Accurate prediction of tree growth is crucial for sustainable forest management, yet traditional field-based methods for measuring tree dimensions and competitive interactions are labor-intensive and time-consuming, limiting the spatial and temporal scope of forest monitoring. This study evaluated whether unmanned aerial vehicle laser scanning (ULS) could effectively replace traditional field-based methods for predicting tree diameter increment (DBHI) in Chinese fir (Cunninghamia lanceolata) plantations. We established 15 permanent plots in Jiangle County, Fujian Province in 2013 and conducted both field surveys and ULS scanning in 2023 to calculate diameter increments. Using an interpretable machine learning approach combining feature screening and SHAP analysis, we developed random forest models with Bayesian optimization using three variable sets: field-only, ULS-only, and combined data. ULS-derived variables achieved prediction accuracy statistically equivalent to traditional field methods (R2 = 0.572 vs 0.542; ΔR2 = 0.029, 95

Accurately Crown asymmetry is a common morphological response to heterogeneous environmental pressures, and this asymmetry directly influences the physiological processes of tree growth. Advances in UAV-Lidar technology have facilitated the acquisition of high-precision crown data, providing new pathways to characterize crown structural variations and investigate their impacts on tree growth. In this study, we propose a novel crown asymmetry index (Ihctsa), based on UAV-LiDAR data and time-series tool, to characterize crown asymmetry. Using a Cunninghamia lanceolata plantation in southeastern China as a case study, we analyzed the size-dependent effects of crown asymmetry on tree growth. The results revealed a significant reversal in the effect of crown asymmetry as tree size increased. Specifically, crown asymmetry exhibited a significant negative effect on the growth of small DBH trees, whereas it showed a significant positive effect on the growth of large DBH trees. Compared to the traditional Mean Circularity Index (IMC), the Ihctsa index captures the structural characteristics of crown variations and can determine the DBH threshold value (≈ 19.77 cm) for growth response at which transition from negative to positive. Furthermore, the LiDAR-based crown overlap index (CI2) demonstrated superior methodological suitability for separating thinning effects compared to the classic Hegyi index (CI1). This study demonstrates the feasibility and effectiveness of utilizing LiDAR for spatial competition analysis, providing a new methodological reference for forest ecology research.

Tree height is a key indicator of forest growth, yet identifying its drivers remains challenging owing to complex interactions and limited data. We analyzed longitudinal data (2010–2021) from ten permanent 20 × 20 m plots at the Jiangle State-Owned Forest Farm. To address the limitations of small sample sizes, this study proposes a novel integrated framework combining Bayesian hyperparameter-optimized Random Forest (RF) and Bootstrap Bayesian Networks (BN). Based on 23 variables retained after multicollinearity screening, the RF model identified 21 potential predictors, such as diameter at breast height (Dg), stand basal area (BA), and stand density (N). A key methodological contribution was the implementation of the Bootstrap BN with permutation tests, which significantly enhanced model robustness and edge reliability. The results revealed significant interactions: Dg and MAP exerted direct effects on tree height, whereas SDI influenced tree height indirectly through mediation effects. Notably, the proposed Bootstrap BN model achieved an 84

The Longmen Shan orogenic belt, at the eastern margin of the Tibetan Plateau, is a prominent tectonic deformation zone formed by eastward plateau extrusion against the Sichuan Basin. Marked north-south contrasts exist in landscape morphology, deformation mechanisms, and erosion history, yet their geodynamic origins remain debated. Using numerical models constrained by thermochronological data and geomorphic observations, we evaluated how surface erosion parameters, tectonic uplift velocity, and lithological contrasts shape fluvial landscapes and longitudinal profiles. Without tectonic uplift, differences between the Min River (southern segment) and Fu River (northern segment) profiles can be explained by erosion duration. Starting from an initial topography resembling the modern upper Min River, the landscape requires 5-20 Myr of additional erosion to evolve into one similar to the present-day upper Fu River. This process involves low erosion on the plateau and basin sides but substantial erosion in the intervening orogenic belt, matching spatial patterns in thermochronological ages. The results with tectonic uplift indicate that preserving steep topography in the southern segment requires higher vertical uplift velocity, while lower velocities maintain the subdued northern relief, consistent with present-day uplift patterns. Spatial variations in erosion parameters linked to lithological heterogeneity also influence knickpoint formation and profile segmentation. These results indicate that the geomorphic contrast between the southern and northern Longmen Shan arises from the combined effects of heterogeneous uplift and spatially variable surface processes. Sustaining steep topography since the Miocene requires high uplift velocity in the Minshan region and southern segment of the Longmen Shan, corresponding to the Huya Fault and southern Longmen Shan Fault Zone, likely marking the eastern boundary of the Songpan-Ganzi block.

