Environmental Research Letters

Agroforestry is a cornerstone of Natural Climate Solutions, yet the hierarchical importance of its soil organic carbon (SOC) drivers remains poorly resolved across heterogeneous tropical landscapes. Current global assessments predominantly rely on categorical system typologies that mask the continuous influence of biophysical drivers, leaving the reliability of mitigation estimates unclear. Here, we synthesize 643 observations from 54 field studies in Latin America and the Caribbean to decouple the determinants of SOC sequestration using a machine-learning framework. We show that baseline soil carbon stocks and temporal kinetics override management design, collectively explaining [~]85% of sequestration variability, whereas system typology and species richness contribute marginally (R2<0.10). While the median SOC storage rate was 0.26 Mg C ha{superscript 1} yr{superscript 1}, accumulation followed a distinct non-linear trajectory: sequestration intensity peaked early before decelerating sharply after a critical inflection at year 7. Critically, sequestration is governed by a robust negative feedback from initial SOC stocks, which cross a zero-net-gain threshold at [~]80 Mg C ha{superscript 1}. Depth-resolved analyses reveal that subsoil layers (up to 55-75 cm) exhibit a cumulative relative response up to fourfold greater than surface horizons, indicating that conventional shallow monitoring could systematically underestimates long-term stabilization potential. Our findings demonstrate that current carbon accounting frameworks, rooted in generic system averages (IPCC Tier 1), are structurally limited by their inability to account for baseline-dependent saturation feedbacks and non-linear effects. Transitioning toward Tier 3 context-aware, depth-explicit modeling is therefore essential to transform agroforestry from a broad practice into a precision-based, high-integrity Natural Climate Solution. HighlightsO_LISoil carbon sequestration in tropical agroforestry is primarily controlled by baseline soil conditions and temporal dynamics rather than system typology. C_LIO_LIDepth-resolved analyses reveal long-term carbon stabilization processes overlooked by surface-based assessments. C_LIO_LICarbon accumulation is strongly front-loaded, declining sharply after system establishment. C_LIO_LIContext-dependent responses challenge generic carbon accounting frameworks and highlight the need for predictive, site-specific deployment of agroforestry. C_LI

Reliable freshwater access drives terrestrial wildlife movements and habitat use globally. The small, rain-fed seasonal pools critical for dryland wildlife persistence are vulnerable to rising temperatures and unstable precipitation regimes projected under climate change. In southern Africa, which is expected to warm rapidly by 2100, the drying and disappearance of surface water may cause a breakdown in seasonal migrations of large, area-sensitive, and water-dependent wildlife species. Furthermore, the disappearance of ephemeral water may concentrate wildlife around remaining surface water, increasing resource competition and human-wildlife conflict. An accurate understanding of the dynamics and drivers of seasonal surface water will therefore be critical to wildlife and human health as climate change intensifies. Here, we present a framework and empirical analysis of fine-scale surface water mapping in the 520,000km2 Kavango Zambezi Transfrontier Conservation Area (KAZA), the worlds largest terrestrial conservation area. From 2019-2025, we implemented Otsu thresholding on median Automated Water Extraction Index imagery from 10m Sentinel-2 MSI, leveraging high wet season contrast between vegetation and water as a dry season positive mask. We created >35 quasi-monthly KAZA-wide Ephemeral Surface Water (ESW) rasters (mean classification accuracy 87%, compared to 50% accuracy for existing water products), and found wet season precipitation drivers of non-riparian water fill levels did not extend into the dry season. Then, using GPS data from 27 African savanna elephants (Loxodonta africana), which typically visit water every 48 hours, we compared elephant water visitation rates based on ESW to existing 30m Global Surface Water (GSW) maps. Models using ESW estimated 99% of elephant data came within a 48-hour window, compared to 42% for GSW, suggesting that ESW is a better proxy for actual wildlife water use in animal movement modeling. As aridification threatens to diminish surface water resources, we must model the drivers of wildlife movements at the scale of wildlife needs. With ESW, we provide fine scale accessible surface water data and a straightforward coding architecture for applications beyond KAZA.

