechanisms differ between inside and outside the park?

2. Materials and Methods

2.1. Study Area

As shown in Figure 1, the QJYNP pilot area is located in Kaihua County, Quzhou City, Zhejiang Province, China, covering a total area of 252 km2. It was officially designated as one of China’s first national park pilot sites in the second half of 2015. The pilot area was established by integrating existing protected areas, including the Gutianshan National Nature Reserve and Qianjiangyuan National Forest Park [42], together with ecological corridors connecting these areas. QJYNP is located within the Huangshan–Huaiyushan Mountains region, one of China’s 32 key Biodiversity Conservation Priority Areas, and preserves a large, well-preserved area of low-altitude mid-subtropical evergreen broad-leaved primary forest. The QJYNP also provides important habitat for first-class nationally protected species, including the black muntjac(Muntiacus crinifrons) and Elliot’s pheasant (Syrmaticus ellioti), both of which are listed as Near Threatened by the IUCN [43]. Currently, the Chinese government has designated a Core Protection Zone of 72 km2 within the park, where all human activities are strictly prohibited except for scientific research. As the headwaters of the Qiantang River, QJYNP plays a crucial role in soil conservation and carbon sequestration [44].
As one of the most densely populated national park pilot areas in China, the QJYNP provides a representative case for examining conservation effectiveness, human activity pressures, and spillover effects in a coupled socio-ecological system. Following similar methodologies [36], the spatial boundary of the national park’s ecological impact was estimated at approximately 12 km, based on an analysis of the spatial proportions of construction and forest land within the buffer zone (Supplementary Materials S1, Tables S1 and S2). Accordingly, the study area was divided into three zones: (1) the QJYNP (NP); (2) the 0–12 km buffer zone outside the national park (Buffer), which is the adjacent area affected by ecological spillover effects; and (3) the 12–24 km control zone outside the national park (Out), which is an area subject to neither conservation policy constraints nor the spillover effects of the national park (Figure 1).

2.2. Research Framework and Data Sources

The basic framework of this study is shown in Figure 2. Considering that QJYNP is characterized by a densely populated low-altitude evergreen broad-leaved forest landscape facing combined pressures from agricultural land use, infrastructure expansion, and ecotourism development, a single indicator is insufficient to capture its multifaceted conservation status. Following the ecosystem health framework originally proposed by Costanza [45] and subsequently adapted for national park conservation assessment by Zhu [46], conservation effectiveness index (EEI) is evaluated across three dimensions: ecosystem vigor (EV), ecosystem organization (EO), and ecosystem services (ES). To further reveal the spatial drivers underlying EEI heterogeneity across management zones, drawing on recent applications of interpretable machine learning in ecological and conservation research [47], this study employs a random forest model combined with SHAP values to identify the relative importance and nonlinear directional effects of driving factors on EEI within and beyond the national park [47,48]. Compared to traditional methods such as linear regression, random forests can effectively capture the nonlinear relationships between driving factors and the EEI [48,49]. To isolate the causal effects of national park establishment from confounding factors arising from non-random park siting, we applied a Propensity Score Matching combined with Difference-in-Differences (PSM-DID) quasi-experimental framework to construct spatially comparable treatment and control groups and identify net policy effects [34,35]. The datasets used in this study are listed in Table 1. Specifically, land-use, vegetation growth, and climate data are used to calculate ecological conservation effectiveness, while topography, nighttime lights, population density, and transportation data are used to support counterfactual assessments and driver analysis. All data were projected to the CGCS2000/3-degree Gauss-Kruger zone 39 coordinate system. A 500 m × 500 m grid was used, and the average of the data within each grid cell was calculated to represent the level of each analysis unit.

2.3. Construction of the Ecological Conservation Effectiveness Index

EO was characterized using landscape metrics that capture fragmentation, shape complexity, and diversity. These metrics were calculated using a moving-window approach with a 30 m grain size and an 840 m rectangular window, the size of which was determined based on semivariogram analysis to capture the spatial autocorrelation range of landscape patterns in the study area (Supplementary Materials, Figure S1). After addressing multi-collinearity via Variance Inflation Factor (VIF) screening, three representative metrics were retained for analysis: Edge Density (ED), Mean Perimeter-to-Area Ratio (PARA_MN), and Shannon’s Diversity Index (SHDI) [16]. Ecosystem vigor (EV) reflects regional vegetation productivity, biomass accumulation, and the capacity to maintain biodiversity, which can be quantified using NDVImax, aboveground biomass, and the biodiversity maintenance function index. In this study, NDVI_MAX and aboveground forest biomass were selected to represent vegetation growth status. Additionally, in subtropical low-altitude regions, low-altitude areas with abundant precipitation help maintain habitats for biodiversity. Therefore, following established methodologies, the biodiversity maintenance capacity was calculated using the following formula [50]:
B M F = N P P m e a n   × F p r e × F t e m   × F a l t
where B M F is the Biodiversity Maintenance Function Index; N P P m e a n     is the multi-year average Net Primary Productivity; F p r e   is the normalized multi-year average precipitation factor; F t e m is the normalized multi-year average temperature factor;   F a l t is the normalized elevation factor. To avoid extreme cases, F p r e   and F t e m   select the average values of the study year and the adjacent years as the reference.
As a biodiversity hotspot in the Yangtze River Delta region and the source of Zhejiang’s largest river, QJYNP plays a vital role in habitat maintenance, water resource conservation, and carbon sequestration [51,52]. To quantify ecosystem service provision, this study used 30 m resolution raster data and the InVEST (v3.17.1) model to estimate three key services: habitat quality, water yield, and carbon storage. Habitat quality was assessed using the Habitat Quality module, which evaluates habitat suitability based on land-use types while incorporating the intensity of threat sources, their effective impact distances, and distance-decay functions. Detailed model descriptions and mathematical formulations for the three ecosystem service modules are provided (Supplementary Materials, Text S1). Habitat sensitivity to different threats was also considered, with model parameters and weights calibrated based on previous regional studies [47,48,49] (Supplementary Materials, Tables S3 and S4). Water yield was estimated using the InVEST Water Yield module, which calculates the multi-year average annual water production for each land-use unit. To better align with the local hydrological characteristics of Quzhou, the seasonal constant Z was calibrated from its default setting of 1.5 to 0.8 (2015), 0.8 (2020), and 1.0 (2024), based on the calculated water yield coefficients for each period. Carbon storage was quantified using the InVEST Carbon Storage module, which estimates ecosystem carbon stocks based on vegetation biomass and soil organic matter. Carbon pool density parameters for each land-cover type were derived from published data on China’s subtropical forests (Supplementary Materials, Table S5) [46].
After normalizing all raw indicators and extracting 500 m grid means via nearest-neighbor analysis, we applied Principal Component Analysis (PCA) separately to each dimension for each study year, using the first principal component (PC1) as the integrated score for EO, EV, and ES, following the methodologies [50,51,52]. For all three study years, the eigenvalue of PC1 exceeded 1, and the variance explained by PC1 was greater than 60% in each dimension, indicating that PC1 effectively captured the dominant information of the original indicators. To ensure consistency in evaluation criteria across time periods, Grey Relational Analysis (GRA) was employed to assign weights to the three dimensions by pooling standardized data from 2015, 2020, and 2024 (Table A1). GRA quantifies the proximity of each indicator to a reference sequence representing the ideal optimal state, with the distinguishing coefficient set to ρ = 0.3. To mitigate the influence of extreme values, the 95th percentile (P95) of standardized indicators was adopted as the reference sequence rather than the conventional maximum value. A sensitivity check using the 90th percentile (P90) as an alternative reference sequence yielded identical dimension weight rankings, confirming the robustness of the weighting results to the choice of reference threshold. This model can be expressed as:
E E I j = ∑ k = 1 3   w k × X j k
where E E I j denotes the Ecological Conservation Effectiveness Index for the j -th grid cell; w k is the relative weight assigned to the k -th dimension of ecosystem services (i.e., EO, EV, or ES); and X j k represents the normalized value of the k -th dimension in grid cell j .

