Revealing the dynamic changes in tree spatial patterns holds significant implications for the restoration practices of natural secondary forests. We analyzed the dynamics of tree spatial patterns using 18 years (2005–2023) of repeated survey data from a 4-ha subalpine natural secondary forest on the eastern Qinghai–Tibet Plateau, China. The results showed: (1) Tree density decreased by 40.31% over time. The dominant tree species, Minjiang fir (Abies fargesii var. faxoniana), exhibited an expanding population with abundant young individuals and few old trees. In contrast, the pioneer species Betula albosinensis, Acer sinense, and others exhibited declining population trends. (2) Dominant tree species showed aggregated patterns at small scales (< 15 m), and their aggregation scales decreased after 18 years. (3) Spatial associations varied in species pairs over 18 years. B. albosinensis and Minjiang fir changed from repulsion to independence, while B. albosinensis and A. sinense changed from attraction to independence. However, Minjiang fir and Picea asperata exhibited no spatial association at either census. (4) We found that density-dependent mortality occurred in both the Minjiang fir and B. albosinensis populations at different periods throughout the survey. Intense competition prevailed in the subalpine natural secondary forests. Therefore, thinning can help alleviate this competition. Our study provides long-term evidence for successional pathways and species-specific dynamics in subalpine natural secondary forests, informing future restoration planning.
Understanding the spatiotemporal variability of climate–growth relationships is critical for predicting forest responses to environmental change and guiding sustainable forest management. However, research on this subject in subtropical conifer species remains limited. We quantified spatiotemporal growth patterns and climatic drivers of Pinus massoniana Lamb across 32 sites spanning the climate gradient of subtropical China. Across the study area, the overall growth of P. massoniana exhibited an increasing trend, with growth declines observed only at a few sites in the central region. Temperature was positively correlated with growth in colder northern regions but negatively correlated with growth in warmer southern regions, whereas growth in the central region showed a significant positive correlation with relative humidity. This indicates that tree growth is primarily constrained by region-specific climatic factors, including high temperatures in the south, drought in the central region, and low temperatures in the north. Furthermore, drought stress has increasingly limited the growth of central and southern P. massoniana populations over time. Simulations based on the VS-Lite model further revealed divergent growth phenological patterns: Northern trees display a unimodal growth pattern, while central and southern populations exhibit a bimodal pattern. Under three projected future climate change scenarios, tree growth in the central and southern regions is projected to exhibit more pronounced declines. This study highlights the importance of accounting for the spatiotemporal variability of climate–growth relationships when developing forest management plans, including those for future plantations. Effective management should prioritize region-specific strategies to mitigate the impacts of climate change, ensuring the resilience of P. massoniana subtropical forests.
Habitat heterogeneity may influence conspecific negative density dependence (CNDD) in karst forests through direct effects and indirect pathways mediated by shifts in functional traits, with consequences for species coexistence and diversity maintenance. However, how habitat heterogeneity and functional traits jointly regulate CNDD, and whether trait-mediated effects differ among habitat contexts, remain unclear. Using 177,893 individuals representing 251 woody species from the 25-ha Maolan Forest Dynamics Plot in Southwestern China, we estimated species-specific CNDD within hydrological, light, and topographic heterogeneity classes. We tested whether resistance-tolerance traits (including mycorrhizal association and shade tolerance) and competition-colonization traits (seed dispersal mode) explained CNDD variation and modulated its response to habitat heterogeneity. Reliable estimates were obtained for 92 species (36.65%). Habitat heterogeneity affected CNDD in a dimension-specific manner: Terrain ruggedness (TR) weakened CNDD, whereas topographic wetness index (TWI) strengthened it. CNDD differed among trait categories: it was strongest in arbuscular mycorrhizal (AM) species, stronger in autochorous and zoochorous than in anemochorous species, and strongest in intermediate shade-tolerant species. Structural equation models showed that habitat heterogeneity indirectly influenced community-level CNDD by shifting importance-value-weighted trait composition toward ectomycorrhizal (EcM), anemochorous, and shade-tolerant species and away from AM, zoochorous, and shade-intolerant species. Trait modulation was habitat-dependent, with the strongest contrast between the TR and TWI gradients: AM and EcM species showed positive slopes along TR but negative slopes along TWI. In full models, increasing TR heterogeneity weakened CNDD in anemochorous species with AM or EcM associations, whereas increasing TWI heterogeneity strengthened CNDD in shade-tolerant zoochorous species. Significant trait-specific responses generally remained consistent with the baseline direction within each habitat dimension. These findings indicated that CNDD in karst forests is jointly shaped by habitat filtering, dispersal, mycorrhizal association, and plant ecological strategies, highlighting the importance of habitat-trait matching for forest regeneration and diversity maintenance.