AbstractThe ability to accurately assess ecosystem C budgets across scales from individual sites to continents is essential for C accounting, management, and ultimately mitigating climate change. State data assimilation (SDA) provides a framework for harmonizing observations with models, while robustly accounting for and reducing multiple sources of uncertainty. In this study, we employed a hybrid SDA framework that combines process-based terrestrial biosphere modeling, hierarchical Bayesian inference, and machine learning to harmonize bottom-up and remotely-sensed data streams for 8,000 pre-selected 1km2 locations across North America within a hybrid structure. Combining bottom-up soils data (SoilGrids) with spectral (MODIS and Landsat) and microwave (SMAP) remote sensing helps constrain the major C and water stocks through space and time. Machine learning is used both to identify and correct systematic errors in the process model (SIPNET) and to interpolate the pre-selected locations onto a 1km grid, making it computationally feasible to generate annual ensemble maps of the North American carbon budget. Furthermore, the uncertainties for each variable were reduced compared to those from observations or models alone. Spatiotemporal analysis showed a slight decrease in aboveground biomass (AGB) across the western US, a loss of leaf area across the boreal, and a slight greening of the Alaskan tundra. The uncertainty trends suggest a significant reduction in the uncertainty about soil organic carbon (SOC), the largest C reservoir. Validation results show that we accurately estimate C pools, compared to the assimilated data streams and held-out observations of AGB from GEDI, ICESat-2, and the US FIA, and SOC from the ISCN network. Our ML-debiasing algorithm further improved the accuracy of major C pools (AGB, SOC). In general, our continental SDA framework will facilitate global C MRV (monitoring, reporting, and verification) by providing accurate and precise C-cycle estimates, along with their corresponding spatiotemporal uncertainties.

Peatlands are an important terrestrial carbon sink which, when drained, can produce substantial CO2 efflux. Low productivity forestry planted on drained peatlands can become a net carbon source if losses from drained soils exceed sequestration by the trees. Decision support tools which assist resource allocation and intervention planning in forest-to-bog restoration are needed to mediate this substantial environmental harm. Predicting carbon mitigation benefits associated with forest-to-bog restoration is a major challenge, however, due to the lack of long-term monitoring programs and the fact that mitigation times depend on processes distant from the intervention. Here we introduce the PEATREST life cycle assessment (LCA) which predicts carbon fluxes associated with forest-to-bog restoration, including due to processes far from restored sites. The LCA estimates mitigation timescales defined as the time following intervention at which the restored peatland is predicted to sequester or store more carbon than the forestry would have if retained. HighlightsO_LIHere we develop a novel forest-to-bog Life cycle assessment (LCA) tool C_LIO_LIThe LCA predicts carbon mitigation times following peatland restoration C_LIO_LIThe model combines a variety of process-based and empirical sub-models C_LIO_LIExample implementations for two different restoration scenarios are explored C_LIO_LISensitivity analysis highlights the model inputs that most impact outcomes C_LI Graphical abstract(A single, concise figure that serves as a visual summary of the main research findings described in your manuscript.) O_FIG O_LINKSMALLFIG WIDTH=200 HEIGHT=80 SRC="FIGDIR/small/715261v1_ufig1.gif" ALT="Figure 1"> View larger version (18K): org.highwire.dtl.DTLVardef@f243f5org.highwire.dtl.DTLVardef@14bc4c7org.highwire.dtl.DTLVardef@164261borg.highwire.dtl.DTLVardef@1db3b_HPS_FORMAT_FIGEXP M_FIG The PEATREST Life cycle assessment (LCA) generates compound time series of carbon sequestration and carbon storage for two scenarios: the forest-to-bog peatland restoration (PR) and a counterfactual (CF) of forestry retention. By comparing the two scenarios, the LCA predicts the carbon mitigation timescales (vertical dashed lines). These are defined as the time following harvesting at which the peatland is predicted to sequester more (emit less), or to have stored more (lost less) carbon, than the forestry would have if retained. C_FIG