2.4. Identification of Driving Factor Importance Based on the RF-SHAP Model

Given that QJYNP is embedded in a heterogeneous landscape where topographic, climatic, and anthropogenic drivers interact non-linearly with ecological outcomes, and that this study aims to identify dominant drivers and their threshold effects rather than to make spatial predictions, the RF–SHAP combination was selected as the most appropriate approach for the third research question. The random forest model uses EEI as the dependent variable and nine independent variables: topography (dem, slope), climate (tem, prec), population (pop), nighttime light (ntl), and road and water body accessibility (NEAR_road, NEAR_water, Road_Density) [53,54]. Population density data was truncated at the 99th percentile and then transformed using the natural logarithm (ln(pop + 1)) to reduce the impact of outliers [55]. The model was trained on 80% of the samples and tested on 20%. Model fit was evaluated using R2, RMSE, and MAE on the test set. The model achieved an R2 of approximately 0.60 on the test set. The RF models showed moderate but acceptable explanatory performance across the three years, with R2 values of 0.5824, 0.6036, and 0.5990, RMSE values of 0.0965, 0.0782, and 0.0547, and MAE values of 0.0779, 0.0627, and 0.0436, respectively. The stability of these performance metrics across three independently fitted yearly models indicates that the identified driver–EEI relationships are temporally robust rather than period-specific. Given the potential spatial autocorrelation in the grid-based dataset, the RF-SHAP analysis was used primarily to interpret the relative importance, direction, and nonlinear response patterns of the driving factors, rather than to pursue high-accuracy spatial prediction of EEI. Grounded in the Shapley value principle [56], this study used TreeExplainer to compute the additive marginal contribution of each driver to EEI predictions. Global feature importance was determined by the mean absolute SHAP values, while SHAP dependency plots were employed to characterize the directional effects and nonlinear response patterns of each factor [41,56,57].

2.5. Identifying Conservation Effectiveness via Propensity Score Matching

Using a Logit model in R (v4.2.2), propensity scores were estimated based on six covariates: elevation, slope, precipitation, population, distance to roads, and distance to water bodies. A 1:1 nearest-neighbor matching procedure was then performed using the MatchIt package, with a caliper of 0.2 to restrict matches to comparable treatment and control units. Matching quality was assessed using standardized mean differences (SMDs) before and after matching, with an absolute SMD below 0.1 considered to indicate acceptable covariate balance between the treatment and control groups (Supplementary Materials, Figure S2). Subsequently, a two-period DID estimation was applied to the matched samples to isolate policy-driven effects. Two comparison schemes were established: (1) direct effects (NP vs. Out) and (2) spillover effects (Buffer vs. Out). These effects were evaluated across two windows to capture phased dynamics: the early stage (2015–2020) and the mid-term cumulative stage (2015–2024). The model is formulated as follows:
Y i t = α + β g T r e a t i g × P o s t t + μ i + λ t + ε i t
where Y i t is the Ecological Conservation Effectiveness Index (EEI) for the grid cell i in year t ; T r e a t   i g is a dummy variable indicating whether the grid cell i belongs to the treatment group g (where NP or Buffer = 1, Out = 0); P o s t t is policy dummy variable; μ i and λ t represent the grid cell and year fixed effects; and β g is the coefficient of interest, representing the Average Treatment Effect on the Treated (ATT) for the group g .
The ATT is formally defined as ATT = E[Y(1) − Y(0)|Treat = 1], that is, the expected difference in EEI between the treated and untreated states for grid cells that actually received the policy treatment, and is interpreted as the causal effect of QJYNP establishment net of confounding factors. The statistical significance of β g was assessed using its associated p-value derived from the standard error of the DID estimator, with ** and *** denoting significance at the 1% and 0.1% levels, respectively, and 95% confidence intervals reported as β g ± 1.96 × SE. To facilitate cross-zone comparison, ATT estimates were further expressed as relative effects, calculated as the ratio of ATT to the mean baseline EEI of the control group (Out) in 2015.

3. Results

3.1. Analysis of the Spatial Patterns and Spatiotemporal Dynamics of Ecological Conservation Outcomes

The spatial distribution of ecological conservation effectiveness indicators in 2015, 2020, and 2024 exhibits distinct spatial heterogeneity (Figure 3a). Specifically, high-value areas for EO were primarily located within national parks and to the south of their boundaries; however, by 2024, the extent of these high-value areas had significantly shrunk, and the overall rating tended to decline; EV high-value areas were predominantly situated in the northern and northwestern parts of the study area, as well as in some peripheral regions. The spatial patterns in 2015 and 2020 were generally similar, whereas by 2024, high-value areas had significantly decreased; ES was most prominent within the national park boundaries, with high-value areas exhibiting strong continuity, but by 2024, the overall grade had declined, and the extent of high-value areas had also contracted significantly. The trend of EEI aligns with that of the individual indicators, exhibiting a spatial gradient characterized by relatively higher values within the park and relatively lower values outside the park, with a more pronounced overall decline evident in the latter part of the study period. To quantify the pattern of change, this study classified the 2015–2024 changes into three categories (increase, stability, and decrease) based on standard deviation thresholds, and calculated the proportion of each category across all grid cells (Figure 3b). Further analysis of the 2015–2024 change types reveals that “stability” and “decline” were the predominant patterns across all indicators, with the proportion of declining grids generally following the order: EO > EV > ES > EEI.
We compared conservation effectiveness across the three regions by analyzing the average values of EO, EV, ES, and the composite index EEI for each sub-region (Figure 4). Across all years, mean values generally followed the pattern NP > Buffer > Out, suggesting that the level of ecological conservation effectiveness in national parks is overall higher than in buffer zones and external areas. From a temporal perspective, the EEI declined overall across the three sub-regions from 2015 to 2024, with a more pronounced decline during 2020–2024; however, the ranking of sub-regions remained generally consistent.

