Vegetation Water Status and Topography Shape Drought-Induced Mortality of Two Pine Species in Southwest China

Increasingly frequency of extreme droughts has triggered widespread tree mortality across biomes, yet, the species-specific drivers and spatial patterns of drought-induced tree mortality at regional scales remain poorly understood, hindering effective forest management. In 2023, an extensive die-off of Pinus yunnanensis and Pinus kesiya var. langbianensis occurred in central Yunnan Province, Southwest China. This study integrated unmanned aerial vehicle (UAV) plot surveys, remote sensing, and environmental data to assess the extent and drivers of this Pinus mortality. Field observed mortality rates rang from 1.05% to 6.47% for P. yunnanensis and 0.53% to 11.16% for P. kesiya var. langbianensis, with most dead trees occurring on south-facing slopes. Random forest models revealed that mortality in P. yunnanensis was jointly influenced by soil potassium, stand age, and canopy height, whereas in P. kesiya it was primarily associated with habitat slope. For both species, dry season normalized difference water index (NDWI) was the most consistent top predictor, underscoring the pivotal role of vegetation water status. Predicted mortality hotspots were concentrated at the edges of species’ ranges, coinciding with regions that experienced recurrent extreme droughts over the past decades. This study provides the first spatially explicit and species-specific assessment of drought-induced mortality for these pines, offering a framework that integrates UAV-derived metrics with multiple source data to predict forest die-off under climate extremes.

Increasingly frequency of extreme droughts has triggered widespread tree mortality across biomes, yet, the species-specific drivers and spatial patterns of drought-induced tree mortality at regional scales remain poorly understood, hindering effective forest management. In 2023, an extensive die-off of Pinus yunnanensis and Pinus kesiya var. langbianensis occurred in central Yunnan Province, Southwest China. This study integrated unmanned aerial vehicle (UAV) plot surveys, remote sensing, and environmental data to assess the extent and drivers of this Pinus mortality. Field observed mortality rates rang from 1.05% to 6.47% for P. yunnanensis and 0.53% to 11.16% for P. kesiya var. langbianensis, with most dead trees occurring on south-facing slopes. Random forest models revealed that mortality in P. yunnanensis was jointly influenced by soil potassium, stand age, and canopy height, whereas in P. kesiya it was primarily associated with habitat slope. For both species, dry season normalized difference water index (NDWI) was the most consistent top predictor, underscoring the pivotal role of vegetation water status. Predicted mortality hotspots were concentrated at the edges of species’ ranges, coinciding with regions that experienced recurrent extreme droughts over the past decades. This study provides the first spatially explicit and species-specific assessment of drought-induced mortality for these pines, offering a framework that integrates UAV-derived metrics with multiple source data to predict forest die-off under climate extremes.

Extreme drought; Remote sensing; Tree mortality; Unmanned aerial vehicle (UAV); Pinus yunnanensis ; Pinus kesiya var. langbianensis 1 Introduction Drought events are increasing in frequency and severity under ongoing climate change, and extreme droughts are now recognized as a primary driver of widespread tree mortality (Allen et al., 2010; Choat et al., 2018). Large-scale die-offs have profound ecological and socio-economic consequences, including habitat fragmentation, reduced carbon sequestration, and disruption of forestry-dependent livelihoods (Meyer et al., 2005; Pfeifer et al., 2011; Jones et al., 2018). Over the past two decades, drought-induced die-offs have escalated globally across tropical, temperate, and boreal biomes (Anderegg et al., 2013; Hammond et al., 2022). For example, notable cases include extensive die-off of Pinus halepensis in the Mediterranean basin (Camarero et al., 2015), widespread Eucalyptus mortality in Australia (Matusick et al., 2013), oak and pine dieback in Spain (González De Andrés et al., 2022), and severe beech and spruce mortality in Europe (Obladen et al., 2021; Martinez Del Castillo et al., 2022). Climate models project further increases in drought frequency, intensity, and duration under continued warming (IPCC, 2023), highlighting the urgent need to identify the drivers of mortality and map vulnerable areas (Sturm et al., 2022). Despite the growing global evidence, regional assessments remain limited, particularly in tropical and subtropical forests subject to pronounced seasonal drought, such as the Asian monsoon region, where forests support critical ecological and cultural functions (Corlett, 2016) Yet, even within the same drought events, mortality is highly heterogeneous, reflecting modulation by local biophysical factors such as topography, soil properties, and stand structure (Van Gunst et al., 2016; Varani et al., 2024; Chen et al., 2024; Jääskeläinen et al., 2025). For example, mortality tends to be higher on lower elevations, steeper, and south-facing slopes with coarse-textured, shallow, and nutrient-poor soils, where water limitation exacerbates hydraulic failure (Guarín and Taylor, 2005; Sardans and Peñuelas, 2007; Rozas and Sampedro, 2013; Crouchet et al., 2019; Varani et al., 2024). Similarly, denser stands with taller or older trees experience elevated mortality due to intensified competition and reduced physiological resilience (Van Gunst et al., 2016; Stovall et al., 2019; Johnston et al., 2025). Consequently, climate variables alone are insufficient to explain or forecast landscape-scale tree mortality. Integrating site-specific environmental and structural drivers is therefore necessary to capture both the mechanisms and spatial patterns of forest die-off. To complement these local predictors, scalable monitoring approaches such as unmanned aerial vehicle (UAV) and satellite-derived vegetation indices provide practical tools for assessing forest vulnerability across landscapes. Indices including the Normalized Difference Vegetation Index (NDVI) and the Normalized Difference Red-Edge Index (NDRE) are widely used proxies for forest health, effective for tracking canopy growth and dieback dynamics (Breshears et al., 2005; Swain et al., 2011; Buras et al., 2018; Moreno-Fernández et al., 2022; Italiano et al., 2023). Multiple studies have demonstrated that these indices can effectively capture both physiological and morphological changes during the transition from tree health to mortality, making them highly suitable for monitoring forest condition in drought-prone ecosystems (Klouček et al., 2024; Trubin et al., 2024; Yan et al., 2025). However, NDVI can saturate in dense canopies, while NDRE partially alleviates saturation but may lose sensitivity at high leaf area. By contrast, the Normalized Difference Water Index (NDWI; Gao, 1996) is sensitive to canopy water content, making it particularly suited for detecting water stress and delineating high-risk areas during drought. Together, these spectral indices complement local environmental and structural predictors in forecasting drought-induced mortality. During 2022-2023, Yunnan Province experienced a unprecedented eight-month extreme drought, driven by concurrent precipitation deficits and amplified evapotranspiration (Ma et al., 2024; Wang and Ma, 2024), resulting in indirect economic losses exceeding 5 billion yuan (Yunnan Provincial Government Memorial Hall, 2023). This event triggered extremely extensive mortality of two dominant conifer species, Pinus yunnanensis and Pinus kesiya var. langbianensis, in Yunnan forests. This study developed a species-specific modeling framework integrating UAV imagery (DJI Mavic 3M), Sentinel-2 data, and fine-scale environmental and forest structural predictors, including topographic variables (e.g., elevation, slope), soil nutrients (e.g., total nitrogen, total potassium), stand attributes (tree height, tree density), and vegetation indices (NDVI, NDWI, NDRE). This framework enables identification of key predictors of elevated mortality and quantification of their relative contributions to forest vulnerability. We specifically address three questions: (1) What was the severity and spatial distribution of tree mortality during the extreme drought, and how did patterns differ between the two Pinus species? (2) Which type predictors most strongly influence mortality rates, and do their relative contributions differ between species? (3) Can these predictors be integrated into species-specific models to generate spatially explicit mortality risk maps? The results can provide critical insights for sustainable forest management and climate adaptation strategies in drought-prone regions. 