Vegetation greening and climate variability (CV) control evaporative cooling (EC) and are crucial in mitigating global warming. EC has been investigated because of vegetation greening; however, the direct impact of CV on EC is not fully understood. Therefore, we separated the contributions of vegetation greening and CV to changes in evapotranspiration using the Penman–Monteith model and time-series analysis. The increase in evapotranspiration during 1982–2024 decreased the global temperature by 0.75 °C. CV contributed to EC more than vegetation greening did, at −0.63 and −0.12 °C, respectively. Increased atmospheric moisture demand was the main climatic factor driving EC in the tropical forests of South America and Africa, while vegetation greening played an important role in EC in Eastern and Southern Asia. However, the weak evaporative warming attributable to vegetation browning in Western North America, Southern South America, Southern Africa, and the Amazon was likely due to compound drought and deforestation decreasing transpiration. These results further improve our understanding of CV’s roles and leaf area index changes in regulating EC.
Conspecific negative density dependence (CNDD) is widely regarded as a key mechanism maintaining plant diversity and is classically attributed to host-specific pathogens or herbivores. However, whether animal-mediated seed dispersal may contribute to CNDD-like spatial patterns remains poorly understood. Such animal-mediated seed dispersal behaviors are widespread in forests but are difficult to study because of their stochastic nature. Here, we investigated the distance-dependent spatial distribution of seedlings of two Cyclobalanopsis species around parent trees (at seven distances from 0 to 30 m) in a tropical forest, comparing directions with and without observed nut-storage activity by two squirrel species that actively carve and store nuts away from parent trees. Results showed that seedling numbers exhibited a clear nonlinear distance-dependent pattern along directions with squirrel nut-storage activity, with significantly higher numbers at 20 m than in both same-tree and different-tree controls. Pathogen infection in sapling fine roots did not show a significant distance-dependent pattern. These findings provide a new perspective on CNDD-like pattern formation by highlighting animal-mediated seed redistribution as an additional process that may be associated with seedling spatial patterns.
Mixed-species plantations are increasingly adopted as they may exhibit higher productivity and carbon sequestration potential under suitable species mixtures and management. Assessing the mixing effects typically relies on controlled biodiversity-ecosystem functioning (BEF) experiments and other forest inventory data. Often, the BEF experiments are limited to a short period (e.g., <30 years), and other forest inventory data inherit uncontrolled factors not aligned with the BEF experiments. Results from both are often inconsistent. We used a forest dynamic model to conduct simulation experiments of mixed-species forest plantations. Our simulations varied factors that may affect mixing effects, including levels of species mixture, competition control treatments for naturally regenerated species, and planting density. We informed the simulation model from the BEF experiments and validated the simulated results against forest inventory data. Our results showed that species trait dissimilarity is the primary factor affecting mixing effects, with its relative importance ranging from 20% to 40% across all simulation periods. High niche overlaps intensify competition, and distinct niches facilitate complementarity. The effects of competition control treatments are determined by functional complementarity or redundancy between the naturally regenerated and planted species. A high planting density (e.g., 4,444 stems·ha−1) enhanced interspecific interactions that amplified the effects of species trait dissimilarity up to twofold. We found that the inconsistent mixing effects from the BEF experiments and other forest inventory data are due to not including the self-thinning process, as the BEF experiments usually do not extend to this stage. Post self-thinning (at approximately 80 years after the plantation) leads to the re-optimization of stand structure that sets the basis for enhancing the mixing effects for the long term. Thus, our simulation results underscore the importance of both time span and stand development stages when assessing the mixing effects of forest plantations.