3.2. Analysis of the Importance and Mechanisms of Drivers Shaping the Spatial Patterns of Ecological Conservation Outcomes

There are significant differences in the relative importance of the nine driving factors on the spatial pattern of the EEI (Supplementary Materials, Figure S3). Slope, temperature, population density, and elevation are the dominant factors influencing the spatial variation in EEI, with their mean absolute SHAP values far exceeding those of the remaining variables. The importance of road-related factors shows an overall upward trend, indicating that the disturbance of the ecological pattern by the expansion of the transportation network has been continuously increasing during the study period. The importance of NEAR_water and ntl is relatively limited (Figure 5a,b).
As illustrated in Figure 5c, slope exhibits a monotonic positive relationship with EEI; grid cells with low slopes below 3.5° show significantly negative SHAP values, while positive contributions increase continuously with slope after exceeding this threshold. Elevation displays a pronounced inverted U-shaped response, with positive SHAP contributions concentrated within the 200–800 m range, while negative effects emerge at lower and higher altitudes. Precipitation generally exhibits a positive relationship with EEI, with the zero-crossing point rising from approximately 1202 mm in 2015 to approximately 1836 mm in 2024. Temperature generally exhibits a positive relationship with EEI; SHAP values for high-temperature grids are generally positive, while those for low-temperature grids are slightly negative. Dependency plots reveal a distinct nonlinear threshold effect. Among human activity-related factors, population density exerts a negative inhibitory effect on the EEI and exhibits nonlinear characteristics; once a threshold is exceeded, the negative SHAP values increase sharply, with the maximum negative contribution reaching −0.10 in 2024. The threshold decreased from approximately 1096 people/km2 in 2015 to approximately 202 people/km2 in 2024. Within the QJYNP, grid cells with extremely low population (pop) values exhibit SHAP values concentrated in the positive range; in contrast, grid cells with high population in the control area show the most prominent negative contribution. Distance from roads exhibits an inverted U-shaped response over the three-year period: SHAP values are negative within 50 m of either side of the road, while positive contributions increase beyond 50 m, peaking at approximately 1800 m. Road density exhibits an overall negative relationship with EEI; SHAP values for high-road-density grids remain consistently negative, with their importance rising from last place in 2015 to sixth place in 2024.

3.3. Evaluation of the Direct Policy Effects of the Establishment of QJYNP

Following PSM, 959 pairs of spatially comparable grid samples were constructed. Post-matching diagnostics confirmed that standardized mean differences for all covariates converged significantly, with absolute values remaining below the 0.1 threshold. PSM-DID analysis based on change measures identified the following policy effects (Table 2): In the initial period, the ATT of the treatment group relative to the control group was 0.013. This indicates that, compared to unprotected areas, the treatment-induced changes in EEI within national parks were more favorable, representing a 1.03% relative improvement over the control group’s 2015 baseline. By the mid-term of the policy, the ATT of NP relative to Out shifted to −0.016, corresponding to a cumulative relative decline of approximately 1.23%; this down turn was primarily driven by significant negative transitions during the 2020–2024 period. We further estimated the ATT for the sub-dimensions EO, EV, and ES separately on the matched samples. The results show that the relative improvement in the EEI during the early policy period was primarily driven by positive changes in EV (ATT = 0.031) and ES (ATT = 0.004), while EO exhibited a significant negative trend (ATT = −0.048) during the same period, offsetting the composite index; thus, despite the decline in EO, the composite EEI still exhibited a significant positive relative effect. In contrast, during 2015–2024, the negative contributions of EO (ATT = −0.056) and ES (ATT = −0.020) were dominant. Although EV maintained a small positive net effect (ATT = 0.009), it was insufficient to offset the declines in EO and ES, resulting in a significantly negative cumulative direct effect of the EEI overall (Figure 6).

3.4. Evaluation of the Spillover Policy Effects of the Establishment of QJYNP

The PSM procedure yielded 5740 spatially comparable pairs between the Buffer and Out. According to Table 3 and Figure 7, the ATT for the treatment group (Buffer) relative to the control group (Out) during the early stages of the policy was 0.0007 (p = 0.319), indicating a non-significant effect; indicating an insignificant policy impact equivalent to only 0.05% of the 2015 baseline. In contrast, the mid-term cumulative effect (2015–2024) was significantly negative (ATT = −0.0124), representing a 1.00% decline relative to the baseline. This suggests that the establishment of a national park may be accompanied by negative spillover effects in adjacent buffer zones. Results for individual indicators show that during the initial phase of the national park’s establishment, EV was significantly positive (ATT = 0.0118), but both EO and ES were significantly negative during the same period. The positive and negative effects offset each other, resulting in a non-significant composite EEI in the initial phase (ATT = 0.0007), which was approximately 0.05% of the baseline level. However, during the 2020–2024 interval, EV shifted from positive to negative (ATT = −0.0129). During the mid-phase of the national park’s establishment, the sustained negative contributions of EO and ES remained dominant (EO: ATT = −0.0234, ES: ATT = −0.0136), while the cumulative effect of EV remained insignificant (ATT = −0.0011). Consequently, the cumulative spillover effect of the EEI was overall significantly negative (ATT = −0.0124), amounting to approximately −1.00% of the baseline level.

4. Discussion

4.1. Spatio-Temporal Evolution of National Park Conservation Outcomes and Policy Effectiveness

Using the calculation method based on the change in ATT relative to the baseline, we compared the differences in conservation effectiveness changes between unmatched and matched calculations (Table 4). Regarding direct effects, unmatched data suggest that the EEI decline within the NP was nearly identical to the Out group during 2015–2020, but the NP experienced a 4.10% greater decline than the Out group over the 2015–2024 period. However, PSM-DID identification reveals a more nuanced trajectory: the ATT was significantly positive during the early stage (+1.03% relative to the baseline), before transitioning to a negative cumulative effect (−1.23%) by 2024. This indicates that simple mean comparisons may overestimate the relative decline within the park by failing to account for initial selection bias.
Regarding spillover effects, raw observations show consistent trends between the Buffer and Out groups from 2015 to 2020, but a 0.81% sharper decline in the Buffer zone by 2024. After controlling for covariates via PSM-DID, the early-stage spillover effect remained insignificant, whereas the mid-term cumulative effect was confirmed to be significantly negative (−1.00%). Compared to unmatched observations, the PSM-DID estimates provide a more rigorous estimate of effect size, clearly isolating the phased characteristics of governance performance and the emergence of medium-to long-term negative spillovers in the buffer zone.
The significance of these findings extends beyond the numerical estimates themselves. Methodologically, the contrast between unmatched and matched results shows that conventional comparisons can simultaneously mask early-stage policy gains and overstate mid-term losses, whereas PSM-DID reveals a phased trajectory in which the policy effect transitions from significantly positive to significantly negative, alongside a delayed emergence of negative spillovers in the buffer zone. By integrating multi-dimensional EEI assessment with quasi-experimental causal inference, this study offers a transferable analytical framework for evaluating phased policy effects in other national parks subject to pronounced surrounding urbanization and area-based conservation policies.