2. Materials and methods 2.1 Study area and species The study area (99.65°-103.94°E, 23.40°-26.33°N) encompasses elevations of 987-1895 m and features a pronounced monsoon climate with a dry season from November to April and a wet season from May to October (Fig. 1; Table 1) (Wu and Zhu, 1987). Over the 1950-2023 period, the study area’s mean annual precipitation and temperature were 1086.06 mm and 17.27 °C, respectively (Table S1). The driest site locates in central-northern Yunnan, received an average annual precipitation of 909.70 mm and a mean negative climatic water deficit (CWD) of -142.37mm, whereas the wettest site locates in southwestern Yunnan, recorded an average annual precipitation of 1407.80 mm and a mean positive CWD of 367.87mm (Table S1). In recent years, Yunnan has experienced multiple extreme droughts, with the 2023 dry season being particularly severe (Ma et al., 2024). Compared to the 2013-2022 baseline, the 2023 dry season showed unprecedented temperature increases and precipitation deficits. Specifically, northwestern Yunnan recorded temperature rises exceeding 3°C, while precipitation in southern Yunnan reduced over 15 mm (Fig. S2(a-b)). Potential evapotranspiration (PET) in certain subregions even surpassed 150 mm (Fig. S2c). Notably, the study area showed high spatial overlap with regions characterized by frequent droughts during 2013 - 2022 (Fig. S2(d)). P. yunnanensis and P. kesiya var. langbianensis, are the dominant afforestation species in southern and central Yunnan, respectively (Wu and Zhu, 1987; Liu et al., 2010). P. kesiya var. langbianensis occurs at 700-1800 m on red soil derived from sandstone, slate, or localized limestone, favoring warmer, more humid southern regions. It grows rapidly and regenerates naturally but is relatively sensitive to drought and poor soil. In contrast, P. yunnanensis inhabits colder, drier central and northern areas, thriving on deep, fertile, well-drained acidic sandy soils, and is more tolerant to water stress (Chen et al., 2012). 2. 2 UAV plot survey We conducted aerial surveys in September 2023, January 2024, and March 2025 over forest sites that exhibited signs of tree mortality (Fig. S1). Unmanned aerial vehicle (UAV) equipped with multispectral cameras were flown at a terrain-following altitude of 30 m, with each mission covering areas larger than 4 ha (200 m × 200 m). The UAV imagery was processed using DJI Terra (version3.7.6) to generate complete orthoimage of the mortality-affected areas through reconstruction. To ensure that the observed tree mortality was both drought-induced and accurately dated, we conducted complementary field inspections within all surveyed plots. Increment cores and bark samples were collected from both dead and living trees for anatomical and dendrochronological analyses. These examinations allowed us to exclude alternative mortality agents such as fire, insect outbreaks, or pathogenic infections. Moreover, cross-dating of tree-ring series from live and dead individuals confirmed that nearly all sampled dead trees ceased growth simultaneously in 2023, thereby providing robust evidence that the mortality events occurred in that year and were primarily associated with the extreme drought conditions (data not shown). For the orthoimages, it was further clipped into standardized 200 m × 200 m plots using ArcGIS 10.8 (Fig. 2(a)). Within each plot, dead tree crowns were manually delineated and the plot was subdivided into 100 grid cells (pixels) of 20 m × 20 m (Fig. 2(b)). Tree mortality rate in each cell was calculated as the ratio of the cumulative crown area of dead trees to the cell area (Eq. 1). \begin{equation} \begin{matrix}M\text{ortality}=\frac{c\text{rowm\ area\ of\ dead\ trees\ within\ the\ pixel}}{\text{cell}\text{\ area}}\ \#\left(1\right)\\ \end{matrix}\nonumber \\ \end{equation} To minimize potential biases in subsequent modeling, grid cells adjacent to roads or dominated by bare ground were excluded. In total, 9.5% of the cells were removed from the analysis, with 1.33% excluded due to road adjacency and 8.17% due to insufficient tree cover. For the remaining cells, all other predictor variables, such as elevation (Fig. 2(c)), vegetation indices (Fig. 2(d)), were extracted to align with the cell-level mortality data. 2.3 Topographical and soil nutrient data To generate topographic variables at the same spatial resolution as the tree mortality dataset (20 m), elevation data were derived directly from UAV-based digital elevation models (DEMs) rather than from coarse-resolution existing DEM products (Table 1). The UAV-derived DEMs, with an initial resolution of 2 cm, were aggregated by computing the mean elevation within each 20 m cell. From these aggregated elevation layers, slope and aspect were calculated using standard GIS terrain algorithms (Table S2). In addition, the topographic wetness index (TWI; Eq. 2) was computed for each cell based on the mean elevation, slope, and upslope contributing area. \begin{equation} \begin{matrix}{TWI=ln}\frac{a}{\tan\beta}\ \#\left(2\right)\\ \end{matrix}\nonumber \\ \end{equation} where α represents the upslope contributing area per unit contour length, and β is the local slope angle in radians. These variables (elevation, slope, aspect, and TWI) were subsequently incorporated as topography predictors in the mortality modeling framework. Soil property variables (Table S2), including soil thickness, soil organic carbon density (SOC, 0 - 5 cm), total nitrogen (TN, 0 - 5 cm), total potassium (TK, 0 - 5 cm), and total phosphorus (TP, 0 - 5 cm), were acquired from the National Earth System Science Data Center (http://soil.geodata.cn). These datasets originally had a spatial resolution of 90 m. To align with the 20 m resolution of the tree mortality dataset and enhance spatial consistency across variables, all soil property layers were resampled to 20 m resolution using bilinear interpolation implemented in the terra package in R (R Core Team, 2024; Hijmans, 2025). This method computes each new pixel value as a weighted average of its four nearest neighboring pixels, thereby ensuring smoother transitions between adjacent pixels and minimizing spatial mismatches that might arise from differences in original spatial scales. 2.4 Forest structure and vegetation index This study employed several forest structural variables, including stand canopy height (sourced from : https://www.3decology.org), stand age (sourced from: https://essd.copernicus.org), and stand tree density (sourced from: https://data.tpdc.ac.cn) (Table S2). Both variables have been widely recognized as critical factors influencing drought vulnerability in forest ecosystems (Johnston et al., 2025). Since the original spatial resolution of these datasets was 30 m, all variables were resampled to 20 m resolution using bilinear interpolation to ensure spatial consistency with the mortality dataset (Hijmans, 2025). Vegetation indices, including NDVI, NDRE and NDWI are applied into the analysis (Table S2). All indices were derived from Sentinel-2 satellite imagery, which features a temporal resolution of 5 days and a spatial resolution of 20 m (https://earthengine.google.com). Notably, we downloaded vegetation index data spanning from June 2015 to January 2025, however, due to extensive missing values prior to 2017, those earlier records were excluded from analysis. To capture drought-induced vegetation stress, we calculated the seasonal anomaly (VI_re) for each index, defined as the difference between the mean index value during the 2023 dry season and the corresponding mean during the 2022 dry season. The dry season window was specifically chosen, as vegetation indices during this period are most sensitive to plant water stress (Zhang et al., 2024). Ultimately, all VI_re metrics were incorporated into the mortality prediction model as key predictors, enabling an assessment of the association between remotely sensed vegetation dynamics and observed mortality patterns. 2.5 Statistics We employed random forest (RF) regression to investigate the relationships between tree mortality and the full suite of environmental and vegetation predictors. RF is a machine learning algorithm that uses an ensemble of decision trees to produce predictions, where the final predicted value in regression tasks is obtained by averaging the outputs of all terminal nodes across the trees (Breiman, 2001a). RF was chosen because it can handle large datasets with numerous predictor variables, capture complex nonlinear interactions among predictors, and maintain robustness to multicollinearity, while also accommodating skewed or zero-inflated response distributions (Breiman, 2001b). These attributes are particularly critical in the present study, as the mortality dataset includes a substantial proportion of zero values, rendering it unsuitable for conventional linear or nonlinear regression approaches (Fig. S3). Consistent with these advantages, RF has been extensively applied in studies of drought-induced tree mortality (Paz‐Kagan et al., 2017; Howe et al., 2022; Queally et al., 2025). After assembling the full set of predictor variables, we assessed pairwise correlations among predictors to reduce redundancy (Fig. S4). Although RF is generally robust to multicollinearity, excessive correlation can still bias variable importance measures and reduce interpretability (Molnar, 2020). A correlation threshold of |r| = 0.7 was therefore adopted (Queally et al., 2025). For each correlated pair exceeding this threshold, a two-step variable selection process was applied. In the first step, an RF “full model” including all predictors was fitted, and variables were ranked by their importance, measured as the percentage increase in mean squared error (%IncMSE) (Fig. S5). For each correlated pair, the variable with lower importance was excluded. If the two variables exhibited comparable importance (i.e., differences were negligible relative to permutation variability), the pair was passed to Step 2 instead of discarding either variable. In the second step, we examined each remaining correlated pair with respect to their correlations with all other predictors. In cases where one variable exhibited consistently higher correlations across the predictor set, the variable with lower overall correlation was retained. The final set of predictors was then used to fit species-specific RF models for P. yunnanensis and P. kesiya var. langbianensis . Model hyperparameters were optimized via a combination of 10-fold cross-validation and grid search, implemented using the caret package in R (Kuhn, 2008). Model performance was evaluated using the coefficient of determination (R²) and root mean squared error (RMSE). To interpret model behavior and quantify the contribution of each predictor, we computed Shapley additive explanations (SHAP) values for all variables. SHAP is a post-hoc model interpretation method grounded in cooperative game theory, which decomposes a model prediction into additive contributions from each predictor (Shapley, 1953). This approach allows for a more nuanced understanding of how individual variables influence the predicted mortality rates, both in magnitude and direction, across the full range of observations (Lundberg and Lee, 2017). Finally, to further explore nonlinear and threshold effects, we generated partial dependence plots (PDPs) using the pdp package in R (Greenwell, 2017). PDPs depict the marginal effect of a single predictor on mortality while averaging over the distribution of all other predictors, thereby providing complementary insights into the functional relationships between variables and observed mortality (Friedman, 2001). 3 Results 3.1 Tree mortality rates across sites and species Across all surveyed plots, the average mortality of P. yunnanensis ranged from 1.05% to 6.47%, with the three sites with the highest mortality rate concentrated in central-northern Yunnan (mean = 5.87%; Fig. 3(a); Table 1). In contrast, P. kesiya var. langbianensis exhibited a broader range of mortality, from 0.53% to 11.17% (mean =8.59%), with the highest mortality rate occurring in southwestern Yunnan (Fig. 3(b); Table 1). Despite these distinct spatial patterns and ranges, the Mann-Whitney U test indicated no statistically significant difference in mortality rates between the two species across plots (p = 0.38; Fig. 3(c)). 3.2 Impact factors on tree mortality After eliminating redundant predictors and optimizing hyperparameters, we developed optimal random forest models for predicting the mortality rates of the two tree species (Fig. 4). The model for P. kesiya var. langbianensis achieved higher explanatory power (R² = 0.506) compared with that for P. yunnanensis (R² = 0.213). However, the P. yunnanensis model exhibited lower prediction error (RMSE = 4.681%) relative to P. kesiya var. langbianensis (RMSE = 5.869%) (Fig. 4(a-b)). Variable importance based on mean absolute SHAP values was presented in Fig. 5(a, c). For P. yunnanensis, the most influential predictors were total potassium (TK), stand age, NDWI_re, and stand canopy height. Among topographic factors, only the topographic wetness index (TWI) showed a notable contribution (mean SHAP = 0.30). At the categorical level, predictors ranked in importance as: Soil (32.98%) > Forest structure (30.61%) > Topography (24.80%) > Vegetation (11.61%) (Fig. 6). In contrast, mortality in P. kesiya var. langbianensis was most strongly influenced by slope, total nitrogen (TN), NDWI_re, and stand density. Notably, slope emerged as the dominant predictor (max mean SHAP = 1.25), surpassing all others. The categorical ranking was: Topography (39.61%) > Soil (30.46%) > Forest structure (17.78%) > Vegetation (12.15%) (Fig. 6). Importantly, NDWI_re consistently ranked among the top three predictors for both species, underscoring its robust capacity to detect drought-induced mortality. Beeswarm plots of SHAP values (Fig. 5(b, d)) further revealed the directional effects of key predictors. For P. yunnanensis, higher TK content and older tree age were associated with elevated mortality risk. For P. kesiya var. langbianensis, steeper slopes increased mortality, whereas higher TN content and greater stand density tended to reduce it. 3.3 Partial dependence To gain a more intuitive understanding of how predictor variables influence mortality, we generated partial dependence plots (PDPs) of SHAP values for the six most important variables in each mortality model. For P. yunnanensis, higher mortality was observed when TK exceeded 17.62 g·kg -1, stand age was greater than 54.37 years, stand canopy height exceeded 6.12 m, or NDWI_re fell below -0.16 (Fig. 7 (a-d)). Both TWI and TN exhibited two distinct inflection points in their relationships with SHAP values (Fig. 7 (e-f)). Mortality was relatively low when TWI ranged from -1.87 to 0.25 and TN from 1.17 to 1.51 g·kg -1 . For P. kesiya var. langbianensis, mortality increased markedly when slope exceeded 47.9°, NDWI_re was below -0.14, TWI was less than -1.29, or elevation was lower than 1580.35 m (Fig. 7 (g, i, k, l)). Similarly, TN and tree density also showed dual inflection patterns, with lower mortality occurring when TN ranged from 1.23 to 1.37 g·kg -1 and tree density from 145.01 to 463.55 trees·ha -1 (Fig. 7 (h, j)). 3.4 Mortality prediction across Pinus ranges Based on the optimal random forest models, we predicted the drought-induced mortality rates of P. kesiya and P. yunnanensis within their respective distribution ranges in Yunnan Province for 2023 (Fig. 8). For P. yunnanensis, the predicted mortality rate ranged from 2.48% to 22.32%, with a mean of 7.08% and a standard deviation of 1.19%. Approximately 16.7% of the distribution area exhibited mortality rates above 15%. High mortality rates were primarily concentrated in the northeastern and southeastern portions of the range, whereas lower values occurred mainly in the central and northwestern regions. For P. kesiya var. langbianensis, the predicted mortality rate ranged from 2.30% to 17.27%, with a mean of 7.58% and a standard deviation of 1.25%. In contrast to P. yunnanensis, large areas within its distribution exhibited relatively high mortality rates, with 41.3% of the area exceeding 15%. Regions with lower mortality were mostly located in the central and northeastern parts of the range. Taken together, areas with high predicted mortality largely overlapped with the peripheral zones of both species’ distributions in Yunnan Province, which also correspond to regions that have experienced a higher frequency of extreme drought events over the past decade (Fig. S2(d)). 4 Discussion This study provides a spatially explicit assessment of drought-induced mortality for P. yunnanensis and P. kesiya var. langbianensis, in Yunnan, SW China. Mortality was spatially heterogeneous, mainly on south-facing slopes, with rates of 1.05-6.47% and 0.53-11.16% for P. yunnanensis and P. kesiya var. langbianensis, respectively. Species-specific drivers were identified: soil potassium, stand age, and canopy height for P. yunnanensis, and slope for P. kesiya var. langbianensis, while NDWI consistently highlighted vegetation water status for both. Integrating UAV, soil, stand, and spectral data enabled species-specific risk maps, revealing hotspots at range edges and recurrent drought zones. These findings provide key insights for predicting vulnerability under extreme drought. 4.1 Key predictors of tree mortality A common predictor across both species was the remotely sensed NDWI_re, which consistently ranked among the top variables (Fig. 5(a, c)). Specifically, higher NDWI_re values (> -0.14 to - 0.16) were associated with lower mortality, underscoring the utility of canopy water status as a direct indicator of vitality during drought stress (Anderson et al., 2010; Marusig et al., 2020). This convergence suggests that canopy-level water availability is a fundamental constraint on survival, regardless of species-specific traits or habitats. Traditionally, drought-induced tree mortality has been predicted using indirect environmental variables such as climate, soil, and topography, which fail to directly capture forest physiological status (Keshavarz et al., 2014; Jääskeläinen et al., 2025). In contrast, NDWI provides a proximate indicator of canopy water availability, with higher values reflecting active transpiration and lower mortality risk. Accordingly, NDWI has been widely integrated with other vegetation indices and environmental variables to map tree mortality from regional to global scales (Sturm et al., 2022; Zhang et al., 2024; Vulova et al., 2025). In contrast, the dominant mortality drivers differed markedly between the two pines. For P. yunnanensis, mortality was closely associated with soil total potassium (TK), showing a sharp increase when TK exceeded 17.6 g·kg⁻¹ (Fig. 7(a)). Although potassium generally promotes osmotic regulation and drought tolerance, its uptake is restricted under extreme drought, leading to nutrient imbalance and physiological stress (Sardans and Peñuelas, 2007; Mu et al., 2023). Moreover, excessive potassium can stimulate growth that exacerbates water competition under water-limited conditions, thereby intensifying die-off risk (Stoneman et al., 1997; Battie‐Laclau et al., 2014). These mechanisms hav been