The voluntary blue carbon market is severely bottlenecked by outdated methodologies that apply broad, coast-level carbon averages across low-resolution spatial units, systematically failing to account for micro-site ecological realities and critical socio-political constraints. To resolve this structural deficit, this paper introduces the High-Resolution Geographically-Explicit Blue Carbon Assessment (HiGEBCA) pipeline, an innovative geospatial architecture that shifts site intelligence from monolithic raster grids to a topologically verified, hyper-dimensional polygon infrastructure. Operating on 1,601 distinct mangrove features across Colombia, the pipeline mathematically binds 47 ecological attributes to each polygon, integrating Monte Carlo uncertainty propagation, climate-stratified soil organic carbon, and rigorous biodiversity quantification spanning 293 taxon-code pairs. A diagnostic CatBoost machine learning emulator (R{superscript 2} = 0.926) deployed within the pipeline empirically demonstrates that local climate classes and biodiversity metrics drive over 96% of the variance in carbon density, proving that traditional broad biome classifications are inadequate for accurate micro-site valuation. Crucially, the HiGEBCA framework pioneers the integration of operational reality into natural capital assessment. When applied to Colombias theoretical national estate of 276,430 hectares (containing an estimated 478 million tCO2e), the pipeline executes a rigorous REDD+ white space assessment alongside hard mathematical filters for legal land tenure, armed conflict, and regulatory overlap. This strict governance filtration shatters the illusion of massive, easily accessible natural capital, systematically reducing the viable, investment-grade portfolio to a highly de-risked 4,000 to 12,000 hectares. Designed for cross-jurisdictional replication, the HiGEBCA pipeline establishes a new, transparent standard for prioritizing high-integrity blue carbon assets, providing a quantitative mandate for investors seeking to maximize climate impact, capture biodiversity premiums, and definitively mitigate operational risk.

Mountain ecosystems face unprecedented pressures from anthropogenic activities and climate change, challenging the productivity of these vital habitats. In the Tien Shan mountains, understanding localized responses to these pressures is often hindered by the coarse spatiotemporal resolutions of available data. To address this, we combined high-resolution satellite imagery (1997-2021) to map land-cover dynamics in the Naryn oblast, Kyrgyzstan across a gradient of grazing intensities. We classified and quantified land-cover distribution over 24 years, investigating the roles of topography, elevation, and anthropogenic disturbances as drivers of change. Our results identify intermediate elevations, high degrees of disturbance, and the interaction between the two as the primary contributors to recent transitions in grassland, forest, and barren habitats. By integrating Landsat analysis-ready data, European Space Agency WorldCover dataset and digital elevation models at fine spatial scales, we provide valuable contemporary and historical landscape and habitat-level insights and a high-resolution framework for disentangling climate-driven shifts from land-use impacts. These findings highlight the urgency of localized management in remote, data-poor regions where rapid environmental change threatens both biodiversity and pastoral livelihoods. Our work serves as a critical baseline for characterizing the adaptability of semi-arid mountain rangelands under escalating global and regional pressures.

O_LITropical forest restoration is critical for mitigating biodiversity loss and climate change, including in forests impacted by selective logging. Active restoration through liana cutting and enrichment tree planting can substantially accelerate carbon recovery, potentially reducing economic pressures to convert logged forests. But its long-term biodiversity impacts remain largely unknown. C_LIO_LIUsing over two decades of bird survey data from Borneos largest logged-forest restoration project, we quantified occupancy patterns for 176 species across primary, naturally regenerating, and actively restored logged forests spanning a 30+ year post-logging chronosequence. C_LIO_LIForest-dependent, threatened and near-threatened species generally declined through time in actively restored areas, whereas many species in naturally regenerating forests progressively recovered toward primary forest levels. Between 17-40% of 66 threatened or near-threatened species had consistently lower occupancies in actively restored than in naturally regenerating forest. Across species of global conservation concern, median occupancies in restored areas remained [~]22% below primary forest even 50 years after harvests, compared with only [~]6% lower under natural regeneration. C_LIO_LIArboreal insectivores, frugivores, and predatory species appeared most negatively affected by active restoration, with 27-49% of arboreal gleaning insectivores (of 62), 13-30% of arboreal frugivores (of 40), and one-third of predatory species (of 15) showing higher occupancy in naturally regenerating forests. Sallying insectivores also showed a possible but uncertain response, whereas ground-associated frugivores and insectivores were largely unaffected by restoration treatment. C_LIO_LIConcerningly, even 50 years post-logging, up to 52% of 50 high forest-dependency species retained distinct occupancies in actively restored compared with primary forest, suggesting persistent negative impacts of vine-cutting and/or tree planting activities on avian populations. C_LIO_LISynthesis and applications. Our findings indicate that despite substantial carbon benefits, active restoration within selectively logged forests may impede the recovery of forest-dependent biodiversity. This challenges the common assumption embedded within nature-based climate solutions that carbon and biodiversity outcomes will necessarily align. Nonetheless, despite the persistent declines in bird communities, actively restored forests continued to provide key habitat for many species. Active interventions may thus still contribute to broader biodiversity conservation objectives if they protect logged areas from conversion, potentially via carbon payments. C_LI