Tree height (H) and height to crown base (HCB) are fundamental variables in individual-tree growth modeling, yet they are commonly modeled independently and estimated solely within the frequentist paradigm, which disregards their biological coupling. Joint models that incorporate stand structural complexity remain largely absent for Chinese fir (Cunninghamia lanceolata (Lamb.) Hook.) plantations. Here, we developed an integrated framework that couples SHapley Additive exPlanations (SHAP)-guided variable selection with a hierarchical Bayesian nonlinear seemingly unrelated mixed-effects model (HB-NSURMEM), using 2827 trees from 25 even-aged plots. We screened 45 candidate covariates by random-forest SHAP analysis, embedded the selected covariates within a bivariate Logistic system (standard form for H, inflection-point form for HCB), and estimated the system under both frequentist (NSURMEM) and hierarchical Bayesian frameworks, with frequentist estimates serving as informative priors. The plot-level Gini coefficient had the highest SHAP value for both responses but was not significant in the hierarchical models, its ranking reflecting a between-plot screening signal rather than an independent fixed effect. The covariate effects that persisted across both frameworks were confined to the height equation, in which a distance-dependent competition index (CI18) lowered the asymptotic height and a distance-independent crowding index (CI7) delayed its attainment; crown-base position, by contrast, was governed predominantly by between-plot variation rather than by any individual-tree covariate. The HB-NSURMEM and frequentist NSURMEM showed similar in-sample accuracy. Under leave-one-plot-out validation, the Bayesian model showed higher population-level accuracy without local calibration (R2=0.49 vs. 0.42 for H; 0.34 vs. 0.29 for HCB), whereas the two frameworks converged after local calibration. Beyond point prediction, the Bayesian formulation recovered the full cross-equation dependence that the block-diagonal frequentist structure cannot express—a strong residual correlation (ρε = 0.818) and plot-level correlations between the two asymptotes (ρ = 0.74) and between the H location parameter and crown-base inflection (ρ = 0.68). Tree-level SHAP screening can identify informative between-plot contrasts, but hierarchical modeling is required to distinguish them from independent covariate effects, and joint Bayesian estimation provides coherent uncertainty quantification and a transferable template for joint allometric modeling in other plantation systems.

Fractional Vegetation Cover (FVC) is an important indicator for evaluating the quality of regional ecological environments and sustainable development. As a core region of China's socio-economic development, the Beijing-Tianjin-Hebei area experiences vegetation dynamics that are significantly influenced by both intensive human activities and climate change. This study aims to reveal the spatiotemporal patterns of vegetation coverage in the Beijing-Tianjin-Hebei region from 2000 to 2023, and to identify the major climatic driving factors, with the goal of providing scientific support for regional ecological conservation and coordinated development policymaking. On this basis, a combination of methods including the Theil-Sen (Sen’s) slope estimator, Mann-Kendall (MK) test, Hurst exponent analysis, and partial correlation analysis was comprehensively applied. An in-depth analysis was conducted on the spatiotemporal dynamics of FVC, its future persistence, and its relationships with temperature and precipitation. The results indicate that: (1) Spatially, the FVC in the Beijing-Tianjin-Hebei region exhibits a distinct "northwest-high, southeast-low" distribution pattern. Temporally, the region experienced three phases: rapid increase (2000–2010), significant decline (2010–2015), and restorative growth (2015–2023).(2) Vegetation improvement areas (47.90%) and degradation areas (41.68%) show strong spatial heterogeneity, and up to 59.83% of the region exhibits anti-persistent characteristics in vegetation dynamics.(3) Precipitation serves as a positive driving factor for improved FVC, while the influence of temperature exhibits spatial heterogeneity. The results of this study highlight the need for future regional ecological management to shift from solely pursuing increases in vegetation coverage toward enhancing ecosystem quality and resilience, and to develop adaptive strategies to address potential ecological risks.