widely reported in other drought-prone forests (Chambi-Legoas et al., 2020; Schönbeck et al., 2021; Touche et al., 2022). Mortality in P. yunnanensis was also strongly shaped by intrinsic attributes such as tree height (> 6.12 m) and age (> 54.37 years) (Fig. 7(b,d)). Taller and older individuals showed heightened vulnerability, reflecting the dual effects of hydraulic limitation and age-related declines in defense capacity (Bennett et al., 2015; Xu et al., 2018; Camarero et al., 2024; Gao et al., 2025). For example, a global meta-analysis showed that taller trees face roughly twice the drought-induced mortality risk of shorter ones, largely due to carbon starvation and pest susceptibility (West et al., 1999; McDowell et al., 2008; Stovall et al., 2019; Queally et al., 2025). Conversely, other studies suggest that deep rooting may enhance drought survival in taller individuals, due to access deeper soil water and nutrient reserves. (Duursma et al., 2011; Castagneri et al., 2012; Grote et al., 2016; Bose et al., 2024). Tree age exerts a similarly dual influence: young trees may reduce water loss through smaller crowns but suffer from weaker competitive ability, whereas older trees benefit from stronger stomatal regulation yet face age-related declines in physiological defenses, increasing vulnerability to stress and pests (Luo and Chen, 2011; Gazol et al., 2017; Camarero et al., 2024). Overall, these findings suggest that P. yunnanensis mortality is largely governed by the interaction between resource dependence and ontogenetic thresholds. By contrast, P. kesiya mortality was dominated by extrinsic environmental stressors. Slope exerted the strongest influence, with mortality surging above 54.37° (Fig. 7(g)), consistent with the shallow soils, coarse textures, and rapid runoff that constrain water and nutrient retention on steep terrain (Schwinning and Ehleringer, 2001; Dietrich and Perron, 2006). Additionally, increased topographic exposure can also elevate the risk of mechanical damage from strong winds or rockfalls, as well as susceptibility to pest and pathogen attacks, further weakening trees’ resilience to environmental stress. Soil nitrogen and stand density also exhibited bimodal effects in mortality of P. kesiya, with mortality elevated at both low and high values (Fig. 7(b, j)). Nitrogen limitation compromises photosynthesis and defense (Herms and Mattson, 1992; Aber et al., 1998), while excessive nitrogen promotes imbalanced growth and xylem cavitation risk (Herms and Mattson, 1992; Güsewell, 2004). Similarly, low-density stands lack microclimatic buffering (Davis et al., 2019; Rambo and North, 2009), while high-density stands intensify competition for water and nutrients (Xue et al., 2010; Kamara and Kamruzzaman, 2020; Forrester et al., 2025). These results indicate that P. kesiya, often occupying fragmented and topographically habitats, is primarily constrained by physical-habitats rather than intrinsic size-related factors. The drivers of drought-related mortality for the above two pine species indicated that P. yunnanensis mortality is mainly linked to physiological and demographic traits, whereas P. kesiya mortality is predominantly dictated by site conditions and stand structure. This difference underscores the need for species-specific management strategies-with conservation of P. yunnanensis focusing on nutrient dynamics and demographic regulation, while the management of P. kesiya should prioritize avoiding topographically vulnerable sites and stand density optimization. 4.2 Spatial prejection of mortality rates The spatial patterns of predicted mortality for P. kesiya var. langbianensis and P. yunnanensis in 2023 revealed shared distribution and distinct species-specific spatial distribution. Primarily, high mortality rates for both species are predominantly clustered along the periphery of their respective distribution ranges (Fig. 8). This finding corroborate prior research indicating elevated mortality risk in edge populations of tree species, likely attributable to their limited capacity for acclimatory adjustments or tolerance to intensified aridity (Anderegg et al., 2019; Fettig et al., 2019; Camarero et al., 2021; Martinez Del Castillo et al., 2022). Populations in these marginal habitats are more vulnerable to additional stressors, such as drought-induced hydraulic failure or nutrient imbalances, making them mortality hotspots under climate extremes. However, differences in the climatic niches and geographic distributions of the two species likely contribute to the observed disparity in predicted mortality. P. kesiya is primarily distributed in the central and southern regions of Yunnan, characterized by warm and humid climates, where individuals may be less pre-adapted to severe or prolonged drought (Liu et al., 2010). Consequently, the frequent occurrence of extreme drought events from 2013-2022 in these areas may have exerted a disproportionately strong impact on P. kesiya, driving its generally higher predicted mortality relative to P. yunnanensis . In contrast, P. yunnanensis, which occupies the central and northern parts of Yunnan, is more commonly exposed to cooler and drier conditions (Wu and Zhu, 1987). Such climatic adaptation may confer greater drought resilience, allowing populations to withstand extreme events more effectively, despite the occurrence of high mortality in some northern range edges and high drought-frequency zones. The strong spatial overlap between high-mortality zones and areas of recurrent extreme drought highlights the central role of cumulative drought stress as a mortality driver (Gazol et al., 2025). Repeated exposure to drought can induce progressive xylem cavitation, deplete non-structural carbohydrate reserves, and weaken defensive capacity against biotic agents, thereby accelerating mortality (Allen et al., 2010; Anderegg et al., 2013). Nevertheless, caution is warranted in interpreting these spatial patterns, as high-mortality predictions also largely coincide with the extent of field sampling plots. This overlap raises the possibility of spatial sampling bias-overrepresenting high-stress habitats while underrepresenting more resilient sites (Dormann et al., 2007). To enhance the robustness and generality of future mortality models, broader sampling efforts across both high- and low-risk regions are needed, coupled with the integration of physiological trait data that capture species-specific drought tolerance strategies. Such improvements would not only refine predictive accuracy but also provide an empirical foundation for management interventions, such as selective thinning, assisted migration of drought-tolerant provenances, or restoration initiatives targeting marginal populations. Ultimately, these strategies could strengthen the adaptive capacity of Yunnan’s pine forests to withstand the intensifying impacts of climate extremes (McDowell et al., 2020). Our study provides novel insights into the drivers of mortality in P. kesiya and P. yunnanensis, yet several limitations should be acknowledged. Mortality estimates based on crown dieback may not fully capture underlying physiological processes. Differences in spatial resolution and the relatively limited set of predictors may also have introduced uncertainty and overlooked critical mechanisms. In addition, the restricted plot coverage constrains the generalization of our models. Future efforts should focus on incorporating predictors with higher ecological relevance and finer resolution, and on integrating physiological indicators and multi-scale datasets, to develop more mechanistic and accurate models that better inform forest management under intensifying drought. 5 Conculsions This study presents the first spatially explicit assessment of landscape-scale mortality events in P. yunnanensis and P. kesiya var. langbianensis during the severe 2023 drought in Yunnan, Southwest China. By integrating UAV-based plot surveys with remote sensing and environmental data, we demonstrate contrasting species-specific drivers of tree mortality. Stand age, soil nutrient availability, and canopy structure were key predictors for P. yunnanensis, whereas slope was the dominant factor for P. kesiya var. langbianensis . Such divergence stress the need for species-specific management strategies-with conservation of P. yunnanensis focusing on nutrient dynamics and demographic regulation, while protection of P. kes iya should prioritize topographically vulnerable sites and stand density optimization to buffer against drought-induced die-off. Despite these differences, NDWI decline during the dry season consistently ranked among the top predictors for both species, emphasizing the critical role of vegetation water stress in drought-induced mortality. Notably, mortality hotspots were concentrated at the range margins of both species, regions that have experienced recurrent extreme droughts over the past decade. These findings highlight the ecological thresholds imposed by cumulative climate stress and habitat suitability. This study establishes a scalable framework for mapping vulnerability and targeting adaptive management practices in pine forests facing increasing climatic extremes. Conflict of interests The authors declare that they have no known competing financial interests or personal relationships that could have appeared to influence the work reported in this paper.

