Environmental and Earth Sciences

Abstract: Punjab’s rice transplantation regulation seeks to reduce groundwater demand by moving cultivation closer to monsoon rainfall. This study examines how the hydroclimatic conditions supporting that rationale have changed over time. Using CHIRPS v2.0 rainfall and ERA5-Land temperature data for 2000–2024, we compare rainfall magnitude, interannual rainfall variability, temperature, and Hargreaves reference evapotranspiration between 2000–2008 and 2009–2024 for May–July. The district groupings assigned to Punjab’s three transplantation zones in 2025 provide the spatial units for this retrospective analysis. June conditions changed differently across zones. In Zone 1, mean rainfall declined by 9.55%, rainfall variability increased, and temperature and reference evapotranspiration rose, indicating less favourable conditions within the hydroclimatic alignment framework. In Zones 2 and 3, June rainfall increased by 173.19% and 63.19%, respectively, alongside higher temperature and evaporative demand. Rainfall variability increased in Zone 2 but remained broadly stable in Zone 3, producing mixed patterns of alignment. July accounted for a larger share of May–July rainfall across all three zones. These findings suggest that the environmental basis of transplantation regulation requires spatially differentiated reassessment. They do not establish changes in groundwater extraction or the suitability of individual prescribed dates. Adaptive crop-calendar governance could retain the conservation rationale of regulation while periodically reviewing timing against changing conditions. Its feasibility depends on coordination with labour, electricity, inputs, and procurement, linking cultivation timing to both immediate production needs and longer-term groundwater conservation.

Abstract: This study analyzes the contemporary hydrological dynamics, energy generation, and agricultural irrigation projections of the Tigris River Basin through the "Water-Energy-Environment" nexus. It evaluates the impacts of climate change and infrastructural developments on upstream (Turkey) and downstream (Iraq) ecosystems. Findings reveal a critical "Investment-Resource Paradox" in the upper basin: despite climate-induced droughts reducing natural discharge by 20–30%, Turkey’s water-dependent agricultural and hydroelectric capacities are expanding rapidly. Conversely, downstream Iraq faces a hydrological collapse due to a "triple squeeze": upstream regulations by Turkey, damming of eastern tributaries by Iran, and inefficient domestic irrigation. This crisis has caused severe ecological degradation, including the desiccation of the Mesopotamian Marshlands and saltwater intrusion, acting as a threat multiplier for socio-economic instability. The study concludes that traditional supply-oriented management is unsustainable. Securing the nexus requires a shift towards demand management, focusing on agricultural efficiency (digital micro-irrigation), binding environmental flow protocols, and a tripartite (Turkey-Iraq-Iran) multilateral basin diplomacy mechanism to govern shared climate risks.

Abstract: Artificial intelligence is part of the human-built technosphere. The technosphere is the global system of machines, structures, and supply chains. It depends on the biosphere, and it competes with species and systems of the biosphere for energy, water, land, and materials. The biosphere is already outside its safe operating space. Seven of nine planetary boundaries have been crossed. Human use of the energy fixed by plants on land is near three times natural productivity. AI now helps draft, rank, and screen decisions in farming, energy, land use, water, and conservation. It also adds a fast-growing resource load of its own. This paper argues that AI used for biosphere-relevant decisions should not decide ecological hard limits by reading free text. Training on human preferences shapes ordinary behavior but does not encode ecological limits. Refusals that rest on a model reading prose can fail when the request is reworded. The paper presents Biosphere Sentinel, a reference design. In it, a language model may draft advice and fill an action form. A separate rule program outside the model then decides hard limits on coded, governed fields and returns one of four outcomes. The paper reports what the project tested and what it did not. A base model given the full constitution as a prompt hedged on both adversarial hard-limit queries. A coded energy-source field, with non-certification for off-list entries, ended a free-text failure cycle. Wrong approvals were zero across eight reference cases, and 161 locked checks passed. The decisive test, whether a language model can fill the coded fields dependably, was designed but not run. The paper records the peer review that led to withdrawal, the reasons the project was halted, the governance hazards any such design carries, and the conditions under which the work could resume.