4.2. Driving Mechanisms of Spatial Variation in Ecological Conservation Outcomes

The RF-SHAP analysis reveals the evolving drivers of spatial heterogeneity in the EEI. Among natural factors, slope consistently maintained high importance over the three-year period, with its threshold (approximately 3.5–4.4°) remaining largely stable. This is primarily because flat areas with low slopes have long been used for agricultural production and construction development, resulting in persistently low ecological quality. In contrast, steep-slope areas have retained relatively intact natural vegetation cover due to high development costs, which aligns with the general patterns of ecological landscapes in mountainous protected areas. Elevation exhibited a stable inverted U-shaped response pattern, with the 200–800 m range acting as a positive contributor, closely aligning with the primary distribution range of low-elevation evergreen broad-leaved forests in the study area; areas below 200 m are characterized by dense human activity, while those above 800 m tend to have sparse vegetation, with both extremes exerting a suppressing effect on the EEI. The positive driving effect of precipitation persisted over the three-year period, but the threshold at which precipitation began to exert a positive effect increased from approximately 1202 mm to 1836 mm. This likely reflects that, against the backdrop of climate warming, the moisture requirements for the study area’s ecosystem to maintain a high-quality state have increased, causing the ecological benefit threshold for precipitation to shift upward. Furthermore, the rising importance of temperature and road-related factors underscores the escalating cumulative ecological costs of regional infrastructure expansion and baseline climatic shifts over time.

4.3. Causes of Declining National Park Conservation Effectiveness and Negative Spillover Effects

PSM-DID results reveal that the direct effect was significantly positive in the early pilot phase, but turned negative in the mid-phase (2015–2024), while the buffer zone also exhibited a significant cumulative negative spillover effect (ATT = −0.0124). These two phenomena are not isolated but collectively reflect the complex pressure structure facing the study area following national park establishment. RF-SHAP analysis indicates that grid cells with higher temperatures and more abundant precipitation generally exhibited higher EEI values during the same period; however, the overall decline in EEI from 2020 to 2024 was not attributable to shifts in climatic conditions but rather to changes in transportation infrastructure and land use patterns.
The study area is characterized by traditional agricultural production and dense human activity, and it also serves as a major ecotourism destination, including sites such as Wuyuan and Huangshan. Following the formal establishment of the national park, the construction of tourism service facilities within the park was strictly restricted; however, the branding effect of national park designation has instead significantly driven the rapid expansion of the surrounding tourism industry, directly leading to the expansion of construction land and increased landscape fragmentation within the buffer zone. Statistical data indicate that the tourism industry in the three main counties where the study area is located expanded substantially from 2015 to 2024, with annual tourist visits increasing from 22.70 million to 45.55 million and total tourism revenue rising from 17.1 billion CNY to 41.3 billion CNY. Land use analysis within the buffer zone shows that construction land expanded from 4.28 km2 in 2015 to 4.83 km2 (+12.9%), while forested land decreased by 14.07 km2 over the same period, providing direct evidence of intensifying infrastructure development following national park establishment. Tourism-related activities have become a major component of local livelihoods, potentially driving infrastructure expansion and land-use change in areas surrounding the national park. As a result, development pressure may have shifted from the strictly protected core area to the buffer zone, contributing to the observed negative spillover effects [58]. The decline in EEI and the emergence of negative buffer zone spillover fundamentally reflect the spatial reorganization of human activities triggered by national park establishment, with climatic spatial variation serving as the natural backdrop rather than the primary driver of temporal changes.

5. Conclusions

By integrating the VOS framework, causal inference, and interpretable machine learning models, this study assessed the spatiotemporal dynamics, net policy effects, and driving mechanisms of conservation performance in the QJYNP from 2015 to 2024. The results showed that, while the potential influence of regional development stage shifts and anomalous years cannot be fully excluded, the establishment of QJYNP initially generated a significant positive direct effect (+1.03%), which turned negative in the mid-term (−1.23%). Spillover effects in the buffer zones were not significant in the early stage but became significantly negative in the mid-term. This early-stage positive effect would remain undetectable through conventional time-series comparisons, given the overall declining trend, highlighting the practical value of quasi-experimental methods for providing evidence-based justification for national park establishment and offering a transferable evaluation framework for other national parks facing similar urbanization and development pressures. Geographical and climatic factors were the dominant drivers of spatial heterogeneity in the EEI, while the importance of road-related factors continued to rise, and the ecological suppression threshold of population density decreased significantly. The spatial shift in tourism-related development toward the buffer zone is likely an important mechanism underlying the negative spillover effect. It is imperative to incorporate the buffer zone into a unified management framework and strengthen spatial guidance and intensity control of tourism development, thereby achieving the dual objectives of conservation benefits within the park and coordinated regional development.
Although this study provides useful evidence for evaluating conservation effectiveness in QJYNP, several limitations should be acknowledged when interpreting the findings. First, due to data availability, this study relies on three temporal cross-sections, which limits the ability to test the parallel trends assumption underlying the PSM-DID framework. As a result, the estimated policy effects provide useful quasi-causal evidence, but should be interpreted with caution due to data limitations. Future research could integrate species surveys and socioeconomic data to construct a more comprehensive evaluation system and introduce spatial cross-validation to improve the reliability of model evaluation.

Supplementary Materials

The following supporting information can be downloaded at: https://www.mdpi.com/article/10.3390/land15050863/s1, Table S1: Statistical analysis of forest area changes within the 0–20 km circular buffer zone; Table S2: Statistics on the proportion of construction land in the circular buffer zone; Table S3: Habitat suitability and sensitivity of LULC types to threats; Table S4: Characteristics and weights of habitat threat factors; Table S5: Carbon density (t/ha) for each LULC type and carbon pool; Figure S1: Semivariogram-based determination of the optimal moving window size for landscape metric calculation; Figure S2: Assessment of covariate balance and propensity score distribution before and after matching, (a) Covariate balance, (b) Propensity score distribution chart; Figure S3: Performance and interpretability analysis of the RF model for EEI; Text S1: The formulas used in the InVest model.

Author Contributions

Conceptualization, C.W. and L.L.; methodology, C.W.; writing—original draft preparation, C.W.; writing—review and editing, J.L.; visualization, Y.W.; supervision, L.L. All authors have read and agreed to the published version of the manuscript.

Funding

This research was funded by the following projects: the Project of Joint Laboratory for International Cooperation on Capacity Building of National Parks, grant number 2025-HXZX-038; the Science Popularization Project of the National Park Development Center, National Forestry and Grassland Administration of China, grant number 20261710706; and the Fundamental Research Funds for the Central Universities, grant number PTPY202605.

Data Availability Statement

Dataset is available upon request from the authors.