Aber, J., McDowell, W., Nadelhoffer, K., Magill, A., Berntson, G., Kamakea, M., McNulty, S., Currie, W., Rustad, L., Fernandez, I., 1998. Nitrogen saturation in temperate forest ecosystems. BioScience. 48, 921–934. https://doi.org/10.2307/1313296. Allen, C.D., Macalady, A.K., Chenchouni, H., Bachelet, D., McDowell, N., Vennetier, M., Kitzberger, T., Rigling, A., Breshears, D.D., Hogg, E.H. (Ted), Gonzalez, P., Fensham, R., Zhang, Z., Castro, J., Demidova, N., Lim, J.-H., Allard, G., Running, S.W., Semerci, A., Cobb, N., 2010. A global overview of drought and heat-induced tree mortality reveals emerging climate change risks for forests. For. Ecol. Manag. 259, 660–684. https://doi.org/10.1016/j.foreco.2009.09.001. Anderegg, W.R.L., 2015. Spatial and temporal variation in plant hydraulic traits and their relevance for climate change impacts on vegetation. N. Phytol. 205, 1008–1014. https://doi.org/10.1111/nph.12907. Anderegg, W.R.L., Kane, J.M., Anderegg, L.D.L., 2013. Consequences of widespread tree mortality triggered by drought and temperature stress. Nat. Clim. Change. 3, 30–36. https://doi.org/10.1038/nclimate1635. Anderson, L.O., Malhi, Y., Aragão, L.E.O.C., Ladle, R., Arai, E., Barbier, N., Phillips, O., 2010. Remote sensing detection of droughts in Amazonian forest canopies. N. Phytol. 187, 733–750. https://doi.org/10.1111/j.1469-8137.2010.03355.x. Battie‐Laclau, P., Laclau, J., Domec, J., Christina, M., Bouillet, J., De Cassia Piccolo, M., De Moraes Gonçalves, J.L., Moreira, R.M.E., Krusche, A.V., Bouvet, J., Nouvellon, Y., 2014. Effects of potassium and sodium supply on drought‐adaptive mechanisms in Eucalyptus grandis plantations. N. Phytol. 203, 401–413. https://doi.org/10.1111/nph.12810. Bennett, A.C., McDowell, N.G., Allen, C.D., Anderson-Teixeira, K.J., 2015. Larger trees suffer most during drought in forests worldwide. Nat. Plants. 1, 15139. https://doi.org/10.1038/nplants.2015.139. Bose, A.K., Doležal, J., Scherrer, D., Altman, J., Ziche, D., Martínez-Sancho, E., Bigler, C., Bolte, A., Colangelo, M., Dorado-Liñán, I., Drobyshev, I., Etzold, S., Fonti, P., Gessler, A., Kolář, T., Koňasová, E., Korznikov, K.A., Lebourgeois, F., Lucas-Borja, M.E., Menzel, A., Neuwirth, B., Nicolas, M., Omelko, A.M., Pederson, N., Petritan, A.M., Rigling, A., Rybníček, M., Scharnweber, T., Schröder, J., Silla, F., Sochová, I., Sohar, K., Ukhvatkina, O.N., Vozmishcheva, A.S., Zweifel, R., Camarero, J.J., 2024. Revealing legacy effects of extreme droughts on tree growth of oaks across the Northern Hemisphere. Sci. Total Environ. 926, 172049. https://doi.org/10.1016/j.scitotenv.2024.172049. Breiman, L., 2001. Random Forests. Mach. Learn. 45, 5–32. https://doi.org/10.1023/A:1010933404324. Breshears, D.D., Cobb, N.S., Rich, P.M., Price, K.P., Allen, C.D., Balice, R.G., Romme, W.H., Kastens, J.H., Floyd, M.L., Belnap, J., Anderson, J.J., Myers, O.B., Meyer, C.W., 2005. Regional vegetation die-off in response to global-change-type drought. Proc. Natl. Acad. Sci. 102, 15144–15148. https://doi.org/10.1073/pnas.0505734102. Buras, A., Schunk, C., Zeiträg, C., Herrmann, C., Kaiser, L., Lemme, H., Straub, C., Taeger, S., Gößwein, S., Klemmt, H.-J., Menzel, A., 2018. Are Scots pine forest edges particularly prone to drought-induced mortality? Environ. Res. Lett. 13, 025001. https://doi.org/10.1088/1748-9326/aaa0b4. Camarero, J.J., Gazol, A., Sangu, G., Vicente-Serrano, S.M., 2015. To die or not to die: early warnings of tree dieback in response to a severe drought. J. Ecol. 103, 44-57. https://doi.org/10.1111/1365-2745.12295. Camarero, J.J., Gazol, A., Sangüesa-Barreda, G., Vergarechea, M., Alfaro-Sánchez, R., Cattaneo, N., Vicente-Serrano, S.M., 2021. Tree growth is more limited by drought in rear-edge forests most of the times. For. Ecosyst. 8, 25. https://doi.org/10.1186/s40663-021-00303-1. Camarero, J.J., Pizarro, M., Gernandt, D.S., Gazol, A., 2024. Smaller conifers are more resilient to drought. Agric. For. Meteorol. 350, 109993. https://doi.org/10.1016/j.agrformet.2024.109993. Castagneri, D., Nola, P., Cherubini, P., Motta, R., 2012. Temporal variability of size–growth relationships in a Norway spruce forest: the influences of stand structure, logging, and climate. Can. J. For. Res. 42, 550–560. https://doi.org/10.1139/x2012-007. Chambi-Legoas, R., Chaix, G., Tomazello-Filho, M., 2020. Effects of potassium/sodium fertilization and throughfall exclusion on growth patterns of Eucalyptus grandis W. Hill ex Maiden during extreme drought periods. New. For. 51, 21–40. https://doi.org/10.1007/s11056-019-09716-x. Chen, F., Wang, J., Sun, B., Chen, X., Yang, Z., Duan, Z., 2012. Relationship between geographical distribution of Pinus yunnanensis and climate. For. Res. 25, 163–168. https://doi:10.3969/j.issn.1001-1498.2012.02.009. Choat, B., Brodribb, T.J., Brodersen, C.R., Duursma, R.A., López, R., Medlyn, B.E., 2018. Triggers of tree mortality under drought. Nature. 558, 531–539. https://doi.org/10.1038/s41586-018-0240-x. Corlett, R.T., 2016. The Impacts of Droughts in Tropical Forests. Trends Plant Sci. 21, 584–593. https://doi.org/10.1016/j.tplants.2016.02.003. Crouchet, S.E., Jensen, J., Schwartz, B.F., Schwinning, S., 2019. Tree Mortality After a Hot Drought: Distinguishing Density-Dependent and -Independent Drivers and Why It Matters. Front. For. Glob. Change. 2, 21. https://doi.org/10.3389/ffgc.2019.00021. Davis, F.W., Synes, N.W., Fricker, G.A., McCullough, I.M., Serra-Diaz, J.M., Franklin, J., Flint, A.L., 2019. LiDAR-derived topography and forest structure predict fine-scale variation in daily surface temperatures in oak savanna and conifer forest landscapes. Agric. For. Meteorol. 269–270, 192–202. https://doi.org/10.1016/j.agrformet.2019.02.015. Dietrich, W.E., Perron, J.T., 2006. The search for a topographic signature of life. Nature. 