The arcuate tectonic belt in the northeast Tibetan Plateau has been a contentious topic regarding its formation and evolution, owing to its distinctive geological structure as the lateral growth boundary of the plateau. In this research, leveraging geological and geophysical data, a three-dimensional finite element numerical model is employed to explore the impact of lateral and vertical inhomogeneities in lithospheric strength on the northeast Tibetan Plateau’s growth and the arcuate tectonic belt’s formation and alteration. Additionally, the kinematic and deformation traits of the arcuate tectonic belt, such as regional motion velocity, stress, and crustal thickness during shortening and strike-slip deformation, are comparatively analyzed. The findings indicate that the arcuate tectonic belt takes shape when the weakly strengthened Tibetan Plateau is impelled into the Yinchuan Basin after being obstructed by the robust Alax and Ordos blocks during lateral expansion. Intense shear deformation occurs at the block boundaries during the arc tectonic belt’s formation. The weak middle-lower crust, serving as a detachment layer, facilitates the plateau’s lateral growth and crustal shortening and thickening without perturbing the overall deformation characteristics. It is verified that the arcuate tectonic belt was formed during the NE-SW compression phase from around 9.5 to 2.5 Ma, accompanied by significant crustal shortening and thickening. Since 2.5 Ma, within the ENE-WSW compression process, the internal faults of the arcuate tectonic belt are predominantly strike-slip, with no pronounced crustal shortening and thickening. Only local topographical modification is conspicuous. This study will enhance our comprehension of the Tibetan Plateau’s uplift and lateral growth process and furnish a foundation for investigating the formation of arcuate tectonic belts.

This article has been retracted: please see Elsevier policy on article withdrawal (https://www.elsevier.com/about/policies-and-standards/article-withdrawal).This article has been retracted at the request of the Editor-In-Chief.Post-publication, an investigation conducted on behalf of the journal by Elsevier’s Research Integrity & Publishing Ethics team discovered suspicious changes in authorship between the original submission and the revised version of this paper. During revision, the authors Jamshid Ali, Mohammad Javed Ansari and Saleh H. Salmen were added to the revised paper without explanation and without exceptional approval by the journal editor, which is contrary to the journal policy on changes to authorship. The editor reached out to the authors for an explanation, but they failed to provide a satisfactory explanation to these changes.The investigation also found a significant increase of citations to papers published by the author, Shoaib Ahmad Anees, between the original submission and the revised version of this article. In summary, one paper by the author was cited in the original version of the article. This increased to twenty papers in the revised version of the article.The Editor has determined that the authorship and the findings of the article cannot be relied upon, and has decided to retract the article.The authors disagree with the retraction and dispute the grounds for it.

The Tibetan Plateau's lower crustal composition and rheology dictate the tectonic and surface deformation in response to the India-Asia collision. Yet, their spatial variations under the Plateau remain unclear. In this study, we forward calculated seismic velocities (i.e. Vp and Vs) for three types of rocks presumed to constitute the continental lower crust. By comparing the simulated and observed velocities, we infer the current lower crustal composition of the Tibetan Plateau. The results show that W Tibet and Pamir is predominantly composed of eclogite, suggesting ongoing intense metamorphism. Our simulations reveal the presence of a weak felsic granulite layer in SE Tibet, potentially indicative of lower crustal flow. In contrast, the felsic component identified in eastern Lhasa is likely attributed to magmatic underplating. Our results further suggest that the lower crust of W Qiangtang and W Lhasa primarily consists of mafic granulite, indicating the likely preservation of Asian proto-rocks. Furthermore, by linking lower crustal composition to rheology, this study demonstrates how the crust has predominantly deformed in response to the India-Asia collision across the Tibetan Plateau.