Current climate models project an increased frequency and intensity of droughts in the Eurasian forest-steppe. Although forests in this area are at risk from increasing aridization, our understanding of the forest vegetation response to increasing drought stress remains limited. Here, we examined a network of 12 new tree-ring width (TRW) chronologies of Scots pine (Pinus sylvestris L.) in the forest-steppe zone of Western Siberia to determine the effects of climate variability on tree radial growth. The data were generalized using principal component analysis (PCA) into three regional chronologies characterized by a diverse climatic signal. While the northern sites demonstrated a weak growth–climate response, TRW at the southern sites strongly correlated with the six-month standardized precipitation evapotranspiration index (SPEI6) in February–July (r = 0.65, p < 0.001). Based on this relationship, we defined the years with the extremely low summer SPEI6 (July) in the southern forest-steppe of Western Siberia for the period from 1850 to 2020. Our reconstruction identified the most severe droughts during the last two centuries in 1851, 1890, 1904, 1911, 1931, 1952, 1975–1977, 1989, 2004, 2010, and 2012. These results indicate the increasing effect of drought stress on Scots pine populations along the aridity gradient in the forest-steppe ecotone, emphasizing the potential vulnerability of trees under changing climate.
Coarse woody debris constitutes a fundamental structural and trophic component of forest ecosystems. Despite its recognized ecological importance, the role of coarse woody debris in shaping spatial organization and habitat use of sympatric species remains poorly understood. This study investigated the connection between the quantity and spatial distribution of coarse woody debris and individual space use of the yellow-necked mouse (Apodemus flavicollis) and the bank vole (Clethrionomys glareolus) in Białowieża Forest, Poland. We hypothesized that variation in coarse woody debris availability affects home range location, resulting in an above-average coarse woody debris amount in the home range. Using fine-scale radiotelemetry, 50 individuals (29 A. flavicollis, 21 C. glareolus) were tracked, yielding 2378 locations. Home range and core area sizes were estimated with autocorrelation-corrected kernel density estimators. Mean home ranges amounted to 3847 m2 for A. flavicollis and 2719 m2 for C. glareolus, with core areas comprising about 20% of total range. Coarse woody debris coverage and volume in the areas used by the tracked animals were higher (often significantly) than mean values for the study sites. It was true within home ranges, core areas, and especially near telemetry location points. We found that amounts of coarse woody debris were significantly higher than in the study sites both in the home ranges of A. flavicollis (dead wood coverage was 0.34 m2 per 100 m2 higher than the study site average) and in the core areas of C. glareolus (dead wood volume was 0.35 m3 per 100 m2 higher than the study site average). Both species preferentially used microhabitats rich in coarse woody debris, underscoring its role as a critical determinant of spatial behavior. By linking high-resolution telemetry with habitat mapping, this study demonstrates that coarse woody debris contributes to spatial niche differentiation and habitat selection among sympatric small mammals. These findings highlight the ecological necessity of maintaining coarse woody debris in forest ecosystems as a key element of biodiversity-oriented forest management.
As climate change intensifies drought severity and frequency, the stability of large-scale afforestation programs hinges on understanding the growth limits and drought-response patterns of key tree species. Robinia pseudoacacia is a critical species for ecological restoration in northern China, yet its nonlinear growth thresholds and drought-response patterns across hydro-climatic gradients remain poorly quantified. Here, we combined a regional dendrochronological network of 24 sites spanning a semi-arid to humid gradient with boosted regression trees (BRT) and partial least squares path modeling (PLS-PM) to disentangle the drivers of radial growth and drought resilience. Our results revealed divergent drought-response patterns across the gradient. Populations in arid sites exhibited a descriptive “elastic compensation” pattern, characterized by low resistance but strong post-drought recovery, whereas populations in humid sites showed a more conservative growth-maintenance pattern with higher resistance (Rt) and moderate recovery (Rc). BRT and threshold analyses identified critical thresholds for growth decline: The upper thermal threshold was substantially higher in arid sites (22.90 °C) than in humid sites (17.21 °C), indicating a region-specific shift in growth-temperature sensitivity. Furthermore, a distinct vapor pressure deficit (VPD) threshold of 0.55 kPa was detected in arid regions, above which radial growth declined within the observed data range. PLS-PM further revealed region-specific pathway structures: In arid zones, radial growth was positively associated with climate indicators and atmospheric drought and negatively associated with soil chemical properties, whereas in humid zones, climate was linked to radial growth mainly through atmospheric- and soil-drought pathways. These findings challenge uniform management approaches and suggest that future afforestation planning should consider hydrothermal thresholds and region-specific drought-response patterns to improve risk assessment for R. pseudoacacia plantations under a warming and drying climate.