439, 411–418. https://doi.org/10.1038/nature04452. Duursma, R.A., Barton, C.V.M., Eamus, D., Medlyn, B.E., Ellsworth, D.S., Forster, M.A., Tissue, D.T., Linder, S., McMurtrie, R.E., 2011. Rooting depth explains [CO2] x drought interaction in Eucalyptus saligna . Tree Physiol. 31, 922–931. https://doi.org/10.1093/treephys/tpr030. F. Dormann, C., M. McPherson, J., B. Araújo, M., Bivand, R., Bolliger, J., Carl, G., G. Davies, R., Hirzel, A., Jetz, W., Daniel Kissling, W., Kühn, I., Ohlemüller, R., R. Peres‐Neto, P., Reineking, B., Schröder, B., M. Schurr, F., Wilson, R., 2007. Methods to account for spatial autocorrelation in the analysis of species distributional data: a review. Ecography. 30, 609–628. https://doi.org/10.1111/j.2007.0906-7590.05171.x Fettig, C.J., Mortenson, L.A., Bulaon, B.M., Foulk, P.B., 2019. Tree mortality following drought in the central and southern Sierra Nevada, California, U.S. For. Ecol. Manag. 432, 164–178. https://doi.org/10.1016/j.foreco.2018.09.006. Forrester, D.I., Plaga, B.N.E., Bauhus, J., 2025. The effects of tree size, stand density, and tree-species mixing on stand level and tree level light absorption and light-use efficiency: a review. Curr. For. Rep. 11, 15. https://doi.org/10.1007/s40725-025-00247-7. Friedman, J.H., 2001. Greedy function approximation: A gradient boosting machine. Ann. Stat. 29. https://doi.org/10.1214/aos/1013203451. Gao, B., 1996. NDWI—A normalized difference water index for remote sensing of vegetation liquid water from space. Remote Sens. Environ. 58, 257–266. https://doi.org/10.1016/S0034-4257(96)00067-3. Gao, D., Xaybouangeun, N., Zaw, Z., Yang, R., Fan, Z., 2025. Growth resilience of Pinus latteri to extreme drought events across aridity gradients in southern Laos. Glob. Ecol. Conserv. 57, e03424. https://doi.org/10.1016/j.gecco.2025.e03424. Gazol, A., Camarero, J.J., Anderegg, W.R.L., Vicente‐Serrano, S.M., 2017. Impacts of droughts on the growth resilience of Northern Hemisphere forests. Glob. Ecol. Biogeogr. 26, 166–176. https://doi.org/10.1111/geb.12526. Gazol, A., Pizarro, M., Hammond, W.M., Allen, C.D., Camarero, J.J., 2025. Droughts preceding tree mortality events have increased in duration and intensity, especially in dry biomes. Nat. Commun. 16, 5779. https://doi.org/10.1038/s41467-025-60856-5. González De Andrés, E., Gazol, A., Querejeta, J.I., Igual, J.M., Colangelo, M., Sánchez‐Salguero, R., Linares, J.C., Camarero, J.J., 2022. The role of nutritional impairment in carbon‐water balance of silver fir drought‐induced dieback. Glob. Change. Biol. 28, 4439–4458. https://doi.org/10.1111/gcb.16170. Greenwell, B., M., 2017. Pdp: an r package for constructing partial dependence plots. R J. 9, 421. https://doi.org/10.32614/RJ-2017-016. Grote, R., Gessler, A., Hommel, R., Poschenrieder, W., Priesack, E., 2016. Importance of tree height and social position for drought-related stress on tree growth and mortality. Trees. 30, 1467–1482. https://doi.org/10.1007/s00468-016-1446-x. Guarín, A., Taylor, A.H., 2005. Drought triggered tree mortality in mixed conifer forests in Yosemite National Park, California, USA. For. Ecol. Manag. 218, 229–244. https://doi.org/10.1016/j.foreco.2005.07.014. Güsewell, S., 2004. N : P ratios in terrestrial plants: variation and functional significance. N. Phytol. 164, 243–266. https://doi.org/10.1111/j.1469-8137.2004.01192.x. Hammond, W.M., Williams, A.P., Abatzoglou, J.T., Adams, H.D., Klein, T., López, R., Sáenz-Romero, C., Hartmann, H., Breshears, D.D., Allen, C.D., 2022. Global field observations of tree die-off reveal hotter-drought fingerprint for Earth’s forests. Nat. Commun. 13, 1761. https://doi.org/10.1038/s41467-022-29289-2. Herms, D.A., Mattson, W.J., 1992. The Dilemma of Plants: To Grow or Defend. Q. Rev. Biol. 67, 283–335. https://doi.org/10.1086/417659. Hijmans, R. J. 2023. Terra: Spatial Data Analysis. R package version 1.7-46 [Computer software]. https://cran.r-project.org/package=terra. Howe, M., Peng, L., Carroll, A., 2022. Landscape predictions of western balsam bark beetle activity implicate warm temperatures, a longer growing season, and drought in widespread irruptions across British Columbia. For. Ecol. Manag. 508, 120047. https://doi.org/10.1016/j.foreco.2022.120047. Italiano, S.S.P., Camarero, J.J., Borghetti, M., Colangelo, M., Pizarro, M., Ripullone, F., 2023. Radial growth, wood anatomical traits and remote sensing indexes reflect different impacts of drought on Mediterranean forests. For. Ecol. Manag. 548, 121406. https://doi.org/10.1016/j.foreco.2023.121406. IPCC, 2023. Climate Change 2023: Synthesis Report. Contribution of Working Groups I, II and III to the Sixth Assessment Report of the Intergovernmental Panel on Climate Change [Core Writing Team, H. Lee and J. Romero (eds.)]. IPCC, Geneva, Switzerland, pp. 35-115. https://doi.org/10.59327/ipcc/ar6-9789291691647. Jääskeläinen, J., Junttila, S., O’Sullivan, H., Cheng, Y., Horion, S., Vastaranta, M., 2025. Quantifying the drivers of tree mortality: A case study from urban recreational boreal forest. Urban For. Urban Green. 104, 128672. https://doi.org/10.1016/j.ufug.2025.128672. Ji, Y., Zhou, G., Li, Z., Wang, S., Zhou, H., Song, X., 2020. Triggers of widespread dieback and mortality of poplar (Populus spp.) plantations across northern China. J. Arid Environ. 174, 104076. https://doi.org/10.1016/j.jaridenv.2019.104076. Johnston, J.D., Stephens, P., Schmidt, M., Jeronimo, S.M.A., Brackett, A., Wuenschel, A., Ulrich, D.E.M., Paine, R., Duloisy, L., Greenler, S., Larson, A.J., Churchill, D.J., 2025. Significant mortality of old trees across a dry forest landscape, Oregon, USA. For. Ecol. Manag. 593, 122907. https://doi.org/10.1016/j.foreco.2025.122907. Jones, G.M., Keane, J.J., Gutiérrez, R.J., Peery, M.Z., 2018. Declining old‐forest species as a legacy of large trees lost. Divers. Distrib. 24, 341–351. https://doi.org/10.1111/ddi.12682. Kamara, M., Kamruzzaman, Md., 2020. Self-thinning process, dynamics of aboveground biomass, and stand structure in overcrowded mangrove Kandelia obovata stand. Reg. Stud. Mar. Sci. 38, 101375. https://doi.org/10.1016/j.rsma.2020.101375. Keshavarz, M.R., Vazifedoust, M., Alizadeh, A., 2014. Drought monitoring using a Soil Wetness Deficit Index (SWDI) derived from MODIS satellite data. Agric. Water Manag. 