The RSEI index has been used widely in ecological environment quality assessment， but its application in high vegetation coverage areas in southern China remains relatively limited. To monitor and evaluate the ecological environment quality in the region more effectively， indices of humidity （WET）， heat （LST）， and dryness （NDBSI） factors and the comprehensive vegetation index mNDVI were introduced to construct the modified remote sensing ecological index （MRSEI） by principal component analysis. Then， the GEE cloud platform and ArcGIS 10.8 platform were used to analyze the spatial and temporal distribution and driving mechanisms of ecological quality in Jiangle County from 2000 to 2020. The results were as follows： ① Compared with RSEI， the average correlation between MRSEI and the principal components was higher. In the three experimental areas， its contrast was increased by 10.538， 2.923， and 8.558， and its entropy was increased by 0.024， 0.046， and 0.025， respectively. ② The overall change of MRSEI in Jiangle County was an increase， with an average annual increment of 0.010. Areas with good ecological quality accounted for the largest proportion of the county， ranging from 33.61% to 38.28%， while the proportion of county areas classified as poor or poorer was about 10%. The proportion of areas defined as stable in ecological change ranged from 42.23% to 59.05%， although this fluctuated significantly from year to year. Moreover， the ecological environment was more prone to deterioration in the southwest and northern regions， and urban construction expanded outward from the central areas. Through the 20 years， the land area with improved ecological quality reached 424.34 km2， indicating a significant improvement in Jiangle County's ecological environment. ③ Land use type and slope were the primary factors influencing spatial variations in ecological environment quality， with annual average precipitation also playing a significant role. The interactions among driving factors led to some degree of improvement， with the interaction between land use type and annual average precipitation having the strongest influence on MRSEI spatial differentiation， contributing 36.1% to the variation. This study provides a scientific basis for ecological environment monitoring and sustainable development in Jiangle County.

Land use and land cover (LULC) classification is essential for environmental monitoring and sustainable land management. The selection of satellite sensors and classification algorithms influences the accuracy of LULC classification. This study evaluates the performance of three satellite sensors, GF-6 (GF-6), S2 (S2), and L9(L9), and three machine learning classifiers, Random Forest (RF), Support Vector Machine (SVM), and Extreme Gradient Boosting (XGBoost), in classifying LULC in Islamabad, Pakistan. The satellite data with high-to-course spatial resolution data was utilized, and a comprehensive pre-processing workflow ensured high-quality imagery. The results indicate that XGBoost, paired with GF-6, achieved the highest overall classification accuracy (94.24

Tree species mixing is a well-known forest management strategy, often playing a beneficial role in forest productivity, stability and ecosystem biodiversity. However, whether species mixing mitigates the impacts of climate anomalies such as drought on tree growth remains controversial. This study compared the growth, climate sensitivity and drought response between pure and mixed forests with different mixing proportions of Chinese fir in Southeastern China using the normalized difference vegetation index (NDVI) and tree-ring width (TRW) respectively. NDVI time series were extracted from 57 monoculture and 50 mixture plots using the Landsat remote sensing satellite spanning 1986–2022, and tree rings were collected from 324 trees in pure forests and 182 trees in mixed forests. Then, the effects of tree-level, neighborhood, and climate drivers were investigated using the interpretable XGBoost model. No evidence showed that Chinese fir benefited from the admixture with Masson pine. Species mixing had neutral or negative effects on the growth of Chinese fir and its drought response, regardless of whether NDVI or TRW was used. NDVI was more sensitive to atmospheric moisture compared to TRW, but differences between pure and mixed forests were more pronounced in TRW. Tree growth and its response to drought in mixed forests were influenced by tree characteristics > neighborhood traits > climate, and their effects varied with the mixing proportions of species. Vapor pressure deficit (VPD) had a greater impact on tree growth than soil moisture. In mixed forests, the negative impact of VPD on tree growth was the smallest in the mixing proportion of 5:5, which also experienced the lowest growth reduction caused by droughts. Our findings suggested that species mixtures did not always facilitate all mixed species, but rather even reduced the growth and resilience to multi-year droughts of a particular species. However, appropriate mixing proportions of species could minimize the downside effects while maintaining the ecological advantages of species mixing.