Abstract: Unraveling the spatial differentiation and multiscale driving mechanisms of cropland vegetation in the Loess Plateau–Qinghai-Tibet Plateau transition zone provides vital scientific support for sustainable agriculture in ecologically fragile regions. Taking Yongjing County, Gansu Province as the study area, based on 30m resolution Enhanced Vegetation Index(EVI)from 2015 to 2024 and 9 topographic, hydrological, and climatic covariates, an integrated multi-model analytical framework coupling Geographical Detector (Geodetector), Random Forest with SHAP(SHapley Additive Explanations), Geographically Weighted Regression(GWR),and Variation Partitioning Analysis (VPA) was established. Furthermore, systematic robustness tests were conducted across four dimensions: vegetation index substitution, discretization schemes, spatial block sizes, and MGWR bandwidths. The results revealed that: (1) Cropland EVI exhibited significant moderate spatial positive autocorrelation(Moran’s I=0.494), with homogeneous spatial clustering being overwhelmingly dominant(95.8%); (2) Under the natural breaks discretization, the Topographic Wetness Index(TWI)yielded the highest explanatory power(q=0.200), but was highly sensitive to discretization methods (q dropped to 0.028 under quantile classification, a discrepancy of 0.177) and displayed an extremely narrow MGWR bandwidth(44 nearest neighbors), identifying it as a locally sensitive factor rather than a global governing factor; (3) Multi-model cross-validation and multidimensional robustness tests identified Digital Elevation Model(DEM),aspect cosine(Aspect_cos, northness), and mean annual temperature(Temp) as the three robust dominant factors. DEM ranked among the top three across all four methods with a discretization range of only 0.005,functioning as the global dominant factor orchestrating hydrothermal redistribution; Aspect_cos had the highest proportion of statistically significant pixels in GWR(77.4%), acting as a steady local dominant factor; Temp exhibited the greatest local coefficient variability(0.046), representing a spatially non-stationary dominant factor characterized by a pronounced north-south effect reversal. Overall, cropland vegetation in the transition zone is synergistically regulated by topography, hydrology, and climate, wherein topography governs macro-scale hydrothermal differentiation, while climatic drivers manifest intense spatial heterogeneity.

Abstract: Cartography spans millennia of human civilization. Maps helped humans interpret and represent the world before writing. This review examines the evolution of cartographic knowledge from Prehistory to the Middle Ages. It focuses on how maps helped people understand, represent and navigate the world. The methodology used bibliographic search and analysis. Sources came from academic libraries, bibliographic databases and repositories. The literature was divided into Prehistory, Antiquity, and Middle Ages. Approximately 20 maps were selected for their innovations or cosmological significance. They were grouped by orientation, material, content, and Earth’s shape for systematic comparison. The study of maps showed that the Earth‑centred model changed over centuries, slowly replaced by heliocentrism. In Antiquity, Greeks proved the Earth is a sphere, a view that remained mainstream throughout the Middle Ages. Map orientation varied, with Babylonian and Ptolemaic maps facing north, mappae mundi facing east, while Islamic ones facing south. Cartographic materials changed over time, with prehistoric maps on rock, Antiquity using clay tablets, metal, papyrus and parchment, while in the Middle Ages, parchment became common. Content changed with prehistoric artefacts used for local land, in Antiquity became mathematical and administrative, Medieval ones showing religious symbols. By the 13th century, maps had become practical navigation tools.

Abstract: Winter Storm Uri (February 2021) caused the largest power grid failure in Texas history, with extensive outages in the state's most populous Harris County. This study integrates NASA Black Marble VNP46A2 nighttime light (NTL) observations with the CDC Social Vulnerability Index (SVI) to characterize post-storm recovery across 1,070 census tracts. Recovery was defined as the first day tract-level NTL reached at least 90% of its pre-storm baseline for three consecutive days. Contrary to the expectation that vulnerable communities recover more slowly, spatial regression revealed a weak, statistically significant negative association between SVI and recovery time (β = −0.556, p = 0.050), with spatial dependence also significant (ρ ≈ 0.47). Local Indicators of Spatial Autocorrelation identified significant clusters of slow recovery in lower-SVI suburban areas and fast recovery in higher-SVI urban neighborhoods. This association was sensitive to the recovery definition, reversing signs under longer persistence requirements. The spatial organization of urban infrastructure, rather than vulnerability alone, may plausibly explain these patterns, though not directly testable with available data. These findings caution against assuming a universal vulnerability–recovery relationship in infrastructure disasters and highlight the sensitivity of nighttime-light-based equity assessments to how recovery is operationally defined.