Climate change is expected to alter species assemblages by affecting the outcome of competition between species. Investigating processes of competition remains challenging particularly in tree communities, as they unfold over extensive spatio-temporal scales. Here, we developed a deep-learning approach to leverage a novel database of 135 million simulated local-scale tree responses to climate across continental Europe to investigate changes in the competitiveness of nine major tree species under different scenarios of climate change. Specifically, we trained a Deep Neural Network on local process model projections to investigate climate change effects on indicators of competitive strength and species dominance. We found decreasing competitive strength for all investigated evergreen coniferous species across their distribution, while major deciduous broadleaved species such as Quercus robur and Fagus sylvatica increased in competitiveness. Changes in tree species competition with climate differed locally, but most investigated species lost competitive strength at their warm range edges. As a consequence of these changes, up to 19% of Europes forests could experience a change in the dominant tree species until the end of the 21st century. Our results suggest a profound climate-induced reassembly of Europes forests and identify areas that may require specific attention in forest policy and management.

Making robust and generalizable predictions within ecological systems such as forests remains challenging due to limited data availability and the slow pace of environmental change. To address this, we integrate a semi-empirical environmental process model (PRELES) to support deep learning approaches, specifically artificial neural networks (ANNs). We replicate and extend previous work on process-guided neural networks (PGNN) by introducing new model types and conducting a comprehensive hyperparameter optimisation within systematic nested cross-validation analyses in both data-thinning and extrapolative scenarios. Results show that both data-driven ANNs and PGNNs consistently outperform the stand-alone process model, while PGNNs provide additional advantages over ANNs in data-sparse settings and under transfer scenarios to unseen, changing climatic conditions. We further estimate the generalisation error for data-driven models as a function of the amount of training data, allowing for guidance on model suitability under different data availability. A variable importance analysis using accumulated local effects reveals that both PGNNs and ANNs learn simple, physically plausible relationships, whereas PRELES exhibits a strong bias toward boreal conditions and limited ability to predict unseen, climatically divergent sites. HighlightsO_LIProcess-guided, and plain neural networks outperform a calibrated process-based model (PRELES) in predicting forest ecosystem carbon fluxes. C_LIO_LIProcess-guided neural networks provide advantages over naive neural networks in sparse-data settings and show greater robustness under transferable scenarios with unseen changing climatic conditions. C_LIO_LIVariable-importance analyses using accumulated local effects show that both process-guided and naive neural networks learn simple yet physically plausible relationships between meteorological drivers and target responses, whereas the process model (PRELES) exhibits a better fit toward boreal conditions and limited ability to predict unseen, climatically divergent sites. C_LI

Renewable energy is crucial to achieve climate neutrality, but its rapid expansion can threaten biodiversity through habitat loss or fragmentation and ecological disruption. We present a spatially explicit assessment framework that quantifies biodiversity impacts from land use change associated with renewable energy infrastructure across a broad range of species groups, and identifies siting configurations that balance energy provision and conservation goals. Drawing on metrics from life cycle assessment, combined with species distribution models and siting strategies, we evaluate alternative deployment strategies. Using Switzerland as a case study, we compare three siting strategies (maximizing energy output, minimizing biodiversity impact, and a trade-off approach) for photovoltaic systems, run-of-river hydropower, and wind turbines. For solar and hydropower installations, prioritizing energy efficiency yields the highest cumulative biodiversity losses. However, these impacts can be substantially reduced with only a slight increase in land use by favouring biodiversity protection. For wind installations, strict avoidance of sensitive ecosystems may increase total impacts, as less efficient and therefore additional sites are required to achieve the same annual energy yield. Overall, our results show that trade-off-based siting strategies can effectively balance performance and biodiversity protection, highlighting that renewable energy can be provided without sacrificing sensitive ecosystems.

