Search Results (67)

Global streamflow reanalyses such as GloFAS-ERA5 are available everywhere yet lose fidelity in small, snow- and glacier-fed headwaters that feed Central Asia’s transboundary rivers. We present an explainable, calibrated deep ensemble framework that corrects the GloFAS-ERA5 log-residual at gauges of the Syr Darya and Amu Darya systems using the CA-discharge archive. An entity-aware long short-term memory (LSTM) backbone drives a regime-gated mixture of experts trained under a closed-form mixture continuous ranked probability score (CRPS) and augmented with snow physics constraints and a regime-conditional (Mondrian) conformal layer; skill was assessed under temporal holdout, leave-one-basin-out and prediction in ungauged region protocols, with grouped Shapley value attribution. Correction rendered all 74 gauges skillful, raising the median modified Kling–Gupta efficiency (KGE′) from 0.386 (raw) to 0.825 (flagship); on temporal point skill the framework is statistically tied with gradient boosting (paired Wilcoxon p = 0.49). Under-dispersed raw intervals (90% coverage 0.68) were recalibrated to near-nominal coverage (~0.90), and high-flow exceedance decision skill was moderate (Q90 Brier skill score 0.34, ROC-AUC 0.92), while low-flow (Q10) exceedance showed no skill over climatology. Transfer to ungauged, more glacierized catchments was a measured limit that degraded with glacier fraction and basin area. The framework’s value is calibration, an inspectable (supervised) regime structure and regional physical insight, not point skill superiority. Full article

Mangrove canopy height (MCH) is a fundamental structural variable for monitoring ecosystem health and quantifying carbon stocks. However, MCH retrieval from optical satellite imagery is often constrained by spectral saturation in dense stands and environmental noise in intertidal zones. This study investigates the utility of kernel-based spectral features (KVIs) as non-linear topological enhancements for MCH estimation by integrating GEDI spaceborne LiDAR with Sentinel-2 and Sentinel-1 data across three mangrove ecosystems along the South China coast. Utilizing four regression models under spatial cross-validation, we evaluated the performance of traditional indices, KVIs, and integrated feature sets against GEDI reference measurements. Results indicate that traditional optical indices exhibit limited linear sensitivity to MCH. Rather than serving as universal accuracy boosters, KVIs function as non-linear stabilizers by redistributing spectral values in Hilbert space, which effectively enhances feature representation in high biomass stands and mitigates background noise. Furthermore, model comparisons reveal that while traditional indices yield competitive baseline accuracy in specific architectures (e.g., 1D-CNN), kernel-based features provide nuanced advantages in spatial stability and ensemble dispersion reduction. Ultimately, this study demonstrates that kernel-based features enhance model robustness under rigorous cross-validation, providing a reliable structural foundation for large-scale ecological monitoring and regional carbon dynamics assessments in heterogeneous coastal environments. Full article

The “spin-up” problem, in which convection-permitting models require hours to develop realistic clouds from large-scale initial fields, critically limits short-term severe weather forecasting. Cloud analysis can serve as a feasible approach to directly assimilate hydrometeor information from remote sensing retrievals. In this study, we leverage multi-source remote sensing data, including three-dimensional mosaic radar reflectivity, hourly averaged FY-2G satellite brightness temperature (black-body temperature, TBB), and FY-2G total cloud water products, within a stepwise cloud analysis initialization scheme. The scheme is implemented in a convective-scale ensemble forecasting system (CMA-Meso, 3 km resolution) for a heavy rainfall event. For each ensemble member, three-dimensional hydrometeor increments are independently generated from these remote sensing retrievals and gradually introduced over the first ten time steps, ensuring smooth coordination with the model’s dynamic thermal framework. Quantitatively, the scheme reduces near-surface Continuous Rank Probability Score (CRPS) errors, improves the overall predictive skill by 2.6–7.9% (maximum at the 12 h spin-up period), and increases ensemble spread by 2–5.8%, mitigating under-dispersion. Probabilistic precipitation forecasts show uniform area under the relative operating characteristic curve (AROC) improvements across all thresholds, 1.16–5.77% for light rain, 3.03–8.97% for moderate rain, and 6.00–12.07% for heavy rain, with these maxima consistently occurring at the 12 h spin-up time. Although Brier scores are marginally larger, these AROC gains confirm the enhanced discrimination of convective rainfall. At 500 hPa, CRPS reductions of 7.1–15.6% emerge after 24 h (largest 15.6% for geopotential height at 24 h), zonal wind CRPS is reduced by 2.2% at 12 h, and ensemble spread increases by 3.1–7.0% for all three variables. These improvements, particularly the pronounced benefits during the initial 12 h, demonstrate that the remote sensing-driven cloud analysis effectively shortens spin-up. Mechanistically, the gains arise from physically coordinated hydrometeor-latent heat perturbations and subsequent cloud radiation feedback that continuously regulate thermal-dynamic structures. This study establishes that assimilating diverse remote sensing data via cloud analysis is an effective approach for overcoming spin-up challenges in convective-scale ensembles. Full article

