Ecological Informatics最新文献

{"title":"Satellite embeddings as complementary predictors of standing dead-tree volume fraction in boreal forests","authors":"Anwarul Islam Chowdhury,&nbsp;Mete Ahishali,&nbsp;Mikko Vastaranta,&nbsp;Md. Jamal Uddin,&nbsp;Samuli Junttila","doi":"10.1016/j.ecoinf.2026.104013","DOIUrl":"10.1016/j.ecoinf.2026.104013","url":null,"abstract":"<div><div>Standing dead-tree volume fraction is an important indicator of forest condition and biodiversity, but estimating it beyond field plots remains challenging. Remote sensing can support plot-level prediction, yet it remains unclear whether satellite embeddings (SEs) provide an alternative information source when aerial multispectral imagery (MSI) is unavailable or add information when combined with canopy height model (CHM) and MSI predictors. SEs are precomputed AlphaEarth representations that summarize multi-source satellite observations, including Sentinel-1, Sentinel-2, and Landsat, into annual 10 m layers, reducing preprocessing burden while encoding spectral, spatial, and temporal information. Using 134 field plots from primary and near-natural boreal forests in Finland, we predicted standing dead-tree volume fraction using CHM, MSI, and SE predictors. Random Forest, Gradient Boosting, and XGBoost models were evaluated across seven predictor-set combinations with nested cross-validation. For XGBoost, CHM + SE nearly matched CHM + MSI performance within this dataset (R² = 0.660 vs. 0.657), suggesting SE predictors may provide canopy-condition information when MSI is unavailable. The highest accuracy was achieved with CHM + MSI + SE (R² = 0.708 ± 0.062, RMSE = 0.099 ± 0.017, MAE = 0.075 ± 0.010), indicating a modest benefit from combining predictor sources. Predictions were compressed toward intermediate values, with high fractions underestimated and low fractions slightly overestimated. SHapley Additive exPlanations (SHAP) analysis suggested that SE predictors interacted with canopy-structure and spectral variables, with temporal-change predictors among the most influential. Overall, SE predictors can support plot-level prediction where MSI is unavailable, but their contribution was complementary and context-dependent. Future work should test spatial transferability and longer embedding trajectories before wall-to-wall mortality mapping.</div></div>","PeriodicalId":51024,"journal":{"name":"Ecological Informatics","volume":"98 ","pages":"Article 104013"},"PeriodicalIF":8.5,"publicationDate":"2026-09-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"148855168","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":2,"RegionCategory":"环境科学与生态学","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"OA","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}

{"title":"Quantifying physiological arousal to rural visual configurations: A mixed-effects modelling framework for hypothesis generation","authors":"Jie Xu,&nbsp;Yumeng Qian,&nbsp;Zhiqiang Zhou,&nbsp;Kejia Zha,&nbsp;Junhua Zhang,&nbsp;Shiro Takeda,&nbsp;Donglin Li,&nbsp;Siyu Zhang,&nbsp;Wei Nie","doi":"10.1016/j.ecoinf.2026.104012","DOIUrl":"10.1016/j.ecoinf.2026.104012","url":null,"abstract":"<div><div>Rural revitalization is reshaping village landscapes, but prevailing visual-assessment approaches still rely heavily on greenness, land-use composition, ecological indicators, or subjective ratings. These measures support planning but provide limited evidence about which combinations of natural, built, and open elements are associated with short-term physiological arousal at the human viewing scale. We therefore developed a proof-of-concept framework that converts rural photographs into measurable visual profiles and evaluates short-term autonomic responses during controlled viewing. Four villages in eastern China were selected using five site criteria to capture varied planning and landscape conditions. Candidate scenes were first screened during field walking using wearable sensing and synchronized first-person video, after which 40 photographs were evaluated under controlled laboratory viewing. Visual characteristics were quantified through the proportions of visible scene elements, edge density, color entropy, and exploratory principal component analysis, from which contrasting scene configurations were described post hoc rather than specified in advance. Mixed-effects models examined changes in skin conductance and heart rate variability while accounting for repeated observations within participants, and the physiological results were subsequently interpreted together with the visual-feature profiles at the configuration level. Some visual configurations were associated with higher skin conductance responses, whereas changes in heart