Conflicts of Interest

The authors declare no conflicts of interest.

Abbreviations

The following abbreviations are used in this manuscript:
QJYNPQianjiangyuan National Park
PAProtected Area
PSMPropensity Score Matching
DIDDifference-in-Difference
RFRandom Forest
SHAPSHapley Additive exPlanations
ATTAverage Treatment Effect on the Treated
EVEcosystem Vigor
EOEcosystem Organization
ESEcosystem Service
EEIEcological Conservation Effectiveness Index

Appendix A

Table A1. Weights of EO, EV, and ES dimensions based on GRA.
Table A1. Weights of EO, EV, and ES dimensions based on GRA.
DimensionsEOEVES
Weight0.25670.38020.3631

References

  1. Turner, B.L.; Lambin, E.F.; Reenberg, A. The Emergence of Land Change Science for Global Environmental Change and Sustainability. Proc. Natl. Acad. Sci. USA 2007, 104, 20666–20671. [Google Scholar] [CrossRef]
  2. Maxwell, S.L.; Cazalis, V.; Dudley, N.; Hoffmann, M.; Rodrigues, A.S.L.; Stolton, S.; Visconti, P.; Woodley, S.; Kingston, N.; Lewis, E.; et al. Area-Based Conservation in the Twenty-First Century. Nature 2020, 586, 217–227. [Google Scholar] [CrossRef] [PubMed]
  3. Convention on Biological Diversity. Kunming-Montreal Global Biodiversity Framework; Secretariat of the Convention on Biolog-ical Diversity: Montreal, QC, Canada, 2022; Available online: https://www.cbd.int/gbf (accessed on 30 March 2026).
  4. Heinen, J.T. International Trends in Protected Areas Policy and Management. In Global Issues and Trends in the Protection of Natural Areas; Sladonja, B., Ed.; InTech: Rijeka, Croatia, 2012; p. 18. ISBN 978-953-51-0697-5. [Google Scholar]
  5. Barnes, M.D.; Craigie, I.D.; Dudley, N.; Hockings, M. Understanding Local-Scale Drivers of Biodiversity Outcomes in Terrestrial Protected Areas. Ann. N. Y. Acad. Sci. 2017, 1399, 42–60. [Google Scholar] [CrossRef]
  6. Stoner, C.; Caro, T.; Mduma, S.; Mlingwa, C.; Sabuni, G.; Borner, M. Assessment of Effectiveness of Protection Strategies in Tanzania Based on a Decade of Survey Data for Large Herbivores. Conserv. Biol. 2007, 21, 635–646. [Google Scholar] [CrossRef]
  7. Brodie, J.F.; Mohd-Azlan, J.; Chen, C.; Wearn, O.R.; Deith, M.C.M.; Ball, J.G.C.; Slade, E.M.; Burslem, D.F.R.P.; Teoh, S.W.; Williams, P.J.; et al. Landscape-Scale Benefits of Protected Areas for Tropical Biodiversity. Nature 2023, 620, 807–812. [Google Scholar] [CrossRef]
  8. Zhang, L.; Luo, Z.; Mallon, D.; Li, C.; Jiang, Z. Biodiversity conservation status in China’s growing protected areas. Biol. Conserv. 2017, 210, 89–100. [Google Scholar] [CrossRef]
  9. Jones, K.R.; Venter, O.; Fuller, R.A.; Allan, J.R.; Maxwell, S.L.; Negret, P.J.; Watson, J.E.M. One-Third of Global Protected Land Is under Intense Human Pressure. Science 2018, 360, 788–791. [Google Scholar] [CrossRef] [PubMed]
  10. Zhang, Y.; West, P.; Thakholi, L.; Suryawanshi, K.; Supuma, M.; Straub, D.; Sithole, S.S.; Sharma, R.; Schleicher, J.; Ruli, B.; et al. Governance and Conservation Effectiveness in Protected Areas and Indigenous and Locally Managed Areas. Annu. Rev. Environ. Resour. 2023, 48, 559–588. [Google Scholar] [CrossRef]
  11. Heller, N.E.; McManus Chauvin, K.; Skybrook, D.; Barnosky, A.D. Including Stewardship in Ecosystem Health Assessment. Nat. Sustain. 2023, 6, 731–741. [Google Scholar] [CrossRef]
  12. Xu, W.; Xiao, Y.; Zhang, J.; Yang, W.; Zhang, L.; Hull, V.; Wang, Z.; Zheng, H.; Liu, J.; Polasky, S.; et al. Strengthening Protected Areas for Biodiversity and Ecosystem Services in China. Proc. Natl. Acad. Sci. USA 2017, 114, 1601–1606. [Google Scholar] [CrossRef]
  13. Bowker, J.N.; De Vos, A.; Ament, J.M.; Cumming, G.S. Effectiveness of Africa’s Tropical Protected Areas for Maintaining Forest Cover. Conserv. Biol. 2017, 31, 559–569. [Google Scholar] [CrossRef]
  14. Ma, Y.; Liu, L.; Yao, W.; Zeng, Z.; Zhang, M.; Shang, E.; Zhang, S.; Yang, J. Spatial and Temporal Changes and Assessment of Multi-Species Habitat in Hainan Jianfengling Protected Area. Remote Sens. 2025, 17, 46. [Google Scholar] [CrossRef]
  15. Zalasiewicz, J.; Thomas, J.A.; Waters, C.N.; Turner, S.; Head, M.J. The Meaning of the Anthropocene: Why It Matters Even without a Formal Geological Definition. Nature 2024, 632, 980–984. [Google Scholar] [CrossRef]
  16. Li, C.; Hao, Y.; Zheng, Z.; Xu, X.; Cui, L.; Guan, M.; Guan, S.; Li, T.; Pang, Z.; Li, L.; et al. Determination of Priority Conservation Areas in the Yellow River Source Region Based on Ecosystem Integrity Evaluations. Glob. Ecol. Conserv. 2024, 49, e02800. [Google Scholar] [CrossRef]
  17. Parrish, J.D.; Braun, D.P.; Unnasch, R.S. Are We Conserving What We Say We Are? Measuring Ecological Integrity within Protected Areas. Bioscience 2003, 53, 851–860. [Google Scholar] [CrossRef]
  18. Nicholson, E.; Watermeyer, K.E.; Rowland, J.A.; Sato, C.F.; Stevenson, S.L.; Andrade, A.; Brooks, T.M.; Burgess, N.D.; Cheng, S.-T.; Grantham, H.S.; et al. Scientific Foundations for an Ecosystem Goal, Milestones and Indicators for the Post-2020 Global Biodiversity Framework. Nat. Ecol. Evol. 2021, 5, 1338–1349. [Google Scholar] [CrossRef] [PubMed]
  19. Rapport, D.J.; Costanza, R.; McMichael, A.J. Assessing Ecosystem Health. Trends Ecol. Evol. 1998, 13, 397–402. [Google Scholar] [CrossRef]
  20. Zhou, J.; Wang, X.; Wang, X.; Yao, W.; Tu, Y.; Sun, Z.; Feng, X. Evaluation of Ecosystem Quality and Stability Based on Key Indicators and Ideal Reference Frame: A Case Study of the Qinghai-Tibet Plateau. J. Environ. Manag. 2024, 370, 122460. [Google Scholar] [CrossRef]
  21. Zhang, S.; Zhang, Z.; Zhang, T.; Ju, H. Assessment of Coastal Zone Ecosystem Health in the Context of Tourism Development: A Case Study of Jiaozhou Bay. Ecol. Indic. 2024, 169, 112874. [Google Scholar] [CrossRef]
  22. Wang, T.; Zhang, J.; Qian, Z.; Dong, Y.; Ma, X. Conservation Effectiveness and Heterogeneity of the National Park in Promoting Ecosystem Health: Causal Evidence from Huangshan, China. Land 2025, 14, 1948. [Google Scholar] [CrossRef]
  23. Yadav, A.; Kansal, M.L.; Singh, A. Ecosystem Health Assessment Based on the V-O-R-S Framework for the Upper Ganga Riverine Wetland in India. Environ. Sustain. Indic. 2025, 25, 100580. [Google Scholar] [CrossRef]
  24. Li, W.; Chen, X.; Zheng, J.; Zhang, F.; Yan, Y.; Hai, W.; Yang, J.; Han, C.; Liu, L. Effectiveness and Driving Mechanisms of Ecological Conservation and Restoration in Sichuan Province, China. Ecol. Indic. 2025, 172, 113238. [Google Scholar] [CrossRef]
  25. Du, L.; Liu, H.; Liu, H.; Liu, W.; Li, Z.; Liu, X.; Li, J.; Quan, Z.; Zhang, Y. Temporal-Spatial Variation of Regulating Ecosystem Services and Analysis of Driving Factors in Hainan Tropical Rainforest National Park, China. Front. Earth Sci. 2025, 13, 1449816. [Google Scholar] [CrossRef]
  26. Tang, Z.; Fang, J.; Sun, J.; Gaston, K.J. Effectiveness of Protected Areas in Maintaining Plant Production. PLoS ONE 2011, 6, e19116. [Google Scholar] [CrossRef] [PubMed]