Handeliodendron bodinieri is a rare tree species endemic to the karst mountains of southwestern China and is listed as a national Class II key protected wild plant. Due to its highly specialized habitat requirements and limited reproductive dispersal ability, the species is exceptionally vulnerable to environmental changes. To predict the impacts of future climate change on its habitat suitability while minimizing predictive uncertainties, this study compiled multi-source occurrence data alongside bioclimatic, topographic, and edaphic variables. We deployed an ensemble species distribution modeling framework to project the current and future (2100) distribution patterns and dynamic shifts in suitable habitat for H. bodinieri under three Shared Socioeconomic Pathways (SSP126, SSP370, and SSP585). Three independent evaluation metrics (AUCprg, TSS, and the Boyce index) consistently demonstrated that the ensemble model significantly outperformed individual algorithmic models (GLM, MARS, MAXENT, and RFd). Under current conditions, the weighted-average ensemble model estimated the suitable habitat area for H. bodinieri to be 70,582 km². By 2100, the suitable habitat area is projected to contract under all three SSP scenarios, with reduction rates ranging from 27.38% to 80.43%. Concurrently, the centroid of the suitable habitat is expected to shift southeastward or southward by 13–40 km, accompanied by varying degrees of shifts in mean elevation. Furthermore, we identified climate-insensitive regions across the scenarios, which are primarily clustered along the border between Guangxi and Guizhou provinces. Our study provides a robust scientific foundation for formulating effective conservation and management strategies for H. bodinieri amidst ongoing climate change, ultimately maintaining the sustainable survival of the species. Full article

As integral components of ecosystems, rodents exhibit high sensitivity to environmental fluctuations. As a typical semi-subterranean rodent, the burrowing and foraging behaviors of the mole vole (Ellobius tancrei) profoundly impact local microhabitat structure and vegetation succession. Given its extensive distribution range, investigating the geographical distribution of the mole vole under climate change is critical for understanding its population dynamics and spatial patterns. To explore the impacts of climate change on the potential distribution of this species, we utilized an ensemble model based on 395 occurrence records and 38 environmental variables. We predicted its potential geographical distribution under current (1970–2000) and three future climate scenarios (SSP126, SSP245, and SSP585) across three future periods (2041–2060, 2061–2080, and 2081–2100) and further analyzed the variation trends and centroid migration directions of its global potential distribution. The results revealed the following: (1) Elevation (Dem) and temperature annual range (Bio7) were the dominant environmental factors driving the potential distribution of the mole vole, followed by precipitation of the warmest quarter (Bio18), mean diurnal range (Bio2), annual mean temperature (Bio1), base saturation (BSAT), precipitation seasonality (Bio15), calcium carbonate content (TCARBON_EQ), and available water capacity (AWC). (2) Under current climate conditions, the highly suitable habitats were mainly concentrated in Eurasia, particularly in Turkey, Kyrgyzstan, China, Mongolia, and Russia. (3) Under future climate scenarios, suitable habitats showed simultaneous expansion and contraction, with expansion consistently exceeding contraction across all scenarios and periods, especially under SSP585 in the late 21st century. (4) Relative to the current period, the suitable-habitat centroid shifted northeastward under SSP126, but northwestward under SSP245 and SSP585. This study reveals the dynamic spatial shifts of the mole vole driven by climate change, not only providing crucial scientific support for early spatial warning and precise prevention of agricultural and pastoral rodent damage, but also laying a theoretical foundation for understanding the ecological adaptation mechanisms of subterranean rodents and mitigating potential cross-regional dispersal risks. Full article