rate variability did not differ significantly among the image sets. The observed arousal patterns could not be reduced to greenness, edge density, or color entropy alone; instead, the combined organization of built elements, natural elements, openness, and structural detail provided a more informative basis for interpretation. The contribution lies in adapting field screening, computer vision, physiological sensing, and trial-level mixed-effects modelling as a proof-of-concept framework for hypothesis generation in rural visual assessment. The framework is not intended to infer affective valence, restorative mechanisms, or long-term health effects, but to identify physiologically salient rural visual configurations for future validation through virtual reality (VR), field-based experiments, or valence-sensitive measures.</div></div>","PeriodicalId":51024,"journal":{"name":"Ecological Informatics","volume":"98 ","pages":"Article 104012"},"PeriodicalIF":8.5,"publicationDate":"2026-09-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"148855167","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":2,"RegionCategory":"环境科学与生态学","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"OA","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}

{"title":"Graph Theory, Graph Neural Networks, and Reinforcement Learning in marine ecology: A systematic review of methodologies, theoretical rigour, and future directions","authors":"Siphendulwe Zaza,&nbsp;Marcellin Atemkeng,&nbsp;Taryn S. Murray,&nbsp;Nicole Oyetunji","doi":"10.1016/j.ecoinf.2026.103986","DOIUrl":"10.1016/j.ecoinf.2026.103986","url":null,"abstract":"<div><div>Marine ecosystems are characterised by complex, dynamic, and non-linear interactions that challenge traditional ecological modelling approaches. Recent advances in computational methods, particularly Graph Theory (GT), Graph Neural Networks (GNNs), and Reinforcement Learning (RL), offer promising tools for analysing, predicting, and managing these systems. This study presents the first comprehensive systematic review that jointly examines the application and integration potential of GT, GNNs, and RL in marine ecology. Following a PRISMA-guided methodology, 122 studies published between 2000 and 2026 were analysed to evaluate methodological trends, application domains, theoretical rigour, and practical relevance.</div><div>The results show that GT remains the most widely used and mature approach, particularly for analysing ecological connectivity, food webs, and conservation planning. In contrast, GNNs and RL are emerging methods, with GNNs primarily applied to prediction and classification tasks (e.g., sea surface temperature forecasting and species behaviour analysis), and RL focused on adaptive decision-making in dynamic and uncertain environments, such as fisheries management and autonomous monitoring systems. Despite their individual strengths, these methods are largely applied in isolation. Key challenges identified across the literature include data sparsity, limited interpretability, weak integration of ecological knowledge, and insufficient real-world validation, particularly for RL.</div><div>To address these gaps, this review introduces a formal scoring framework to assess theoretical rigour across six dimensions: ecological relevance, theoretical foundations, uncertainty quantification, scalability, stakeholder utility, and interpretability. Furthermore, the study highlights significant geographical disparities in research contributions and emphasises the need for greater inclusion of biodiversity-rich regions in the Global South. Finally, we propose the development of integrated GT–GNN–RL frameworks as a promising direction for future research, enabling the modelling of adaptive behaviours within networked ecological systems and supporting explainable, data-driven decision-making for sustainable marine conservation.</div></div>","PeriodicalId":51024,"journal":{"name":"Ecological Informatics","volume":"98 ","pages":"Article 103986"},"PeriodicalIF":8.5,"publicationDate":"2026-09-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"148855227","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":2,"RegionCategory":"环境科学与生态学","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"OA","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}