  27. Cao, Z.; Shao, M.; Lu, Z.; Dong, X.; Ma, C.; Yao, P. Quantitative Assessment of Ecological Conservation Effectiveness and Spillover Effects of China’s First Group of National Parks. Biol. Conserv. 2025, 308, 111242. [Google Scholar] [CrossRef]
  28. Oldekop, J.A.; Holmes, G.; Harris, W.E.; Evans, K.L. A Global Assessment of the Social and Conservation Outcomes of Protected Areas. Conserv. Biol. 2016, 30, 133–141. [Google Scholar] [CrossRef] [PubMed]
  29. Xu, Y.; Li, C. A Novel Multidimensional Framework for Bridging Conservation Gaps and Optimizing the System of Natural Reserves for Biodiversity: A Case Study of Dongying, China. Ecol. Indic. 2025, 170, 113088. [Google Scholar] [CrossRef]
  30. Yu, C.; Zhang, Z.; Jeppesen, E.; Gao, Y.; Liu, Y.; Liu, Y.; Lu, Q.; Wang, C.; Sun, X. Assessment of the Effectiveness of China’s Protected Areas in Enhancing Ecosystem Services. Ecosyst. Serv. 2024, 65, 101588. [Google Scholar] [CrossRef]
  31. Rosenbaum, P.R.; Rubin, D.B. The Central Role of the Propensity Score in Observational Studies for Causal Effects. Biometrika 1983, 70, 41–55. [Google Scholar] [CrossRef]
  32. Ferraro, P.J. Counterfactual Thinking and Impact Evaluation in Environmental Policy. New Dir. Eval. 2009, 2009, 75–84. [Google Scholar] [CrossRef]
  33. Andam, K.S.; Ferraro, P.J.; Pfaff, A.; Sanchez-Azofeifa, G.A.; Robalino, J.A. Measuring the Effectiveness of Protected Area Networks in Reducing Deforestation. Proc. Natl. Acad. Sci. USA 2008, 105, 16089–16094. [Google Scholar] [CrossRef]
  34. Baylis, K.; Honey-Rosés, J.; Börner, J.; Corbera, E.; Ezzine-de-Blas, D.; Ferraro, P.J.; Lapeyre, R.; Persson, U.M.; Pfaff, A.; Wunder, S. Mainstreaming Impact Evaluation in Nature Conservation. Conserv. Lett. 2016, 9, 58–64. [Google Scholar] [CrossRef]
  35. Zhang, X.; Zhong, L.; Zhang, H. Using Propensity Score Matching Models to Assess the Protection Effectiveness in Pudacuo National Park, China. Ecol. Indic. 2023, 150, 110222. [Google Scholar] [CrossRef]
  36. Li, S.; Yu, H.; Wei, R.; Yang, J. Can Wuyishan National Park Establishment Increase Regional Carbon Storage? An Empirical Study Based on the DID Model. Ecol. Indic. 2026, 182, 114500. [Google Scholar] [CrossRef]
  37. Zhang, X.; Liao, L.; Huang, Y.; Fang, Q.; Lan, S.; Chi, M. Conservation Outcome Assessment of Wuyishan Protected Areas Based on InVEST and Propensity Score Matching. Glob. Ecol. Conserv. 2023, 45, e02516. [Google Scholar] [CrossRef]
  38. Ma, B.; Zhang, Y.; Huang, Y.; Wen, Y. Socioeconomic and ecological direct and spillover effects of China’s giant panda nature reserves. For. Policy Econ. 2020, 121, 102313. [Google Scholar] [CrossRef]
  39. Wu, R.; Liu, Q.; Wang, H. Spatial Spillover Effects of Ecosystem Service Values in Northeast China Tiger and Leopard National Park Based on Spatial Durbin Model. Ecol. Indic. 2024, 166, 112509. [Google Scholar] [CrossRef]
  40. Ostrom, E. A General Framework for Analyzing Sustainability of Social-Ecological Systems. Science 2009, 325, 419–422. [Google Scholar] [CrossRef]
  41. Yu, D.; Zhou, Z.; Chen, M.; Liu, J.E.; Wang, N.; Zhu, B.; Cao, Y. Identifying the Driving Mechanisms of Ecosystem Health in a Typical Ecologically Fragile Region: A Study Based on the XGBoost–SHAP Model. Ecol. Indic. 2025, 181, 114472. [Google Scholar] [CrossRef]
  42. Ministry of Environmental Protection. China National Biodiversity Conservation Strategy and Action Plan; Ministry of Environmental Protection: Beijing, China, 2011. [Google Scholar]
  43. Bruelheide, H.; Böhnke, M.; Both, S.; Fang, T.; Assmann, T.; Baruffol, M.; Bauhus, J.; Buscot, F.; Chen, X.-Y.; Ding, B.-Y.; et al. Community Assembly during Secondary Forest Succession in a Chinese Subtropical Forest. Ecol. Monogr. 2011, 81, 25–41. [Google Scholar] [CrossRef]
  44. Zhao, X.; He, Y.; Yu, C.; Xu, D.; Zou, W. Assessment of Ecosystem Services Value in a National Park Pilot. Sustainability 2019, 11, 6609. [Google Scholar] [CrossRef]
  45. Costanza, R.; Mageau, M. What Is a Healthy Ecosystem? Aquat. Ecol. 1999, 33, 105–115. [Google Scholar] [CrossRef]
  46. Zhu, H.; Xu, Q.; Zheng, Y.; Cui, J.; Meng, Q. Ecological health assessment of the middle route of the South-to-North Water Diversion Project using an enhanced VORS model. Ecol. Indic. 2025, 172, 113281. [Google Scholar] [CrossRef]
  47. Li, M.; Li, X.; Liu, S.; Lyu, X.; Dang, D.; Dou, H.; Wang, K.; Cao, W.; Zhang, C. Analysis of the Spatiotemporal Dynamics and Driving Factors of Ecosystem Quality in Inner Mongolia from 2005 to 2020. Environ. Technol. Innov. 2024, 35, 103686. [Google Scholar] [CrossRef]
  48. Yu, Z.; Bu, C.; Li, Y. Machine Learning for Ecological Analysis. Chem. Eng. J. 2025, 507, 160780. [Google Scholar] [CrossRef]
  49. Cai, Y.; Zhu, P.; Li, X.; Liu, X.; Chen, Y.; Shen, Q.; Xu, X.; Zhang, H.; Nie, S.; Wang, C.; et al. Dynamics of China’s Forest Carbon Storage: The First 30 m Annual Aboveground Biomass Mapping from 1985 to 2023. Earth Syst. Sci. Data 2025, 17, 6993–7018. [Google Scholar] [CrossRef]
  50. Wei, J.; Ji, Y.; Yang, D.; Liang, F. A Multi-Model Coupling Approach to Biodiversity Conservation Strategies for Nationally Important Agricultural Heritage Systems in the Beijing–Tianjin–Hebei Region. Sustainability 2025, 17, 7959. [Google Scholar] [CrossRef]
  51. Wang, X.; Liu, G.; Jia, H. InVEST Model and Landscape Indices Reveal Habitat Quality Degradation From Land Use Changes in South China. Ecol. Evol. 2025, 15, e72719. [Google Scholar] [CrossRef]
  52. Ran, X.; Li, Y.; Guo, Y.; Wei, H. Ecosystem Services Assessment and Trade-off Synergy Relationships in Qianjiangyuan National Park based on In VEST model. Resour. Environ. Yangtze Basin 2023, 32, 1932–1948. (In Chinese) [Google Scholar]
  53. Jian, K.; Wu, X.; Zhang, Q.; Wang, S. Identification of Key Driving Factors for Ecological Environmental Quality in Hainan Tropical Rainforest National Park Using Causal Inference with Double Machine Learning. Environ. Sustain. Indic. 2025, 28, 101029. [Google Scholar] [CrossRef]
  54. Bai, Y.; Li, A.; Lei, G.; Bian, J.; Zhang, Z.; Nan, X.; Chen, L.; Lin, X.; Deng, Y.; Shao, H. Identifying Spatiotemporal Dynamics and Driving Factors of Landscape Pattern in Giant Panda National Park (GPNP) Using an Interpretable Machine Learning Technology. Ecol. Indic. 2025, 178, 114121. [Google Scholar] [CrossRef]
  55. Zhao, Z.; Lei, Y.; Jiang, X.; Leng, Y.; Li, W.; Zhao, W. Explainable Machine Learning Reveals Drivers of Ecological Quality in a Karst Watershed: Implications for Sustainable Management of the Mawei River Basin, China. Int. J. Digit. Earth 2026, 19, 2600876. [Google Scholar] [CrossRef]
  56. Lundberg, S.M.; Erion, G.; Chen, H.; DeGrave, A.; Prutkin, J.M.; Nair, B.; Katz, R.; Himmelfarb, J.; Bansal, N.; Lee, S.-I. From Local Explanations to Global Understanding with Explainable AI for Trees. Nat. Mach. Intell. 2020, 2, 56–67. [Google Scholar] [CrossRef] [PubMed]