Global ocean models at non-eddying resolutions currently used for subseasonal to seasonal to decadal (S2S2D) prediction suffer from a severe deficit in internal variability. In ensemble data assimilation (DA), this can lead to under-dispersed ensembles that require inflation schemes. However, inflation corrections do not persist into the forecast phase, causing ensemble spread to collapse at longer lead times. This study evaluates an alternative approach: addressing the variability deficit directly within the model physics using a “Backscatter Package” (BackPack) consisting of the stochastic Stanley, stochastic GM+E, and Leith+E backscatter parameterizations. Implemented within a global MOM6/CESM framework at nominal $2/3°$ resolution using the DART ensemble DA package, the BackPack’s impacts are compared against cutting-edge adaptive inflation. The results demonstrate that the BackPack substantially increases internal variability and ensemble spread, successfully lowering the amount of required inflation. While reductions in ensemble-mean state errors are modest, the BackPack significantly improves ensemble calibration, assessed using a novel spread–error calibration ratio metric. Although this study only addresses the data assimilation phase, we expect that physics-based BackPack schemes may provide a physically sustainable pathway to maintain spread during the subsequent forecast phase. Full article

Metaheuristic algorithms remain a widely used class of solvers for solving complex, non-convex optimization problems where gradient information is unavailable, yet two failure modes continue to limit their practical reach: premature convergence caused by inadequate exploration diversity in late iterations and population stagnation that persists even when individual agents are nominally assigned to the exploration phase. This paper proposes the Stagnation-Aware Aquila Optimizer (SAAO), a hybrid algorithm that addresses both failure modes by embedding three targeted mechanisms into the Aquila Optimizer (AO) framework: (i) an adaptive exploration probability that responds to global fitness-improvement history; (ii) individual-level stagnation counters that force exploration re-entry for any agent that fails to improve for more than 30 consecutive iterations, regardless of the global phase schedule; and (iii) a diversity-maintenance module that reinitializes completely stagnant agents via random sampling or opposition-based learning. The biological repertoire of search operators is simultaneously enriched by incorporating four physics-grounded operators from the Animated Oat Optimization (AOO) algorithm centroid-guided dispersal, elite-guided dispersal, hygroscopic rolling, and spring ejection, alongside the original AO operators, yielding six complementary update rules partitioned equally between exploration and exploitation. The SAAO was evaluated against nine state-of-the-art algorithms on the CEC2015 benchmark and CEC2022 under identical experimental settings. The SAAO achieved the best Friedman mean rank on both suites and delivered competitive or superior performance against the nine baselines, with Wilcoxon rank-sum tests confirming statistically significant advantages over most competitors. On three classical engineering design problems, the SAAO achieved competitive outcomes. In a real-world equipment anomaly prediction task, an SAAO-optimized ensemble classifier attained 98.23% accuracy, surpassing the compared baseline models. These results establish SAAO as a robust and computationally tractable optimizer for both benchmark and applied settings. Full article