Marine protected areas (MPAs) are increasingly promoted as climate mitigation tools, yet guidance on their placement to maximize resilience against climate stressors like marine heatwaves remains limited. Here, we develop MPA placement guidelines that explicitly consider a mechanistic pathway through which MPAs could enhance kelp forest resilience to heatwaves: protecting fishery-targeted urchin predators to prevent kelp overgrazing. Using a spatially explicit, tri-trophic model of California kelp forests, we evaluate alternative MPA configurations across a hypothetical coastline where half the habitat experiences an increased probability of experiencing heatwaves. We found that effective MPA placement depends on whether MPAs are being newly established or reconfigured within an existing network, and that among-patch connectivity and spillover played vital roles in the relative effectiveness of different MPA configurations. Changes in resilience occurred primarily at the patch scale, with trade-offs between increased within-MPA resilience and decreased resilience in some fished areas, resulting in minimal coastwide population effects. For example, for new MPAs, large single MPAs within heatwave-prone areas maximized within-MPA resilience gains, while multiple small MPAs in heatwave refugia best supported whole-coast resilience. When reconfiguring established networks, expanding existing MPAs in refugia areas was most effective. We also demonstrate the importance of considering MPA recovery timescales: for example, relocating old MPAs to heatwave refugia yielded minimal short-term benefits due to the loss of rebuilt, previously fished, predator biomass. Our findings demonstrate that climate-adaptive marine planning should explicitly consider the spatiotemporal implications of trophic cascades, connectivity, and transient population dynamics to support ecosystem resilience.

Restoration of degraded areas and post-disturbance rehabilitation after wildfire encompass critical approaches for reducing and reversing impacts of wind erosion and sand and dust storms (SDS). However, the broad outcomes of dryland restoration and rehabilitation for wind erosion and SDS remain underexplored. Wind erosion is an emerging issue in the Great Basin of the western United States, exacerbated by invasive annual grasses and associated wildfire. Here, we assess potential wind erosion and SDS responses to wildfire, restoration, and post-wildfire rehabilitation treatments at the regional scale in the Great Basin. We used 13 years of rangeland monitoring data, the Aeolian EROsion model, and the Land Treatment Digital Library to produce counterfactual model-predictions to estimate treatment effects. Our results revealed reductions in aeolian sediment fluxes (Ln Q < 0 g m-1 d-1) across wildfire-affected regions (mean {+/-} SE: -0.070 {+/-} 0.077 Ln Q), restoration treatments in unburned areas (range: -0.867 {+/-} 0.398 to 0.480 {+/-} 0.253 Ln Q), and post-wildfire rehabilitation (range: -0.821 {+/-} 0.183 to 1.278 {+/-} 0.909 Ln Q). In particular, aerial seeding and soil disturbance restoration treatments, and post-wildfire closure-treatments had higher perennial grass cover and the most decreased Ln Q compared to untreated controls. These results represent an important regional scale assessment of wind erosion responses to restoration and post-wildfire rehabilitation. Our findings underscore the application of integrating wind erosion and SDS mitigation into restoration and post-disturbance rehabilitation programs to provide land managers with strategies to reduce land degradation while fostering ecosystem resilience.

Estimates of the marginal cost per disability-adjusted life year (DALY) averted from government health expenditure (GHE) provide an empirical basis for allocating scarce health resources to maximise population health. Existing cross-country estimates have informed priority setting in several countries and international policy discussions but are based on data that are now more than a decade old. Since then, patterns of health expenditure, disease burden, and global health financing have changed substantially. This paper provides updated estimates of the marginal cost per DALY averted for 92 low- and middle-income countries (LMIC) by applying previously estimated elasticities of the effect of GHE on health outcomes from Ochalek et al. (2018) to recent data on mortality, morbidity, population structure, and GHE. Two policy options for improving health in LMIC are assessed: (1) the implications of countries allocating 15% of general government expenditure to health consistent with the Abuja Declaration; and (2) reallocating development assistance for health (DAH) to maximise health across countries. Scenario analyses use the estimated elasticities to reflect diminishing marginal returns to health expenditure when calculating the health gains associated with additional resources. Updated estimates of the marginal costs per DALY averted range from approximately $78 to $15,789 across countries. In most countries (72%), estimates are higher than in the previous analysis, largely reflecting increases in GHE. Increasing domestic expenditure to achieve the Abuja Declaration objective would avert 234 million DALYs but require $563 billion across countries. Reallocating $39.1 billion in existing DAH could avert 133.6 million DALYs. Updated estimates provide an empirical basis for informing both domestic priority setting and the allocation of international health financing. Aligning donor funding with country-specific opportunity costs could substantially increase the global health gains achieved with limited resources.