{"title":"Trajectories and mitigation potentials of agricultural non-CO2 greenhouse gas emissions in China's regional trade","authors":"Chuyao Weng,&nbsp;Yuping Bai,&nbsp;Jiayu Zheng,&nbsp;Xin Xuan,&nbsp;Fan Zhang,&nbsp;Keyi Lyu,&nbsp;Guofeng Wang","doi":"10.1016/j.ecoinf.2026.104007","DOIUrl":"10.1016/j.ecoinf.2026.104007","url":null,"abstract":"<div><div>Projections of the agricultural non-CO₂ greenhouse gas (NCGHG) transfers under diverse regional development pathways offers vital insights for achieving both climate goals and responsible consumption. We calculated the agricultural NCGHG emissions in 30 Chinese provinces from 2007 to 2020, and examines the spatio-temporal dynamics of interregional emission transfers using an environmentally extended multi-regional input-output model. Combining this model with the STIRPAT model, we simulated future embodied NCGHG emissions within regional trade under multiple scenarios. The results revealed an upward trend in net interregional transfers of NCGHG emissions from 87.4 Mt. CO₂eq in 2007 to 194.7 Mt. CO₂eq in 2020. These flows were mainly aggregated in East and South China. East China remained the dominant inflow region, reaching 125.13 Mt. CO₂eq in 2020, while Northwest and Northeast China became the largest net outflow regions. Scenario simulations show that future consumption-based agricultural NCGHG emissions were highest under an extensive development scenario and lowest under green development, with the mitigation gap between them reaching 143.1 Mt. CO₂eq by 2050 and East China evidencing the largest reduction potential. Central, and Southwest China also faced heightened emission pressures, highlighting China's regional disparities in future mitigation potential. Crucially, the results revealed an interregional carbon leakage mechanism: net import regions satisfy their agricultural demands through trade, thereby externalizing the environmental costs of NCGHG emissions to net export regions Our findings provides essential scientific evidence to supporting the need for region-specific mitigation strategies for facilitating China's realization of its carbon peaking and carbon neutrality goals.</div></div>","PeriodicalId":51024,"journal":{"name":"Ecological Informatics","volume":"98 ","pages":"Article 104007"},"PeriodicalIF":8.5,"publicationDate":"2026-09-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"148854602","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":2,"RegionCategory":"环境科学与生态学","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"OA","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}

{"title":"Deep learning prediction of wildfire burned-area extent and burn probability from ignition conditions using topography, fuel, and meteorology in South Korea","authors":"JuGyeong Choi,&nbsp;HeeMun Chae","doi":"10.1016/j.ecoinf.2026.104019","DOIUrl":"10.1016/j.ecoinf.2026.104019","url":null,"abstract":"<div><div>Committing suppression, evacuation, and post-fire recovery decisions the moment a wildfire is dispatched requires knowing, before any perimeter exists, how large it will grow and where it will most likely burn. We present a deep learning framework that forecasts, from geospatial and meteorological conditions available at ignition and without the detailed fuel-load and moisture maps that process-based models require, a calibrated georeferenced burn-probability field for the small- to intermediate-extent fires that dominate the record. A U-Net maps a twelve-channel input stack—ignition prior, topography, fuel, fire duration, and meteorology—over a fixed 4.5 km ignition-centred window (256 × 256 grid, 17.6 m per pixel) to a unit-integral spatial allocation, and a gradient-boosted regression predicts burned-area size from fire duration and aggregated meteorology; the two combine into one coherent field p(x) = E[A]·a(x). We trained and evaluated the framework on South Korean fires, taking ground truth from Sentinel-2 burned-area masks validated against official records (untransformed-area R² = 0.872, log-space R² = 0.752, median detected-to-official ratio 0.82) and predictors from SRTM terrain, satellite fuel proxies, and station and reanalysis meteorology. Over the operating envelope (102 fires; 87 cross-validation, 15 hold-out test), the field was well calibrated in aggregate (expected calibration error 0.025) yet overconfident in its sparse high-probability bins. Its advantage over six reference baselines rested on precision for the rare burned class: per-fire precision–recall AUC reached 0.429 against 0.321 for the strongest baseline (paired Wilcoxon <em>p</em> = 0.008), whereas ranking skill (per-fire ROC-AUC 0.950) exceeded the climatological 0.904 only modestly. Burned-area size was moderately predictable under leave-one-out validation (<em>r</em> = 0.475, under-dispersed), and the finest perimeter remained limited by inputs that cannot represent suppression and spotting. For the routine fires that dominate the record, ignition-time prediction from public data can deliver a calibrated burn-probability field and event-scale size forecast to support fire management and ecological risk assessment.</div></div>","PeriodicalId":51024,"journal":{"name":"Ecological Informatics","volume":"98 ","pages":"Article 104019"},"PeriodicalIF":8.5,"publicationDate":"2026-09-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"148855166","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":2,"RegionCategory":"环境科学与生态学","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"OA","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}