Tree species are increasingly recognized as key drivers of soil biogeochemical processes, yet their combined chemical and physical effects on forest floor dynamics, microclimate, and mineral topsoil carbon stocks remain insufficiently understood. Using the Geographical Arboretum of Tervuren, Belgium, a 120-year-old common garden established on loess-derived silty Retisols, we quantified how tree species identity and mycorrhizal association shape litter traits, humus formation, microclimate buffering, and carbon distribution between the forest floor and the 0–20 cm mineral soil across 24 monospecific stands. Tree species generated distinct humus forms ranging from mull to moder and mor systems, resulting in strong contrasts in forest floor development and carbon stocks. Total carbon stocks, including the forest floor and topsoil, ranged from 7.49 kg·m−2 under arbuscular mycorrhizal (AM)-associated angiosperm stands to 12.95 kg·m−2 under ectomycorrhizal (EcM)-associated gymnosperm stands. EcM-associated gymnosperms stored a significantly larger proportion of the carbon stocks in the forest floor, around 14%, compared with 4%–6% in the other functional groups. Fresh leaf base cation concentrations were strongly associated with forest floor mass and topsoil pH, suggesting that tree species modified carbon cycling mainly through litter-induced changes in soil chemistry near a pedogenetic threshold. Forest canopies strongly buffered temperature extremes relative to the adjacent meadow, with summer maxima up to approximately 10 °C cooler inside the forest, but these physical effects varied mainly with canopy and understory characteristics and did not directly explain soil carbon stocks. Exploratory piecewise structural equation models and targeted carbon models indicated that carbon stocks were more closely linked to litter chemistry and soil acidity than to measured microclimatic conditions. These findings highlight tree species selection as a key driver of long-term soil carbon storage and vertical carbon distribution, which should therefore be explicitly considered in climate-smart forestry, restoration, and assisted migration strategies.
Commercial Eucalyptus plantations have expanded significantly in South America, with most of them being managed under short-rotation regimes for pulp production. Despite the contribution of Eucalyptus plantations to carbon sequestration, there are concerns about high water use and its potential impacts on hydrological processes. In this context, this study aims to develop a spatially explicit bi-objective linear programming framework for stand-level biophysical harvest scheduling in Eucalyptus dunnii plantations for pulp production in Uruguay. This framework integrates Sentinel-1 synthetic aperture radar, Sentinel-2 multispectral imagery, SoilGrids soil property data, and accumulated evapotranspiration (ETa) estimates derived from geeSEBAL-MODIS. Random forest models were applied to estimate above-ground carbon stock (AGC) and stand age (SA), while ETa was derived for 226 forest stands (699 ha) located on Aquic Argiudolls and Typic Albaqualfs soils across a chronosequence of 6 to 9 years. A Pareto-optimal frontier was used to simultaneously maximize the normalized AGC accumulation and the inverted normalized ETa (water-use reduction). The results indicate that a balanced forest management scenario, in which both objectives have equal weight, consists of a 7-year rotation length, corresponding to the peak in water-use efficiency (WUE). Compared to the typical 8-year fixed-rotation scheme, this balanced scenario substantially improved mean WUE (0.723 vs. 0.658 g·L−1) and reduced mean ETa (8,407 vs. 9,401 mm), with only a slight reduction in AGC (60.35 vs. 61.38 Mg·ha−1). This study provides a scalable and cost-effective tool for optimizing harvest scheduling, mitigating potential hydrological impacts while sustaining carbon sequestration in short-rotation Eucalyptus plantations.