132, 37–45. https://doi.org/10.1016/j.agwat.2013.10.004. Klouček, T., Modlinger, R., Zikmundová, M., Kycko, M., Komárek, J., 2024. Early detection of bark beetle infestation using UAV-borne multispectral imagery: a case study on the spruce forest in the Czech Republic. Front. For. Glob. Change. 7, 1215734. https://doi.org/10.3389/ffgc.2024.1215734. Kuhn, M., 2008. Building Predictive Models in R Using the caret Package. J. Stat. Softw. 28, 1–26. https://doi.org/10.18637/jss.v028.i05. Leo Breiman, 2001. Statistical Modeling: The Two Cultures (with comments and a rejoinder by the author). Stat. Sci. 16, 199–231. https://doi.org/10.1214/ss/1009213726. Liu, Y., Liu, Y., Feng, X., Meng, M., Fu, Q., Zhang, Q., Zhang, W., 2010. Research Situation and Prospects of Pinus kesiya var. 1angbianensis Plantation. J. Anhui. agri. Sci. 38, 18982–18983. https://doi:10.3969/j.issn.0517-6611.2010.33.136. Lundberg, S.M., Lee, S.-I., 2017. A unified approach to interpreting model predictions, in: Proceedings of the 31st International Conference on Neural Information Processing Systems, NIPS’17. Curran Associates Inc., Red Hook, NY, USA, pp. 4768–4777. Luo, Y., Chen, H.Y.H., 2011. Competition, species interaction and ageing control tree mortality in boreal forests. J. Ecol. 99, 1470–1480. https://doi.org/10.1111/j.1365-2745.2011.01882.x. Ma, T., Chen, W., Cai, Q., Dong, Z., Wang, L., Hu, P., Gao, L., Garfinkel, C.I., 2024. Attribution analysis of the persistent and extreme drought in southwest China during 2022–2023. Environ. Res. Lett. 19, 114056. https://doi.org/10.1088/1748-9326/ad8171. Martinez Del Castillo, E., Zang, C.S., Buras, A., Hacket-Pain, A., Esper, J., Serrano-Notivoli, R., Hartl, C., Weigel, R., Klesse, S., Resco De Dios, V., Scharnweber, T., Dorado-Liñán, I., Van Der Maaten-Theunissen, M., Van Der Maaten, E., Jump, A., Mikac, S., Banzragch, B.-E., Beck, W., Cavin, L., Claessens, H., Čada, V., Čufar, K., Dulamsuren, C., Gričar, J., Gil-Pelegrín, E., Janda, P., Kazimirovic, M., Kreyling, J., Latte, N., Leuschner, C., Longares, L.A., Menzel, A., Merela, M., Motta, R., Muffler, L., Nola, P., Petritan, A.M., Petritan, I.C., Prislan, P., Rubio-Cuadrado, Á., Rydval, M., Stajić, B., Svoboda, M., Toromani, E., Trotsiuk, V., Wilmking, M., Zlatanov, T., De Luis, M., 2022. Climate-change-driven growth decline of European beech forests. Commun. Biol. 5, 163. https://doi.org/10.1038/s42003-022-03107-3. Marusig, D., Petruzzellis, F., Tomasella, M., Napolitano, R., Altobelli, A., Nardini, A., 2020. Correlation of Field-Measured and Remotely Sensed Plant Water Status as a Tool to Monitor the Risk of Drought-Induced Forest Decline. Forests. 11, 77. https://doi.org/10.3390/f11010077. Matusick, G., Ruthrof, K.X., Brouwers, N.C., Dell, B., Hardy, G.St.J., 2013. Sudden forest canopy collapse corresponding with extreme drought and heat in a mediterranean-type eucalypt forest in southwestern Australia. Eur. J. For. Res. 132, 497–510. https://doi.org/10.1007/s10342-013-0690-5. McDowell, N., Pockman, W.T., Allen, C.D., Breshears, D.D., Cobb, N., Kolb, T., Plaut, J., Sperry, J., West, A., Williams, D.G., Yepez, E.A., 2008. Mechanisms of plant survival and mortality during drought: why do some plants survive while others succumb to drought? N. Phytol. 178, 719–739. https://doi.org/10.1111/j.1469-8137.2008.02436.x McDowell, N.G., Allen, C.D., Anderson-Teixeira, K., Aukema, B.H., Bond-Lamberty, B., Chini, L., Clark, J.S., Dietze, M., Grossiord, C., Hanbury-Brown, A., Hurtt, G.C., Jackson, R.B., Johnson, D.J., Kueppers, L., Lichstein, J.W., Ogle, K., Poulter, B., Pugh, T.A.M., Seidl, R., Turner, M.G., Uriarte, M., Walker, A.P., Xu, C., 2020. Pervasive shifts in forest dynamics in a changing world. Science. 368, eaaz9463. https://doi.org/10.1126/science.aaz9463. Meyer, M.D., Kelt, D.A., North, M.P., 2005. Nest trees of northern flying squirrels in the sierra nevada. J. Mammal. 86, 275–280. https://doi.org/10.1644/BEH-110.1 Molnar, C., 2020. Interpretable Machine Learning. Leanpub. Moreno-Fernández, D., Camarero, J.J., García, M., Lines, E.R., Sánchez-Dávila, J., Tijerín, J., Valeriano, C., Viana-Soto, A., Zavala, M.Á., Ruiz-Benito, P., 2022. The interplay of the tree and stand-level processes mediate drought-induced forest dieback: evidence from complementary remote sensing and tree-ring approaches. Ecosystems. 25, 1738–1753. https://doi.org/10.1007/s10021-022-00793-2. Mu, Z., Asensio, D., Sardans, J., Ogaya, R., Llusià, J., Filella, I., Liu, L., Wang, X., Peñuelas, J., 2023. Chronic drought alters extractable concentrations of mineral elements in Mediterranean forest soils. Sci. Total Environ. 905, 167062. https://doi.org/10.1016/j.scitotenv.2023.167062. Obladen, N., Dechering, P., Skiadaresis, G., Tegel, W., Keßler, J., Höllerl, S., Kaps, S., Hertel, M., Dulamsuren, C., Seifert, T., Hirsch, M., Seim, A., 2021. Tree mortality of European beech and Norway spruce induced by 2018-2019 hot droughts in central Germany. Agric. For. Meteorol. 307, 108482. https://doi.org/10.1016/j.agrformet.2021.108482. Paz‐Kagan, T., Brodrick, P.G., Vaughn, N.R., Das, A.J., Stephenson, N.L., Nydick, K.R., Asner, G.P., 2017. What mediates tree mortality during drought in the southern Sierra Nevada? Ecol. Appl. 27, 2443–2457. https://doi.org/10.1002/eap.1620. Pfeifer, E.M., Hicke, J.A., Meddens, A.J.H., 2011. Observations and modeling of aboveground tree carbon stocks and fluxes following a bark beetle outbreak in the western United States: carbon cycling following bark beetle outbreak. Glob. Change Biol. 17, 339–350. https://doi.org/10.1111/j.1365-2486.2010.02226.x. Queally, N., Zheng, T., Ye, Z., Kovach, K.R., Pavlick, R., Shafron, E., Schneider, F.D., Townsend, P.A., 2025. Functional traits from imaging spectroscopy inform patterns of forest mortality during Sierra Nevada drought. Glob. Change Biol. 31, e70246. https://doi.org/10.1111/gcb.70246. Rambo, T.R., North, M.P., 2009. Canopy microclimate response to pattern and density of thinning in a Sierra Nevada forest. For. Ecol. Manag. 