{"title":"Indicator species of multifunctional predatory solitary wasps (Hymenoptera: Pemphredonini) and their biogeographic regionalization in China","authors":"Nawaz Haider Bashir,&nbsp;Li Ma,&nbsp;Muhammad Naeem,&nbsp;Longfei Yao,&nbsp;Jinghong Li,&nbsp;Huanhuan Chen,&nbsp;Qiang Li","doi":"10.1016/j.ecoinf.2026.103999","DOIUrl":"10.1016/j.ecoinf.2026.103999","url":null,"abstract":"<div><div>Predatory wasps (Hymenoptera: Pemphredonini) are primarily predatory solitary wasps that provide important biological control functions through larval prey provisioning and also contribute secondarily to pollination through adult flower visitation. However, these ecological functions may be threatened by the effects of 21st-century climate change on insect biodiversity. The potential impacts of future climate change on predatory wasps in China remain poorly understood. This study therefore aimed to evaluate potential changes in the climatic suitability of regionally representative predatory wasps and to provide evidence for their monitoring and conservation in China. Rather than selecting focal species a priori for separate MaxEnt analyses, we applied a sequential ecological-informatics workflow that first delineated assemblage-based biogeographic regions, then identified statistically supported regional indicator species, and finally projected the climatic suitability of those focal taxa under current conditions and the ACCESS-CM2 SSP3–7.0 scenario for 2050 and 2070. Multi-species overlays were subsequently used to identify spatial concentrations of suitable climate for monitoring and field validation. Three assemblage-based biogeographic regions were identified, and canonical correspondence analysis (CCA) provided partial rather than definitive support for climatic differentiation among them. Among 103 species, 14 were identified as statistically supported indicator species across the three regions. Under the high-emission SSP3–7.0 scenario simulated with the ACCESS-CM2 global climate model, 9 of the 14 indicator species (approximately 64%) had a lower mean projected climatically suitable area across 2050 and 2070 than under current conditions. Multi-species overlays further identified areas where suitable climatic conditions were concentrated across several indicator species, providing spatial priorities for monitoring, field validation, and conservation planning. Monitoring regional indicator species may provide insights into broader assemblage-level changes within their associated biogeographic regions. The resulting regional indicators and suitability-overlap areas provide testable priorities for regionally stratified monitoring, although the indicator species should not be interpreted as substitutes for direct monitoring of all species within their associated regions.</div></div>","PeriodicalId":51024,"journal":{"name":"Ecological Informatics","volume":"98 ","pages":"Article 103999"},"PeriodicalIF":8.5,"publicationDate":"2026-09-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"148855165","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":2,"RegionCategory":"环境科学与生态学","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"OA","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}

{"title":"A deep-learning based approach to detect and classify animals flying near wind turbines using thermal surveillance cameras and open-source software","authors":"John Yarbrough,&nbsp;Isabelle Cunitz,&nbsp;Jessica Schipper,&nbsp;Sora Ryu,&nbsp;Michael Lawson,&nbsp;Bethany Straw,&nbsp;Cris Hein,&nbsp;Paul Cryan","doi":"10.1016/j.ecoinf.2026.103909","DOIUrl":"10.1016/j.ecoinf.2026.103909","url":null,"abstract":"<div><div>Rapid growth in wind energy creates a need for wildlife-monitoring tools that can process long-duration video efficiently while preserving ecologically meaningful information on animal activity near operating wind turbines. To address this need, we developed an open-source, proof-of-concept thermal-video workflow combining computer vision and convolutional neural networks (CNNs) to detect and classify moving biological objects near a single wind turbine, and to preserve object-level outputs for downstream analysis. Using held-out imagery from the same wind turbine and camera system used for model development, the final workflow distinguished