Mesic resources, the late-season herbaceous vegetation found in riparian areas and wet meadows, provide disproportionately important forage and habitat across western U.S. rangelands, yet their response to climatic variability and anthropogenic influences remains poorly understood. Using a 40-year Landsat time series (1984-2024), we quantified trends in late-season productivity (NDVI) across 4.5 million hectares of the sagebrush biome and applied random forest models to distinguish between temporal and spatial predictors of mesic resource productivity. We identified a fundamental shift in how mesic resources respond to drought: from 1984 to 2004, mesic productivity was strongly correlated with drought severity (Palmer Drought Severity Index, R{superscript 2} = 0.92), but this relationship weakened substantially in the next two decades (2005-2024; R{superscript 2} = 0.28), during which time productivity increased despite persistent aridity. Temporal modeling identified rising atmospheric CO2 concentrations as the strongest predictor of this shift, consistent with enhanced plant water-use efficiency under CO2 fertilization. Spatially, large agricultural valley floodplains act as anthropogenic refugia, sustaining productive mesic resources through flood irrigation and subsequent groundwater recharge into late summer. These findings suggest that human water management and physiological shifts in vegetation are currently buffering mesic systems against meteorological drought throughout U.S. rangelands. However, this apparent buffering is spatially heterogeneous and may mask vulnerability to groundwater depletion, shifts in precipitation regimes, and woody encroachment. Sustaining these vital ecosystems will require conservation approaches that go beyond climate monitoring to include balanced management considering both agricultural and ecological water needs and constraints.

We urgently need to restore degraded tropical forests to mitigate the climate and biodiversity crises, but how to do so rapidly and cost-effectively remains an open question. Here we provide a long-term, landscape-scale assessment of the effectiveness of enrichment tree planting and liana cutting, the two most common restoration interventions used across many tropical regions. Leveraging one of the worlds largest and longest running forest restoration experiments, we used repeat airborne laser scanning to track the 3D structural recovery of 500 ha of selectively logged rainforest in Borneo. Over an 18-year period, enrichment planting increased mean canopy height by 1.6 m relative to unplanted controls. Remarkably, liana cutting increased canopy height more than four times faster (3.7 m over just 9 years). This recovery was jointly driven by accelerated canopy gap closure, enhanced tree growth, and a 50% reduction in tree mortality. Given that liana cutting is around 10 times cheaper to implement than enrichment planting, our results suggest it provides a cost-effective, scalable solution to accelerate the structural recovery of logged tropical forests.

Food costs are more significantly impacted by climate change as countries grow. It is well known that climate change has an impact on the productivity of most agricultural goods, but it is unclear how specifically it will affect food costs. The present research explores how the North Atlantic Oscillation (NAO) index, a widely used climate indicator, affects food prices around the world. This is achieved by applying a robust bivariate Hurst exponent (robust bHe). The research creates a color map of this coefficient using a window-sliding technique over various intervals of time, displaying an illustration that changes overtime. Additionally, the NAO index and global food prices are examined for causal connections using variable-lag transfer entropy using a window-sliding technique. The results show that notable rises in a number of international food prices for long as well as short periods are associated with significant increases in the NAO index. Furthermore, the causative function of the NAO index in influencing global food costs is confirmed by variable-lag transfer entropy. Is highly recommended as it directly connects the research to actionable outcomes for policymakers and the overarching goal of sustainability and food security. This study provides the first direct evidence of a robust, long-range cross-correlation and causal link between the North Atlantic Oscillation (NAO) index and key global food prices. It introduces a novel, robust methodological framework to visualize this time-varying relationship, offering a critical tool for policymakers and forecasting models.

Climate change can affect salmon and steelhead (Oncorhynchus spp.) throughout their anadromous life cycles, yet there have been no assessments of which Canadian populations face the greatest exposure. We developed a framework to quantify relative climate change exposure of salmon and steelhead populations based on the spatial and temporal distribution of different life stages. Exposure was calculated from climate model projections for freshwater and marine climate variables considering unique impact thresholds for each population and life stage. We applied this framework to 60 Conservation Units of Pacific salmon and steelhead in the Fraser River basin, British Columbia. Lake-type sockeye had the highest exposure, driven by elevated stream temperatures during adult freshwater migration and spawning stages and relatively low thermal tolerance of marine stages. Chinook salmon were the next most exposed, while coho, pink, and chum salmon had relatively low exposure. Uniquely, steelhead exposure was driven by high stream temperatures during incubation. Our framework is broadly applicable, and our findings provide critical input for climate change vulnerability assessments and forward-looking resilience planning for Pacific salmon.