Accurate urban tree species distribution information is essential for urban green space planning and management missions, including urban cooling assessment, carbon sequestration estimation, biodiversity conservation, and species-specific maintenance scheduling. However, pervasive shadows in very-high-resolution imagery of complex urban environments distort spectral fidelity and reduce classification accuracy. To address this problem, this study proposed a shadow-aware deep learning framework for classifying eight major urban tree species from Pléiades imagery in Nanjing, China. The framework consisted of three components: (1) A modified 6-channel shadow detection model was developed by extending the multi-scale spatial attention shadow detection network (MSASDNet) with transfer learning, and an expanded multispectral shadow dataset, Pléiades-SD, was constructed to support knowledge transfer from conventional 3-channel RGB imagery to 6-channel multispectral imagery. (2) The conventional RGB-based shadow restoration method was extended to multispectral restoration with multi-band compensation and boundary smoothing to compensate for shadow-induced spectral distortion while preserving cross-band structural information. (3) An improved DeepLabV3+ model was developed by integrating dual-level normalized attention modules (NAMs) and a multi-scale atrous spatial pyramid pooling (ASPP) fusion strategy for fine-grained tree species classification. The results indicated that the proposed 6-channel shadow detector outperformed the original 3-channel transfer model, increasing precision from 82.80% to 98.18%. Using the complete 6-channel NAM-DeepLabV3+ model, shadow restoration increased mean intersection over union (mIoU) from 76.84% to 78.90% and mean pixel accuracy (mPA) from 90.31% to 92.15% (p = 0.004). These results confirm that restoration improves classification by reducing shadow-induced spectral inconsistency.
In recent years, intensified heat waves and prolonged droughts due to climate change have accelerated the ignition and spread of forest fires. Therefore, developing predictive models and model-based approaches to reduce the probability of fires and mitigate potential damage is critical. This study employs the MaxEnt algorithm to model current and future (2020–2100) forest fire probabilities across Türkiye under the optimistic (SSP1-2.6) and pessimistic (SSP5-8.5) scenarios. By analyzing historical fire records alongside climatic, topographic, and anthropogenic variables, the baseline model demonstrated robust predictive performance, achieving an area under the receiver operating characteristic curve (AUC) of 0.82, a continuous Boyce index (CBI) of 0.85, and a true skill statistic (TSS) of 0.78. The results revealed that fire regimes in Türkiye are primarily governed by moisture and temperature dynamics; specifically, precipitation in the coldest quarter (bio19), annual mean temperature (bio1), and the topographic wetness index (TWI) emerged as the most critical determinants of fire probability. Future projections indicate considerable temporal and spatial transformations in fire probability by the end of the century. Under the optimistic scenario, high-risk areas show a progressive increase while low-risk zones rapidly contract. In the pessimistic scenario, a sharp decline in low-risk areas (from 44.15 to 25.18 Mha) is observed until 2080. Notably, the partial increase in low-risk classes observed in the final quarter of the century is interpreted not as an improvement, but as a reflection of ecological tipping points, such as the depletion of fuel or widespread vegetation loss due to extreme droughts. One of the most significant results is the projected northward shift in the weighted mean center of areas classified as high and extreme forest fire probability. The shift reached approximately 81 km under SSP1-2.6 and 152 km under SSP5-8.5 by the end of the century. This situation indicates that fire-prone environments are gradually expanding into humid forest ecosystems in northern Türkiye, which are currently classified as low-risk. Ultimately, by identifying regions where future vulnerability will intensify, this study provides a strategic foundation for planning effective short- and long-term fire prevention, management, and suppression strategies.
Forest conversion toward climate-resilient non-native tree species is increasingly promoted in temperate Europe, yet biodiversity responses outside the main growing season remain poorly studied. In particular, winter conditions may constrain the biodiversity of forest organisms through reduced resource availability and harsh microclimate conditions, potentially amplifying effects of tree species composition on habitat use and community composition.Here, we investigated how winter bird communities respond to forest stand composition by comparing pure and mixed stands of European beech, Norway spruce, and non-native Douglas fir across 38 forest plots in northwest Germany. Using passive acoustic monitoring, we quantified bird taxonomic and functional diversity from 4476 h of recordings comprising 1,816,452 total detections.Beech-Douglas fir mixtures generally supported lower activity days, taxonomic, and functional diversity than native monocultures. Bird community composition differed consistently among stand types, with pronounced contrasts between beech- and conifer-dominated stands. In addition to stand composition, winter bird activity and diversity were strongly modulated by microclimatic conditions and vertical forest structure.Our results, compared to previous studies conducted during the growing season, clearly show that the effects of non-native Douglas fir on biodiversity under extreme environmental conditions, i.e., in winter, cannot be deduced from findings in the summer season. Therefore, focusing solely on the main growing season carries the risk of incomplete conclusions about the consequences of forest conversion with non-native tree species for the activity and diversity of forest-associated organisms.