biological targets from non-biological motion, including moving wind turbine blades and clouds, with 99% accuracy, and classified bats (90%), birds (83%), and insects (69%). This open-source thermal-video workflow allows the recording of object ID, frame number, timestamp, x/y pixel position, predicted class, and class probability for each biological detection, which enables post-processing of apparent biological object flight paths, time spent within the field of view, and relationships with environmental variables. The models presented and described in this work were trained and tested on imagery from a single wind turbine and camera system as a proof of concept and the quantitative performance metrics are specific to this setup, therefore broader deployment will likely benefit from further site-specific retraining or calibration, though the released models and code provide a practical starting point for transfer learning and initial evaluation at new sites. The code and trained models have been released under permissive licenses to support community evaluation, retraining and reuse.</div></div>","PeriodicalId":51024,"journal":{"name":"Ecological Informatics","volume":"98 ","pages":"Article 103909"},"PeriodicalIF":8.5,"publicationDate":"2026-09-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"148855169","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":2,"RegionCategory":"环境科学与生态学","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"OA","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}

{"title":"Integrating Swin Transformer and UPerNet for high-resolution forest-type mapping in support of ecosystem monitoring","authors":"Congfang Liu,&nbsp;Qing Ji,&nbsp;Fengman Fang,&nbsp;Zhiming Zhang,&nbsp;Wei Wang,&nbsp;Youru Yao,&nbsp;Yuesheng Lin,&nbsp;Qi Wang","doi":"10.1016/j.ecoinf.2026.104016","DOIUrl":"10.1016/j.ecoinf.2026.104016","url":null,"abstract":"<div><div>Accurate fine-scale forest-type classification in mountainous regions is critical for ecological monitoring, forest resource assessment, and sustainable forest management. However, complex terrain and heterogeneous vegetation structures pose significant challenges to the accurate delineation of forest-type boundaries using conventional classification methods. In this study, we proposed a deep semantic segmentation framework that integrates a Swin Transformer encoder with a UPerNet decoder to improve fine-scale classification of forest types in mountainous regions. Gaofen-1 (GF-1) imagery with high spatial resolution (2 m) from the Dabie Mountains in China was used to train the model to distinguish four forest types: broadleaf forest, coniferous forest, mixed conifer–broadleaf forest (hereafter “mixed forest”), and sparse forest. Under three-fold leave-one-region-out cross-validation, the Swin Transformer–UPerNet model achieved a mean Intersection over Union (mIoU) of 83.17% ± 0.48%, a mean class accuracy (mAcc) of 90.74% ± 0.31%, and an overall accuracy (OA) of 93.72% ± 0.87%. The model outperformed U-Net, Swin-UNet, DeepLabV3+, SegFormer, and HRNet in overall classification performance. Independent validation in a separate region of the Dabie Mountains yielded an OA of 83.50% (95% CI: 81.20–85.70%), and cross-regional validation based on the pooled samples from three validation subregions in the Mount Huangshan region yielded an OA of 83.33% (95% CI: 78.75–87.50%). These results indicate that the integration of Transformer-based feature extraction and multi-scale feature fusion provides an effective approach for fine-scale forest-type mapping in mountainous regions. The resulting forest-type information can support forest resource inventories, ecological monitoring, and sustainable forest management.</div></div>","PeriodicalId":51024,"journal":{"name":"Ecological Informatics","volume":"98 ","pages":"Article 104016"},"PeriodicalIF":8.5,"publicationDate":"2026-09-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"148854603","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":2,"RegionCategory":"环境科学与生态学","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"OA","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}