  57. Meng, B.; Wang, M.; Zhang, Z.; Pan, H. Evaluation and Driving Force Analysis of Relative Effectiveness in the Giant Panda National Park in Sichuan, China. Front. Environ. Sci. 2025, 12, 1512640. [Google Scholar] [CrossRef]
  58. Zeng, Y.; Wang, L.; Zhong, L. Tourism Has Intensified Habitat Loss within and around Protected Areas: Evidence from China. J. Appl. Ecol. 2026, 63, e70246. [Google Scholar] [CrossRef]
Land 15 00863 g001
Figure 1. Research scope definition and QJYNP location conditions.
Figure 1. Research scope definition and QJYNP location conditions.
Land 15 00863 g001
Land 15 00863 g002
Figure 2. Research framework for assessing conservation effectiveness and spatial drivers of QJYNP.
Figure 2. Research framework for assessing conservation effectiveness and spatial drivers of QJYNP.
Land 15 00863 g002
Land 15 00863 g003
Figure 3. Spatial patterns and change characteristics of ecological protection effectiveness indicators in the study area for 2015, 2020, and 2024. (a) Spatial distribution of EO, EV, ES, and EEI; (b) Spatial distribution of change types for each indicator and statistical analysis of grid proportions from 2015 to 2024.
Figure 3. Spatial patterns and change characteristics of ecological protection effectiveness indicators in the study area for 2015, 2020, and 2024. (a) Spatial distribution of EO, EV, ES, and EEI; (b) Spatial distribution of change types for each indicator and statistical analysis of grid proportions from 2015 to 2024.
Land 15 00863 g003
Land 15 00863 g004
Figure 4. Distribution characteristics and temporal changes in ecological conservation effectiveness indicators for NP, Buffer, and Out. (a) Distribution characteristics of EO, EV, ES, and EEI across the three zones in 2015, 2020, and 2024; (b) Temporal trends in the mean EEI values across the three zones from 2015 to 2024.
Figure 4. Distribution characteristics and temporal changes in ecological conservation effectiveness indicators for NP, Buffer, and Out. (a) Distribution characteristics of EO, EV, ES, and EEI across the three zones in 2015, 2020, and 2024; (b) Temporal trends in the mean EEI values across the three zones from 2015 to 2024.
Land 15 00863 g004
Land 15 00863 g005
Figure 5. Results of the EEI driver analysis using the RF-SHAP model: (a) Comparison of the importance of drivers in 2015, 2020, and 2024; (b) SHAP swarm plot for 2024; (c) SHAP dependency plot of the main drivers for 2024.
Figure 5. Results of the EEI driver analysis using the RF-SHAP model: (a) Comparison of the importance of drivers in 2015, 2020, and 2024; (b) SHAP swarm plot for 2024; (c) SHAP dependency plot of the main drivers for 2024.
Land 15 00863 g005
Land 15 00863 g006
Figure 6. PSM-DID estimates of the direct effects across indicators and time periods (NP vs. Out).
Figure 6. PSM-DID estimates of the direct effects across indicators and time periods (NP vs. Out).
Land 15 00863 g006
Land 15 00863 g007
Figure 7. PSM-DID estimates of the spillover effects across indicators and time periods (Buffer vs. Out).
Figure 7. PSM-DID estimates of the spillover effects across indicators and time periods (Buffer vs. Out).
Land 15 00863 g007
Table 1. Summary of research data sources and descriptions.
Table 1. Summary of research data sources and descriptions.
Data DescriptionData SourcesResolutionPeriod
Land-useScience and Data Center of the Chinese Academy of Sciences (https://www.resdc.cn/ (accessed on 20 November 2025) 30 m 2015, 2020, 2024
China Annual Nighttime Light Dataset500 m
Maximum Normalized Difference Vegetation Index (NDVI_MAX)30 m
Net Primary Productivity (NPP)30 m
Elevation and Slope30 m
Forest Aboveground Biomass (AGB)Zenodo [49]
https://doi.org/10.5281/zenodo.12747329 (accessed on 20 November 2025)		30 m2015, 2020, 2023
PrecipitationNational Tibetan Plateau Data Center https://data.tpdc.ac.cn/home (accessed on 20 November 2025) 1000 m2014–2025
Temperature 1000 m2014–2025
Potential evapotranspiration1000 m2015, 2020, 2024
Soil dataHarmonized World Soil Database 2.0 https://www.fao.org/soils-portal/data-hub/soil-maps-and-databases/harmonized-world-soil-database-v20/en/(accessed on 20 November 2025) 100 m
PopulationOak Ridge National Laboratory https://landscan.ornl.gov/ (accessed on 25 November 2025) 1000 m2015, 2020, 2024
Road and water networksOpenStreetMap https://www.openstreetmap.org/ (accessed on 25 November 2025)vector data2015, 2020, 2024
Table 2. PSM-DID estimates of the direct effects of national park establishment.
Table 2. PSM-DID estimates of the direct effects of national park establishment.
IndicatorPeriodATT (β)95% CIRange
EEI2015–20200.013 ***[0.010, 0.016]+1.03%
2015–2024−0.016 ***[−0.022, −0.009]−1.23%
2020–2024−0.029 ***[−0.033, −0.024]−2.26%
EO2015–2020−0.048 ***[−0.056, −0.040]−3.85%
2015–2024−0.056 ***[−0.067, −0.046]−4.50%
2020–2024−0.008 ***[−0.012, −0.004]−0.65%
EV2015–20200.031 ***[0.029, 0.033]+5.25%
2015–20240.009 ***[0.004, 0.014]+1.52%
2020–2024−0.022 ***[−0.027, −0.018]−3.73%
ES2015–20200.004 **[0.001, 0.007]+0.35%
2015–2024−0.020 ***[−0.025, −0.014]−1.60%
2020–2024−0.024 ***[−0.028, −0.020]−1.95%
ATT denotes the average treatment effect; the values in square brackets represent the 95% confidence interval (β± 1.96 × SE). *** and ** indicate significance at the 0.1% and 1% levels respectively.
Table 3. PSM-DID estimates of the spillover effects of national park establishment.
Table 3. PSM-DID estimates of the spillover effects of national park establishment.
IndicatorPeriodATT (β)95% CIRange
EEI2015–20200.0007[−0.0006, 0.0020]+0.05%
2015–2024−0.0124 ***[−0.0150, −0.0098]−1.00%
2020–2024−0.0130 ***[−0.0149, −0.0112]−1.05%
EO2015–2020−0.0189 ***[−0.0221, −0.0156]−1.52%
2015–2024−0.0234 ***[−0.0276, −0.0193]−1.89%
2020–2024−0.0046 ***[−0.0059, −0.0032]−0.37%
EV2015–20200.0118 ***[0.0110, 0.0126]+2.08%
2015–2024−0.0011[−0.0030, 0.0008]−0.20%
2020–2024−0.0129 ***[−0.0147, −0.0111]−2.28%
ES2015–2020−0.0099 ***[−0.0116, −0.0082]−0.82%
2015–2024−0.0136 ***[−0.0162, −0.0110]−1.13%
2020–2024−0.0037 ***[−0.0054, −0.0020]−0.31%
ATT denotes the average treatment effect; values in square brackets represent the 95% confidence interval ( β ± 1.96 × SE); *** indicates significance at the 0.1%.
Table 4. Comparison of direct and spillover effects rates between the unmatched and PSM-DID methods.
Table 4. Comparison of direct and spillover effects rates between the unmatched and PSM-DID methods.
MethodNp vs. OutBuffer vs. Out
2015–20202015–20242015–20202015–2024
Not match0.00%−4.10%0.00%−0.81%
PSM-DID1.03%−1.23%+0.05%−1.00%
Disclaimer/Publisher’s Note: The statements, opinions and data contained in all publications are solely those of the individual author(s) and contributor(s) and not of MDPI and/or the editor(s). MDPI and/or the editor(s) disclaim responsibility for any injury to people or property resulting from any ideas, methods, instructions or products referred to in the content.