The rheological structure of the East Asia continent is the key to understanding its broad, heterogeneous, and intense Cenozoic deformation. Based on a refined three-dimensional thermal structure of the lithosphere in this region and the latest strain rate data, we derived a model of the rheological structure of the East Asian continental lithosphere. The strength envelopes, defined by the yield strength of frictional, fractural, and plastic creep, are constrained by the lithological stratification based on previous studies and the depth distribution of earthquakes. The results show large vertical and lateral variations of lithospheric strength in the East Asian continent. A weak lower crust with low effective viscosity is ubiquitous. The rheological structure agrees with the jelly sandwich model in cratons, where the mantle lithosphere is relatively strong. The Tibetan Plateau has the weakest lower crust, with its effective viscosity ranging from 1019 to 1020 Pa center dot s. Its mantle lithosphere is weakened by relatively high temperature; hence, its rheological structure can be described by the cre`me br & ucirc;lee model. The lithospheric scale faults and suture zones in and around the Tibetan Plateau, with low strength or viscosity, correspond to the banana split model. The strength of the lithosphere in the Tibetan Plateau and other zones of active Cenozoic tectonics mainly derive from the crust, while the strength of the cratonic lithosphere is dominated by that of the mantle lithosphere. The rheological heterogeneity controls the lateral growth of the Tibetan Plateau and the widespread and differential deformation in the East Asian continent.

Objective The MS 6.8 earthquake that struck Luding County, Sichuan Province, on September 5, 2022, and its aftershocks have drawn widespread attention, especially concerning the potential risks of induced seismicity associated with the construction of high-dam reservoirs in regions with high seismic intensity. Previous studies have explored the possible link between reservoirs and seismic activity, without reaching a definitive conclusion. This study aims to assess the impact of water storage in the Dagangshan Reservoir on the surrounding strata and its correlation with recent seismic events. Methods Numerical simulation methods were employed using high-precision digital elevation model (DEM) data, fault data, and reservoir water level information to develop a three-dimensional poroelastic finite element numerical model extending from the surface to a depth of 25 km. By analyzing the hydrogeological conditions, lithology of rock masses, and groundwater dynamic changes, this study evaluated the seismic hazard risk of major faults, such as the Moxi Fault, and calculated variations in Coulomb stress and strata pore pressure at the hypocenter during the occurrence of the earthquake. Results The study indicates that, during the MS 6.8 Luding earthquake on September 5, 2022, the pore pressure at the epicenter reached 5 kPa, and the Coulomb stress increased by 3.6 kPa, suggesting that the impoundment of the Dagangshan Reservoir contributed to an increased risk along the Moxi Fault on the northwestern side of the reservoir. Using the source parameters of the MS 5.6 aftershock that occurred on January 26, 2023, it was observed that the impoundment caused a change in Coulomb stress at the epicenter of -0.69 kPa and a pore pressure of approximately 0.32 kPa. It is evident that the reservoir impoundment had a relatively minor impact on the fault activity where the MS 5.6 aftershock occurred and even exhibited a certain inhibitory effect. Moreover, seismic activity was mainly concentrated in two areas on the western side of the reservoir, with both the seismicity and expected magnitudes in these regions reflecting a higher risk of earthquakes. Conclusion This study demonstrates a correlation between the impoundment activities of the Dagangshan Reservoir and the occurrence of the Luding County earthquake and its aftershocks. The spatial distribution characteristics of the earthquakes align with the geological stress adjustment patterns following the reservoir impoundment, which played a promotive role in the occurrence of the MS 6.8 main shock, leading to an increased risk of earthquakes in the Moxi Fault region. This finding is significant for understanding the mechanisms of reservoir-induced earthquakes, subsequent aftershock analyses, and earthquake disaster prevention and mitigation efforts. Significance The results of this study provide new insights into the complex relationship between reservoir water storage and seismic events. This study offers a scientific basis for future assessments of seismic risks of reservoir design and operation, contributing to improved accuracy in earthquake early warnings and efficiency in disaster prevention and mitigation efforts.