Groundwater-dependent ecosystems are biodiversity hotspots that provide habitat for specialised species. The EU Water Framework Directive (WFD) stresses the importance of identifying and protecting these ecosystems. However, they remain poorly mapped in temperate regions, as most studies have focused on (semi-) arid regions, where groundwater use by vegetation is both more prevalent and easier to detect from remote sensing. In this study, we transfer mapping approaches for groundwater-dependent vegetation (GDV) from dry climates into a novel framework for humid climates. To do so, we integrated, ECOSTRESS evapotranspiration data, together with high-resolution remote sensing data, regional geospatial data and field data to identify GDV. To test our framework, we trained and validated Random Forest models with eight predictor variables using 166 ground-truth vegetation plots to map GDV in Saxony-Anhalt (Germany). The final model achieved an overall accuracy of 0.97, identifying 2,067 km2 (41%) of GDV. Currently, only 19% are protected under the EU WFD. The proposed mapping framework offers a new solution for identifying GDV in temperate regions. The new GDV maps can contribute to managing groundwater resources and preserving biodiversity hotspots in regions facing increasing droughts, ultimately supporting implementation of the EU WFD.

Mediterranean forests are becoming increasingly vulnerable under climate change, as the growing frequency and intensity of droughts and heatwaves amplify physiological stress, reduce productivity, and heighten the risk of large-scale disturbances. Yet vegetation activity trends, as revealed by remote sensing, may obscure divergent responses between photosynthetic activity and growth, a critical early warning of forest vulnerability. Therefore, the long-term relationship between photosynthesis and tree growth remains poorly understood at regional scales, especially in Mediterranean areas. To address this challenge, we applied a mechanistic, process-based forest ecosystem model across approximately 2,400 km{superscript 2} of Mediterranean forests in southern Italy, encompassing a heterogeneous landscape characterized by diverse stand structures and species dominance. This framework enabled us to explicitly trace carbon fluxes from gross primary productivity (GPP) through allocation processes to average tree growth. By mean of a factorial approach, we identify over extended areas an emergent spatial pattern of divergence of summer GPP and radial tree growth amplified in space and time by the climate variability of the last two decades and shaped by forest legacy. Our findings reveal also that canopy-level greening can mask structural vulnerability and previsual decline across Mediterranean forests. Data show as an apparent long-term trend in photosynthesis decline during summer, not necessarily translates to tree growth decline. Improving our ability to determine if, where and when a key change in forest behaviour will occurs, remains essential for designing effective restoration measure and anticipating tipping points in forest resilience under accelerating climate change.

In biodiversity offsetting, balancing biodiversity losses with gains can be achieved by using multipliers that define the ratio between the magnitude of biodiversity loss and the area required to deliver equivalent biodiversity gains. Although there is broad scientific consensus that multipliers should be calibrated to deliver no net loss or a net gain for biodiversity, they are often applied without quantitative assessment of the ecological outcomes of offset actions. Here we operationalise the Response-based Habitat Hectare Assessment of Biodiversity Gains (REHAB), a framework where multipliers are informed by an understanding of habitat-specific ecological responses to offset action. To support Finlands national biodiversity offsetting scheme, we harnessed the knowledge of 111 experts to compile ecological attributes and condition matrices for all 388 Finnish habitat types and derive 346 offset action multipliers that represent ecological response functions for 216 habitat type-specific offset actions including restoration, management and passive recovery. Our analysis reveals substantial variation in response-functions, resulting in offset multipliers between 1.3-4,000 across offset actions and habitat types. We find that the fixed multipliers commonly used in offset schemes would result in net loss in 60% of the cases if action- habitat specific responses were not considered. This variability underscores that fixed multipliers cannot deliver reliable biodiversity outcomes and should be avoided in offsetting schemes. The REHAB framework has already been integrated into Finlands national offsetting policy. Other potential areas of application include informing ecosystem restoration planning and assessing biodiversity gains linked to credit issuance in emerging nature-credit markets.