Share and Cite

MDPI and ACS Style

Wang, C.; Wang, Y.; Lu, J.; Li, L. Conservation Effectiveness and Spatial Drivers of Qianjiangyuan National Park: Causal Evidence from a Quasi-Experimental Framework. Land 2026, 15, 863. https://doi.org/10.3390/land15050863

AMA Style

Wang C, Wang Y, Lu J, Li L. Conservation Effectiveness and Spatial Drivers of Qianjiangyuan National Park: Causal Evidence from a Quasi-Experimental Framework. Land. 2026; 15(5):863. https://doi.org/10.3390/land15050863

Chicago/Turabian Style

Wang, Chuqi, Yinglin Wang, Jiwen Lu, and Liang Li. 2026. "Conservation Effectiveness and Spatial Drivers of Qianjiangyuan National Park: Causal Evidence from a Quasi-Experimental Framework" Land 15, no. 5: 863. https://doi.org/10.3390/land15050863

APA Style

Wang, C., Wang, Y., Lu, J., & Li, L. (2026). Conservation Effectiveness and Spatial Drivers of Qianjiangyuan National Park: Causal Evidence from a Quasi-Experimental Framework. Land, 15(5), 863. https://doi.org/10.3390/land15050863

Note that from the first issue of 2016, this journal uses article numbers instead of page numbers. See further details here.

Article Metrics

Back to TopTop