Forest soils play a critical role in regulating atmospheric greenhouse gas (GHG) budgets; however, coordinated multi-gas observations across broad climatic gradients remain limited, constraining our understanding of spatial variability and environmental controls on soil GHG fluxes. Here, we conducted a three-year field study (2019–2021) to quantify soil GHG fluxes across four representative forest ecosystems in China: a tropical montane rainforest (JFL), a subtropical Chinese fir plantation (HT), a temperate deciduous broadleaf forest (BTM), and a boreal larch forest (MH). Soil CO2 emissions were significantly lower at MH (114.24 mg·m−2·h−1) than at the other forests and highest at HT (229.13 mg·m−2 ·h−1). CH4 uptake was the strongest at BTM (60.13 μg·m−2 ·h−1), significantly exceeding the other sites (31.81–34.45 μg·m−2·h−1). N2O emissions were similar between JFL (31.72 μg·m−2·h−1) and HT (34.29 μg·m−2·h−1), and both were significantly higher than those at BTM (21.06 μg·m−2·h−1) and MH (16.54 μg·m−2·h−1). Distinct seasonal dynamics of soil GHG fluxes were observed among forest types, reflecting contrasting hydrothermal regimes. Soil temperature showed consistent correlations with CO2 and N2O fluxes across sites, whereas soil moisture correlations varied among gases and forest types. The estimated Q10 ranged from 1.40 (MH) to 3.63 (JFL) and showed no consistent pattern along the climatic gradient. CO2 and N2O fluxes were significantly and positively correlated across all sites (R2 = 0.25–0.74, p < 0.05), whereas CH4 fluxes showed positive relationships with CO2 and N2O at JFL, but negative or non-significant relationships at the other sites. Total global warming potential (GWP) ranged from 5.22 tCO2−eq·ha−1 at MH to 20.89 tCO2−eq·ha−1 at HT, with CO2 accounting for more than 95% of the total GWP. At all sites, the warming effect of N2O consistently outweighed the cooling effect of CH4 uptake. Overall, our results highlight the importance of accounting for spatial heterogeneity and environmental conditions when assessing forest soil GHG budgets across climatic gradients.
Increasing frequency and intensity of heat and drought events under anthropogenic climate change has triggered widespread forest mortality and canopy dieback worldwide. Previous studies have suggested that such dieback events are often preceded by a marked decline in forest resilience, yet the underlying mechanisms driving this pre-dieback decline remain unclear. Here, we combined multi-scale satellite observations of vegetation dynamics and canopy functional traits to trace resilience trajectories preceding the 2022/2023 extreme heatwave-drought dieback event in northern China’s temperate forests. We found that forests experiencing canopy dieback exhibited a pronounced increase in the standard deviation (SD) of vegetation indices starting approximately a decade prior to dieback, indicating amplified ecosystem variability. Meanwhile, the contrasting trajectories of SD and lag-1 autocorrelation (AR1) revealed a trade-off between ecosystem stability and recovery dynamics, with forests becoming less stable but maintaining faster recovery capacity before dieback. This trend of declining stability was closely associated with higher pre-dieback productivity promoted by favorable climatic conditions, such as warmer and wetter periods. This suggests that while favorable climates temporarily boost productivity, they also amplify ecosystem sensitivity to subsequent climatic extremes. Our findings reveal that apparent greening under a warming climate does not necessarily indicate recovery, but may instead serve as an early warning signal of reduced forest stability and elevated dieback risk. These insights underscore the importance of integrating high-resolution satellite observations with long-term field monitoring to disentangle the mechanisms linking enhanced productivity to declining stability and to improve predictions of forest vulnerability under accelerating climate change.