The middle reaches of the West Liaohe River Basin, a typical semi-arid to semi-humid transition and agro-pastoral ecotone in northern China, exhibit high ecological sensitivity, low resilience, and pronounced fragility. Despite growing concerns, existing studies in this region lack a comprehensive assessment paradigm that effectively couples inherent ecological attributes with nonlinear predictive modeling. To fill this gap, we developed an integrative framework that innovatively combined the SRP conceptual model with a stacking ensemble learning technique. This coupling is methodologically novel because it moves beyond linear assumptions, enables the detection of complex nonlinear response surfaces, and establishes a seamless analytical chain from historical evaluation to future projection. By selecting 13 indicators, including topography, climate, soil, vegetation, and socio-economic factors, the weight was determined by the comprehensive application of the analytic hierarchy process and entropy weight method, and the ecological fragility of the middle reaches of the West Liaohe River Basin from 2000 to 2020 was evaluated at multiple scales. The spatial differentiation driving factors were analyzed using a geographic detector. Therefore, an Ensemble Learning Regression model was used to simulate and predict the ecological fragility pattern in 2030. The results show that from 2000 to 2020, the ecological fragility of the study area showed a decreasing trend overall, with the Ecological Vulnerability Synthetical Index (EVSI) decreasing from 3.48 to 2.68, and the spatial pattern gradually shifting from “high in the northwest, low in the southeast” to “overall stability, local optimization.” The spatial agglomeration of ecological fragility gradually weakened, indicating that high-fragility areas tend to disperse and low-fragility areas expand in contiguous areas, and the ecosystem structure tends to develop towards equilibrium. The driving mechanism shows an evolution characteristic from “soil erosion dominated” to “biological abundance dominated,” with the impact of climate factors first increasing and then stabilizing, and the direct pressure from human activities continuously weakening. Under the assumption that historical trends continue, the ensemble learning model projects that by 2030, the ecological vulnerability pattern will be dominated by Mild and Moderate levels, with the area of extremely vulnerable regions significantly reduced to 0.36%. This study verified the applicability of the SRP model in transitional river basins, and the constructed “evaluation-driving mechanism-prediction” framework can provide a scientific basis for the ecological protection and adaptive management of the West Liaohe River Basin and provide a methodological reference for ecological fragility research in similar areas. However, limitations persist: the indicator system and weight assignment are subject to inherent subjectivity, and the 2030 scenario projection based on the Stacking ensemble learning model relies on the BAU (Business-As-Usual) assumption, which fails to account for abrupt climate extremes or major policy shifts. Future studies should incorporate multi-scenario constraints to reduce predictive uncertainty. Full article

Global climate warming has intensified in recent years, with extreme weather events occurring more frequently and severely impacting ecosystems and social production. According to the “China Climate Change Blue Book (2023),” China’s temperature rise rate exceeds the global average, with increasingly significant impacts on ecosystems. Hyphantria cunea, an invasive forest pest first discovered in China in 1979, has spread widely, causing serious damage to forestry and agriculture and posing a significant threat to China’s ecological security. To address this threat, this study employed seven modeling algorithms (GLM, GBM, CTA, ANN, SRE, FDA, MARS, RF, and MaxEnt) from the R Biomod2 package to develop an ensemble model. The core research objective of this work is to quantify climate-driven range shifts of H. cunea under ongoing global climate change. Previous nationwide SDM studies on invasive forest pests have consistently demonstrated that climatic variables dominate broad-scale nationwide suitable habitat patterns at the macro-regional level. Supplementary topographic, vegetation cover, and human land-use disturbance layers were incorporated to capture fine-scale habitat filtering effects and long-distance pest dispersal facilitated by human activities, which together fully characterize the suitable regional environments of this pest. By integrating climate, topography, vegetation, and human disturbance data, we predicted the potential geographical distribution of H. cunea in China under four future climate scenarios (SSP1-2.6, SSP2-4.5, SSP3-7.0, and SSP5-8.5). The ensemble model achieved excellent performance with TSS and ROC values of 0.901 and 0.984, respectively. Currently, highly suitable areas for H. cunea are concentrated in 12 provinces, including Shandong, Jiangsu, Hebei, Henan, and Anhui, covering 56.33 × 104 km², with Shandong showing the highest proportion (25.48%). The suitable habitat range is projected to expand northeastward, with significant increases under high emission scenarios (SSP5-8.5). Analysis of environmental variables reveals that nighttime light brightness, precipitation in the warmest season, the seasonal temperature variation coefficient, and average temperature in the driest season are key factors influencing H. cunea distribution. Nighttime light brightness shows the highest contribution (27.7%), indicating significant human impact on species spread. Response curves suggest that H. cunea favors warm, humid areas with pronounced seasonal changes. This study demonstrates that climate change will increase H. cunea expansion risk, necessitating strengthened cross-regional monitoring and biological control techniques. These findings provide a scientific foundation for understanding H. cunea spatiotemporal distribution patterns under future climate scenarios and for developing effective prevention and control strategies. Full article