The increasing pressures of urban development and agricultural expansion have significant implications for land use and land cover (LULC) dynamics, particularly in ecologically sensitive regions like the Murree and Kotli Sattian tehsils of the Rawalpindi district in Pakistan. This study's primary objective is to assess spatial variations within each LULC category over three decades (1992-2023) using cross-tabulation in ArcGIS to identify changes in LULC and investigates into forest fragmentation analysis using the Landscape Fragmentation Tool (LFTv2.0) to classify forest into several classes such as patch, edge, perforated, small core, medium core, and large core. Utilizing remote sensing data from Landsat 5 and Landsat 9 satellites, the research focuses on the temporal dynamics in various land classes including Coniferous Forest (CF), Evergreen Forest (EF), Arable Land (AR), Buildup Area (BU), Barren Land (BA), Water (WA), and Grassland (GL). The Support Vector Machine (SVM) classifier and ArcGIS software were employed for image processing and classification, ensuring accuracy in categorizing different land types. Our results indicate a notable reduction in forested areas, with Coniferous Forest (CF) decreasing from 363.9 km2, constituting 45.0% of the area in 1992, to 291.5 km2 (36.0%) in 2023, representing a total decrease of 72.4 km2. Similarly, Evergreen Forests have also seen a significant reduction, from 177.9 km2 (22.0%) in 1992 to 99.8 km2 (12.3%) in 2023, a decrease of 78.1 km2. The study investigates into forest fragmentation analysis using the Landscape Fragmentation Tool (LFTv2.0), revealing an increase in fragmentation and a decrease in large core forests from 20.3% of the total area in 1992 to 7.2% in 2023. Additionally, the patch forest area increased from 2.4% in 1992 to 5.9% in 2023, indicating significant fragmentation. Transition matrices and a Sankey diagram illustrate the transitions between different LULC classes, providing a comprehensive view of the dynamics of land-use changes and their implications for ecosystem services. These findings highlight the critical need for robust conservation strategies and effective land management practices. The study contributes to the understanding of LULC dynamics and forest fragmentation in the Himalayan region of Pakistan, offering insights essential for future land management and policymaking in the face of rapid environmental changes.

Staphylococcus aureus ( S. aureus ) infection is major cause of nosocomial infections. Antibiotic treatment for S. aureus remains the primary solution for managing S. aureus infections, which, however, increases the risk of antibiotic resistance. To broaden the resolutions on S. aureus infection, here we report TurboID-mediated protein proximity technologies to inhibit the growth of S. aureus . To achieve this goal, we utilized synthetic biology techniques to create a fusion protein named N-AgrD-TurboID (Agr-ID). The N-AgrD domain includes auto-inducer peptide (AIP) which combined to the surface AgrC protein on S. aureus . As such, TurboID then catalyzed the production of biotinoyl-5’-AMP anhydride, triggering the biotinylation of surface proteins on S. aureus 25923 which were visualized by using fluorescence microscopy after incubating with Alexa Fluor 647-conjugated streptavidin. The biotinylation of surface protein on S. aureus 25923, S. aureus 43300, and S. aureus 6538 (MRSA) also resulted in growth inhibition and impaired colonization. Moreover, the biotinylation on surface protein further inhibited virulence protein production in S. aureus 25923, as indicated by reduced apoptosis of HEK 293T cells after treatment with S. aureus 25923 lysates. Overall, our work reveals that the biotinylation of surface proteins can inhibit the growth and toxicity of S. aureus 25923, S. aureus 43300, and S. aureus 6538 (MRSA), indicating therapeutic potential in clinical treatment. ### Competing Interest Statement The authors have declared no competing interest.