{"title":"SWARM-IS: Adaptive sensor swarm search strategy and simulation framework for predator surveillance and control","authors":"Hugh Parsons,&nbsp;Sandra Gómez-Gálvez,&nbsp;Liam Brydon-Brown,&nbsp;Rachelle Binny,&nbsp;Bruce Warburton,&nbsp;Katerina Taškova","doi":"10.1016/j.ecoinf.2026.103930","DOIUrl":"10.1016/j.ecoinf.2026.103930","url":null,"abstract":"<div><div>Effective monitoring of predator eradication and control efforts critically depends on reliable detection across diverse and heterogeneous environments. However, existing search strategies do not take full advantage of the developed AI-enabled sensing technologies. In this paper, we present SWARM-IS, a novel adaptive sensor swarm strategy that dynamically optimises the spatial deployment of predator-detecting sensors based on their collective detection histories. The method leverages collective swarm information to iteratively guide sensor placement, enabling responsive adaptation to evolving environmental and ecological conditions. To evaluate the proposed approach, we developed a modular surveillance simulation framework that models varying predator densities, spatial ecology, and sensor detection efficacy. The performance of SWARM-IS is compared against conventional systematic and random search strategies across a range of ecological scenarios. The framework is generalisable and can be readily adapted to other predator surveillance and management contexts. Using this framework, we conducted a case study on brushtail possum surveillance under diverse ecological conditions and detection-efficacy regimes. Results demonstrate that the proposed adaptive swarm-based strategy can achieve up to a 40% absolute improvement in the number of individuals removed from the site compared to conventional approaches, highlighting its potential to enhance predator monitoring and, consequently, improve management outcomes.</div></div>","PeriodicalId":51024,"journal":{"name":"Ecological Informatics","volume":"97 ","pages":"Article 103930"},"PeriodicalIF":8.5,"publicationDate":"2026-08-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"148657289","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":2,"RegionCategory":"环境科学与生态学","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"OA","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}

{"title":"Drivers of tree growth across Europe: An explainable AI analysis of tree- and site-level influences","authors":"Grégory Mermoud,&nbsp;Glory Mary Givi,&nbsp;Raphaël Lüthi,&nbsp;Volodymyr Trotsiuk,&nbsp;Nenad Potočić,&nbsp;Tanja G.M. Sanders,&nbsp;Bruno De Vos,&nbsp;Arthur Gessler,&nbsp;Stefan Hunziker","doi":"10.1016/j.ecoinf.2026.103932","DOIUrl":"10.1016/j.ecoinf.2026.103932","url":null,"abstract":"<div><div>Climate change is increasingly impairing forest ecosystems in Europe, reducing tree vitality and increasing mortality. Crown defoliation and stem growth are widely used indicators of early stress responses, yet their drivers remain difficult to disentangle due to complex, non-linear interactions among climatic, edaphic, and biotic factors.</div><div>Here, we apply explainable AI (XAI) to model annual diameter growth of individual trees from four dominant European tree species (<em>Norway spruce, Scots pine, Common beech, and Oak</em>) using long-term data from the ICP Forests Level II network. Gradient-boosted decision trees trained on tree-level attributes (e.g., defoliation, social class) and plot-level variables (e.g., soil solution chemistry, atmospheric deposition, and topography) outperform linear baselines across all ablations and grouping strategies. Ablation experiments show that plot-level features account for most of the predictive power while defoliation contributes only marginally.</div><div>XAI analyses reveal strong non-linear, species-specific response regimes, including marked growth reductions at high defoliation levels (∼35–60%) and optimal regimes of nitrogen and sulphate deposition, illustrating the capacity of XAI to identify candidate growth-relevant regimes and interactions.</div><div>However, temporally explicit validation using tree-wise cross validation results in a 54% reduction in R² score whereas spatially explicit validation based on plot-wise cross-validation leads to a near-complete collapse in predictive performance, indicating strong reliance on both temporal and spatial autocorrelation and context-specific patterns. Moreover, ablating defoliation features causes attribution to shift toward correlated environmental variables, highlighting the role of proxying and statistical confounding.</div><div>Overall, our results illustrate both the potential and the limitations of XAI for forest ecology: while effective for screening large observational datasets and generating hypotheses, XAI outputs require cautious interpretation, mechanistic understanding and spatially robust validation.</div></div>","PeriodicalId":51024,"journal":{"name":"Ecological Informatics","volume":"97 ","pages":"Article 103932"},"PeriodicalIF":8.5,"publicationDate":"2026-08-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"148657290","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":2,"RegionCategory":"环境科学与生态学","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"OA","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}