Raw sentiment extracted from Chinese financial news is noisy and difficult to use directly for market prediction. This study proposes a mathematical filtering framework that converts noisy Chinese financial news sentiment into reliable quantitative signals for financial market prediction. Three daily sentiment measures were constructed from Chinese financial news: sentiment mean, sentiment dispersion, and polarity imbalance. Seven filtering methods were applied to each measure, including exponential smoothing, autoregressive filtering, ARIMA filtering, moving average smoothing, discrete wavelet transform, Savitzky–Golay filtering, and Kalman filtering. The seven filtered outputs were averaged to produce an ensemble-smoothed sentiment signal. Support vector machines and neural networks were then used to compare the predictive performance of raw and filtered signals for stock index log returns and realized volatility. Filtering reduced the standard deviation of sentiment mean by 48%, sentiment dispersion by 55%, and polarity imbalance by 50%, while mean levels remained stable. Filtered sentiment consistently outperformed raw sentiment across all model configurations. The improvement was larger for realized volatility than for returns: the best support vector machine reduced volatility prediction error by 16.9% and return prediction error by 5.8%. A moderate neural network with 20 hidden neurons achieved optimal performance for both outcomes. Mathematical filtering extracts stable and informative sentiment signals from Chinese financial news. Filtered sentiment is more useful than raw sentiment for predicting market volatility, and the improvement holds across multiple machine learning models. Full article

Under global climate change, ungulate distributions are generally shifting poleward. However, the dispersal pathways, dynamics of suitable habitat, and environmental drivers of the expanding Milu population in the intensively used coastal wetlands of Jiangsu Province remain poorly understood. To support conservation and management, this study used field occurrence data and environmental variables to predict potentially suitable habitat for Milu under current and future climate scenarios. The Biomod2 ensemble modeling framework was applied to assess spatial changes in habitat suitability, and Geographical Detector was used to identify key environmental drivers. Current potentially suitable habitat showed a belt-like pattern along the coast, with the high suitability area covering 0.035 × 10⁴ km². Under future climate scenarios, potentially suitable habitat for Milu is projected to expand in the central and northern coastal areas of Jiangsu, with a substantial increase in the predicted total suitable habitat area. Dis_coastline, BIO14, BIO4, and Pop_density were identified as key factors influencing the distribution of potential suitable habitat for Milu, among which BIO4 and BIO14 were the principal climatic drivers affecting the northward shift in future suitable areas. These results suggest that Milu habitat suitability is jointly shaped by coastline proximity, temperature and water-availability conditions, and population density. Conservation should prioritize the protection of highly suitable habitats, improve patch connectivity, reduce human disturbance, and strengthen wetland protection and vegetation restoration. Full article

Global climate change is causing an increase in extreme rainfall events, which impacts the risk of hydrometeorological disasters. To support disaster mitigation and early warning systems, accurate and reliable rainfall predictions are required. Although ensemble forecasting is widely used to model atmospheric uncertainty, raw ensemble results often exhibit insufficient bias and dispersion. Therefore, post-processing techniques are needed to improve the quality of probabilistic predictions. The most commonly used calibration method is Bayesian Model Averaging (BMA). This study conducted a scoping review of peer-reviewed papers on ensemble forecast calibration using BMA, based on the PRISMA-ScR framework. Furthermore, this study presents a comprehensive bibliometric analysis involving co-authorship networks of productive authors and bibliometric maps with clustered terms. A total of 35 relevant articles were identified from 49 screened publications. The bibliometric analysis revealed that “ensemble forecasting” and “Gaussian distribution” are the most dominant terms in the research network, indicating that Gaussian-based approaches remain more widely used in ensemble forecast calibration studies. In contrast, studies explicitly applying Gamma-based approaches are still relatively limited despite their relevance for modeling asymmetric rainfall data. The results obtained in this study highlight the importance of developing and integrating more appropriate probability distributions, such as those within the Extreme Value Theory framework, into BMA models. These findings suggest that the selection of appropriate probabilistic distributions in BMA-based calibration frameworks plays an important role in improving forecast reliability and the representation of uncertainty in rainfall prediction. Furthermore, the development of more suitable probability distributions, including Extreme Value Theory (EVT)-based distributions, has strong potential to enhance probabilistic calibration performance for asymmetric rainfall data. This approach is expected to improve the accuracy and reliability of extreme rainfall predictions. The findings of this study provide an important contribution to the development of early warning systems for hydrometeorological disasters and support the achievement of Sustainable Development Goals (SDGs). Full article