This study employs a composite lithospheric flexure isostasy model, using Bouguer gravity anomalies and topographic data, to perform a joint inversion of gravity admittance and coherence functions. Using Bayesian statistical inference, the optimal lithospheric flexure parameter model was determined, allowing for the characterization of the spatial distribution and uncertainty estimation of the effective elastic thickness ( Te) and load ratio ( F) along the northeastern margin of the Tibetan Plateau. Building on this foundation, the paper analyzes the variability among different tectonic units and their relationship with seismic activity, incorporating data on thermal-rheological structure to propose potential deep drivers of tectonic movements. The results indicate that the region exhibits a widespread distribution of low F values, suggesting that the current lithospheric flexural state is primarily influenced by surface load. Furthermore, significant variability in effective elastic thickness ( Te) reflects differences in mechanical strength, which correlate with the eastward extrusion deformation and thermal structural state of the Tibetan Plateau. The findings provide a geophysical basis for the quantitative study of the dynamics of the Tibetan Plateau.

Quantifying forest stand parameters is crucial in forestry research and environmental monitoring because it provides important factors for analyzing forest structure and comprehending forest resources. And the estimation of crown density and volume has always been a prominent topic in forestry remote sensing. Based on GF-2 remote sensing data, sample plot survey data and forest resource survey data, this study used the Chinese fir (Cunninghamia lanceolata (Lamb.) Hook.) and Pinus massoniana Lamb. as research objects to tackle the key challenges in the use of remote sensing technology. The Boruta feature selection technique, together with multiple stepwise and Cubist regression models, was used to estimate crown density and volume in portions of the research area's stands, introducing novel technological methods for estimating stand parameters. The results show that: (i) the Boruta algorithm is effective at selecting the feature set with the strongest correlation with the dependent variable, which solves the problem of data and the loss of original feature data after dimensionality reduction; (ii) using the Cubist method to build the model yields better results than using multiple stepwise regression. The Cubist regression model's coefficient of determination (R-2) is all more than 0.67 in the Chinese fir plots and 0.63 in the P. massoniana plots. As a result, combining the two methods can increase the estimation accuracy of stand parameters, providing a theoretical foundation and technical support for future studies.

Knowledge of how different drivers affect tree responses to drought is unprecedentedly imperative in the context of increasing frequency and severity of climatic droughts. Here, to fully understand the drought response complexity of trees, we assessed drought resilience (resistance and recovery) for Chinese fir (Cunninghamia lanceolata) in Southeast China based on tree ring from 324 trees, and used mixed effects model and machine learning (ML) to examine the roles of tree size, predrought growth performances, multiple drought dimensions, and microtopography in affecting tree drought responses. ML were interpreted using a novel of SHapley Additive exPlanations (SHAP) method. Tree responses to drought were primarily driven by tree characteristics (tree size and predrought growth), rather than drought dimensions (intensity, duration and occurrence Timing) and microtopography (elevation and slope aspect). Resistance and resilience increased with tree size and pre-drought growth variability but decreased with drought intensity- quantified by negative climate water balance. Recovery increased with predrought growth rates but decreased with drought duration. The drought intensity threshold for trees fully recovery of tree growth was about -80 mm. Higher elevations and shady slopes favored resistance (resilience) and recovery respectively, which combined with a greater impact of drought in the dry season suggested that the trees suffered more from droughts that only occurred in the dry season, especially at low- and medium-elevation sunny slopes. This study provided a comprehensive insight into tree growth response to drought, and contributed to the understanding of the mechanisms underlying the complexity of drought response. Increasing size diversity in Chinese fir plantations at sunny lower-elevation slopes is a promising measure to cope with the negative effects of drought.