Biodiversity models often rely on opportunistic species occurrence data. However, biases due to different sampling efforts or accessibility can confound biodiversity patterns and obscure ecologically pertinent patterns. Such problems can be specifically challenging when aiming at identifying emerging ecological sensitivity in forested and peri-urban landscapes for spatial planning purposes. Here, we present an eco-informatics framework that decouples sampling bias from environmental suitability to derive planning-ready ecological sensitivity surfaces. First, accessibility-related observation patterns are modeled as a null model (NULL) with the maximum entropy (MaxEnt) algorithm. The NULL model is then removed from the full model (FULL) to produce a bias-corrected suitability signal (FULL-NULL). Second, the residual signal is aggregated to account for geometrical effects and ensure consistency in spatial adjacency through a hexagonal grid. Third, categorical boosting (CatBoost) is used to capture nonlinear associations between rarity-based ecological sensitivity patterns and environmental predictors. Fourth, robustness is evaluated via spatial block-based hold-out validation and local spatial autocorrelation analysis, and model results are interpreted using SHapley Additive exPlanations (SHAP). The resulting surfaces highlight transitional landscapes where ecological vulnerability is elevated but not yet fully expressed, supporting a spatial early-warning perspective. Rather than forecasting future ecological collapse, the proposed framework identifies areas exhibiting latent ecological sensitivity and potential vulnerability under current environmental conditions. Overall, the proposed framework provides a practical and transferable approach to convert biased species occurrence data into meaningful spatially-consistent information that can help to inform regional forest management, conservation planning, and decision-making.
Root exudation is tightly coupled to soil organic carbon (SOC) formation and stabilization in forest ecosystems, yet how tree species and stand density jointly regulate root exudate carbon inputs and their associations with SOC pools across soil depths remains unclear. Here, we investigated pure stands of black locust (Robinia pseudoacacia) and Chinese pine (Pinus tabuliformis) established at five stand-density levels on the Loess Plateau. We quantified root exudation rates (RR, per unit root biomass), root exudate fluxes (RF, integrated over 0–100 cm), and SOC fractions to assess how tree species and stand density affect exudate carbon inputs and SOC pools. Black locust had higher RR in the 0–30 cm topsoil, whereas Chinese pine had greater 0–100 cm RF across density gradients than black locust (6.48–12.24 vs. 4.76–10.58 g C·m−3·year−1). Increasing stand density reduced RR, RF, and SOC mainly in the topsoil, indicating that density effects on belowground carbon input were concentrated near the surface. Associations between RR and SOC pools were depth- and species-specific: In black locust, RR was positively associated with SOC in the 0–30 and 30–60 cm layers but negatively associated with SOC in the 60–100 cm layer, whereas no comparable pattern occurred in Chinese pine. Overall, root exudate carbon inputs explained SOC variation primarily in the 0–30 cm layer. These findings highlight the importance of species- and depth-specific root exudation processes for predicting SOC dynamics and optimizing stand density in plantation forests.
Endogenous post-stratification (EPS) is a statistical method used to approximate traditional post-stratification when population-level counts are unknown. It involves assigning sample units to strata based on predicted values from a model trained on the same sample data. The population is then stratified based on fixed stratum boundaries. We evaluated the EPS method for estimating growing stock volume (GSV) for a management-level forest inventory. In such an inventory, a forest district is divided into age classes, i.e., first-stage strata, by means of auxiliary information on the age of forest stands. Each of these age classes is then divided into three to six second-stage strata using an internal model developed using data on GSV in terrestrial sample plots and with airborne laser scanning (ALS) metrics. The first of our research questions concerned whether traditional post-stratification estimators are relevant for estimating the mean and its variance in the EPS method. The second question was whether the EPS method is more effective than the reference (EXP_REF) method, in which age classes are not divided into second-stage strata. We conducted our research using a Simulated Forest District. This dataset contained simulated GSV values for individual grid cells, which were treated as “terrestrial truth”, as well as a set of original ALS metrics. We used these data to perform 1000 replicate sample simulations, each based on 500 sample plots. The estimated mean GSV for both the entire Forest District and the 11 individual age classes did not differ significantly from the true values. The frequency with which the true mean value fell within the confidence interval for the population mean GSV was aligned with the theoretical frequency for all inventory units. Our results show that the use of the EPS method would allow us to reduce the number of terrestrial sample plots by more than half compared with the method in which age classes are not divided into second-stage strata. Further research could focus on the use of variance estimators other than the simple expansion estimator when systematically located sample plots are used in the EPS method.