Stratospheric Aerosol Injection (SAI) has been investigated as a climate intervention strategy to offset global warming, and regional impacts studies rely on simulations from the Geoengineering Large Ensemble (GLENS). The probabilistic behavior of the GLENS ensemble has not been systematically characterized for Southeast Asia. Because GLENS is a counterfactual experiment combining the Representative Concentration Pathway 8.5 (RCP8.5) forcing with active SAI, comparison with observations cannot validate the SAI response itself. In the early protocol years, the SAI forcing is small, so the early window provides a diagnostic of statistical consistency between the ensemble and the observed climate and of ensemble spread reliability. We compare the 21-member GLENS ensemble for 2020–2025 with ERA5 for daily precipitation and mean and maximum temperature using empirical coverage of the 95% prediction interval, rank histograms with the Jolliffe–Primo decomposition, the Continuous Ranked Probability Score, and the Brier Score for rainfall occurrence. Coverage is well below nominal for all variables, and rank histograms show pronounced U-shapes dominated by the dispersion error component, indicating systematic underdispersion. Because the underlying mechanisms are properties of the ensemble system rather than of the SAI forcing, this underdispersion is expected to persist in the future record, motivating statistical post-processing of GLENS before its use in SAI impact assessments. Full article

The global invasion of the shrub L. camara poses a significant threat to ecosystems. Understanding the roles of human activity and climate in driving its spread is crucial for management. This study aimed to quantify its global invasion dynamics, identify key drivers, and predict future distribution shifts. We constructed a high-precision ensemble species distribution model by integrating historical global occurrence records, multi-source environmental variables (climate and human activity indices), and future climate scenarios (SSP1-2.6 and SSP5-8.5). The global invasion showed a clear four-stage acceleration pattern (1900–1960, 1961–1980, 1981–2000, and 2001–2025). Variable importance and response curve analysis revealed a two-phase “dispersal–colonization” mechanism: human activities (e.g., gross domestic product) acted as a “dispersal amplifier,” while a climatic factor (isothermality) served as a critical “colonization filter.” Under two future climate scenarios assuming unchanged human activity patterns, the potential suitable habitat of L. camara exhibits structural changes while maintaining stable total area. The highly suitable areas continue to shrink, with nearly half the area lost by the end of the century under the high-emission SSP5-8.5 pathway, while low-suitability zones expand significantly—yet the overall suitable habitat remains stable. Under SSP1-2.6, structural changes in suitable habitats occur more gradually. The study clarifies the distinct roles of human activity and climate in the invasion process, providing a scientific basis for differentiated global risk management strategies targeting dispersal pathways and colonization thresholds. Full article

Groundwater systems in semi-arid and industrial regions are increasingly affected by climate-driven non-stationarity and anthropogenic pressure, challenging conventional forecasting approaches. This study develops and evaluates an integrated artificial intelligence framework designed to minimize piezometric head residual dispersion under non-stationary hydroclimatic conditions. The proposed methodology combines Improved Complete Ensemble Empirical Mode Decomposition with Adaptive Noise (ICEEMDAN) and Variational Mode Decomposition (VMD) with a Slime Mould Algorithm–optimized Long Short-Term Memory (SMA–LSTM) model and a CNN–LSTM architecture, which are dynamically fused using an Adaptive Weighting Model (AWM). The framework was applied to long-term groundwater level (1994–2024), groundwater quality (2017–2023), and meteorological datasets to evaluate the predictive robustness across climatic variability regimes. The proposed ensemble achieved a mean absolute error of 0.267 m, root mean square error of 0.429 m, coefficient of determination (R²) of 0.948, and Nash–Sutcliffe efficiency of 0.938, representing substantial residual reduction compared to baseline deep learning models. Residual diagnostics confirmed minimized peak deviations and stable performance under non-stationary conditions. Scenario-based simulations driven by CMIP6 climate projections indicate increasing groundwater stress under future warming trajectories, with amplified variability and declining recharge signals. These findings demonstrate that multi-stage signal decomposition coupled with metaheuristic optimization and adaptive ensemble learning significantly enhances predictive stability and residual minimization in climate-sensitive aquifer systems. The proposed framework provides a transferable, climate-resilient decision-support tool for sustainable groundwater management in industrial and semi-arid regions. Full article