257, 435–442. https://doi.org/10.1016/j.foreco.2008.09.029. Rodríguez, A., Chiti, T., Rey, A., Durán, J., 2020. Forest die-off reduces soil C and N content and increases C stability in a Mediterranean woodland. Geoderma. 359, 113990. https://doi.org/10.1016/j.geoderma.2019.113990 Rozas, V., Sampedro, L., 2013. Soil chemical properties and dieback of Quercus robur in Atlantic wet forests after a weather extreme. Plant Soil. 373, 673–685. https://doi.org/10.1007/s11104-013-1835-5. Sardans, J., Peñuelas, J., 2007. Drought changes phosphorus and potassium accumulation patterns in an evergreen Mediterranean forest. Funct. Ecol. 21, 191–201. https://doi.org/10.1111/j.1365-2435.2007.01247.x. Schönbeck, L., Li, M.-H., Lehmann, M.M., Rigling, A., Schaub, M., Hoch, G., Kahmen, A., Gessler, A., 2021. Soil nutrient availability alters tree carbon allocation dynamics during drought. Tree Physiol. 41, 697–707. https://doi.org/10.1093/treephys/tpaa139. Schwinning, S., Ehleringer, J.R., 2001. Water use trade-offs and optimal adaptations to pulse-driven arid ecosystems. J. Ecol. 89, 464–480. https://doi.org/10.1046/j.1365-2745.2001.00576.x. Shapley, L.S., 1953. Stochastic Games*. Proc. Natl. Acad. Sci. 39, 1095–1100. https://doi.org/10.1073/pnas.39.10.1095. Stoneman, G.L., Crombie, D.S., Whitford, K., Hingston, F.J., Giles, R., Portlock, C.C., Galbraith, J.H., Dimmock, G.M., 1997. Growth and water relations of Eucalyptus marginata (jarrah) stands in response to thinning and fertilization. Tree Physiol. 17, 267–274. https://doi.org/10.1093/treephys/17.4.267. Stovall, A.E.L., Shugart, H., Yang, X., 2019. Tree height explains mortality risk during an intense drought. Nat. Commun. 10, 4385. https://doi.org/10.1038/s41467-019-12380-6. Sturm, J., Santos, M.J., Schmid, B., Damm, A., 2022. Satellite data reveal differential responses of Swiss forests to unprecedented 2018 drought. Glob. Change Biol. 28, 2956–2978. https://doi.org/10.1111/gcb.16136. Swain, S., Wardlow, B.D., Narumalani, S., Tadesse, T., Callahan, K., 2011. Assessment of vegetation response to drought in nebraska using terra-modis land surface temperature and normalized difference vegetation index. GIScience Remote Sens. 48, 432–455. https://doi.org/10.2747/1548-1603.48.3.432. Touche, J., Calvaruso, C., De Donato, P., Turpault, Mp., 2022. Five successive years of rainfall exclusion induce nutritional stress in a mature beech stand. For. Ecol. Manag. 507, 119987. https://doi.org/10.1016/j.foreco.2021.119987. Trubin, A., Kozhoridze, G., Zabihi, K., Modlinger, R., Singh, V.V., Surový, P., Jakuš, R., 2024. Detection of green attack and bark beetle susceptibility in Norway Spruce: Utilizing PlanetScope Multispectral Imagery for Tri-Stage spectral separability analysis. For. Ecol. Manag. 560, 121838. https://doi.org/10.1016/j.foreco.2024.121838. Van Gunst, K.J., Weisberg, P.J., Yang, J., Fan, Y., 2016. Do denser forests have greater risk of tree mortality: A remote sensing analysis of density-dependent forest mortality. For. Ecol. Manag. 359, 19–32. https://doi.org/10.1016/j.foreco.2015.09.032. Vulova, S., Horn, K., Rocha, A.D., Brill, F., Somogyvári, M., Okujeni, A., Förster, M., Kleinschmit, B., 2025. Unraveling the response of forests to drought with explainable artificial intelligence (XAI). Ecol. Indic. 172, 113308. https://doi.org/10.1016/j.ecolind.2025.113308. Wang, L., Ma, S., 2024. Extreme winter-spring drought in Southwest China in 2023: response to the phase transition from La Niña to El Niño. Environ. Res. Lett. 19, 084042. https://doi.org/10.1088/1748-9326/ad64ee. West, G.B., Brown, J.H., Enquist, B.J., 1999. A general model for the structure and allometry of plant vascular systems. Nature 400, 664–667. https://doi.org/10.1038/23251 Wu, Z., Zhu, Y., 1987. Vegetation of Yunnan. Science Press, Beijing, pp. 400-410. Xu, P., Zhou, T., Yi, C., Fang, W., Hendrey, G., Zhao, X., 2018. Forest drought resistance distinguished by canopy height. Environ. Res. Lett. 13, 075003. https://doi.org/10.1088/1748-9326/aacadd. Xue, L., Feng, H., Chen, F., 2010. Time-trajectory of mean component weight and density in self-thinning Pinus densiflora stands. Eur. J. For. Res. 129, 1027–1035. https://doi.org/10.1007/s10342-010-0387-y. Yan, Y., Hong, S., Chen, A., Peñuelas, J., Allen, C.D., Hammond, W.M., Munson, S.M., Myneni, R.B., Piao, S., 2025. Satellite-based evidence of recent decline in global forest recovery rate from tree mortality events. Nat. Plants. 11, 731–742. https://doi.org/10.1038/s41477-025-01948-4. Zhang, Xiang, Zhang, Xu, Terfa, B.K., Nam, W.-H., Zeng, J., Ma, H., Gu, X., Du, W., Wang, C., Yang, J., Wang, P., Niyogi, D., Chen, N., 2024. Mapping global drought-induced forest mortality based on multiple satellite vegetation optical depth data. Remote Sens. Environ. 315, 114406. https://doi.org/10.1016/j.rse.2024.11 Supplementary Material Information & Authors Information Version history Copyright This work is licensed under a Non Exclusive No Reuse License.

Authors Metrics & Citations Metrics Article Usage 215views 130downloads Citations Download citation Dao-Xiong Gao, Rao-Qiong Yang, Pei-Li Fu, et al. Vegetation Water Status and Topography Shape Drought-Induced Mortality of Two Pine Species in Southwest China. Authorea. 27 October 2025. DOI: https://doi.org/10.22541/au.176157213.31136399/v1 DOI: https://doi.org/10.22541/au.176157213.31136399/v1 If you have the appropriate software installed, you can download article citation data to the citation manager of your choice. Simply select your manager software from the list below and click Download. For more information or tips please see 'Downloading to a citation manager' in the Help menu.