Abstract: In mountain areas, tourism changes land use and puts development, environmental fragility and livelihoods at odds, and managing the competition is not always simple. The question this study poses is whether the stakeholders that communities consider to be central to land-use governance in the tourism-pressured Upper Beas Basin of Himachal Pradesh, India, are also found to be effective? The evidence is derived from a structured household survey conducted over three rounds of field visits, between 2024 and 2026, with 300 respondents across nine settlements. We created three composite measures from the Likert-scale responses: a Stakeholder Role Index (SRI), a Stakeholder Effectiveness Index (SEI) and a Community Perception Index (CPI); and calculated a Role-Effectiveness Mismatch (REM) as SRI – SEI. Friedman and Wilcoxon procedures were used to test for differences among the stakeholders, Kruskal-Wallis and Holm-adjusted Mann-Whitney tests were used to test for differences across the tourism contexts, and the results were compared with HC3 regression. The highest scores were achieved by local communities in both role (SRI = 3.435) and effectiveness (SEI = 3.268). The private sector had the highest positive mismatch (REM = +0.381), while the NGOs/Civil Society had the highest negative mismatch (REM = -0.420). The tourism context had little impact on SRI and SEI but was significant for CPI (H = 10.609, p = 0.005) with high and intermediate tourism settlements being significantly different. Overall, the analysis reveals that prominence in governance and perceived effectiveness are two separate dimensions and that REM offers a practical way to locate where-and for which actors-the two diverge in mountain land governance.

Abstract: The occurrence of anthropogenic materials in commercially important fish provides relevant information on environmental exposure in coastal urban systems. This study investigated materials detected in the gastrointestinal tracts of croaker (Micropogonias furnieri) and horse mackerel (Trachurus trachurus) marketed in Santos, Brazil, and Maputo, Mozambique, integrating µ-Raman spectroscopy, frequency analysis, and correspondence analysis with market categories and market-associated socioeconomic profiles. µ-Raman analysis yielded 65 material and compound spectral assignments, including carbonaceous materials, synthetic polymers, cellulosic materials, pigments, dyes, and other chemical signatures. Carbon and graphite were the most frequent components, while polyethylene terephthalate (PET) and polyester-related materials predominated among synthetic polymers. Both study areas showed comparable spectral-occurrence patterns, particularly for PET, graphite, and cellulosic materials. Global chi-square tests showed no statistically significant dependence between material occurrence and either market categories or socioeconomic profiles. However, adjusted standardized residuals and correspondence analysis revealed localized distributional patterns involving specific materials and market categories, which should be interpreted as exploratory rather than causal relationships. Integrating µ-Raman spectroscopy and multivariate analysis provides a useful framework for characterizing chemically diverse materials associated with gastrointestinal exposure in commercially important fish. Because we analyzed only gastrointestinal tissues, the findings do not demonstrate contamination of edible muscle tissues, human dietary exposure, or health risk.

Abstract: Mountain karst landscapes present major challenges for reconstructing past habitation because suitable environments are spatially fragmented and archaeological remains are often dispersed. This study integrates GIS-based multicriteria decision analysis (GIS-MCDA) and machine-learning methods to model habitation suitability in Biokovo Nature Park, Croatia. Twenty-one morphometric, hydrogeomorphological, climatic, and archaeological criteria, together with 804 reference polygons, were used to develop four habitation suitability models: Equal-Weight GIS-MCDA, Analytic Hierarchy Process (AHP) GIS-MCDA, Random Forest (RF), and XGBoost. All models showed strong independent discrimination, with AUC values of 0.9152 for Equal Weight, 0.9447 for AHP, 0.9876 for RF, and 0.9886 for XGBoost. The final XGBoost model indicates that favourable habitation environments are limited and spatially discontinuous, with high and very high suitability occupying only 16.72% of the study area. Suitable zones are concentrated within distinct karst micro-landscapes, particularly around dolines. The results support the interpretation of Biokovo as a selectively used mountain landscape and demonstrate the potential of combining GIS-MCDA and machine-learning approaches for archaeological prospection in Mediterranean and Dinaric karst environments.

Abstract: Flash floods represent one of the most devastating natural hazards worldwide, causing significant loss of life and economic damage due to their rapid onset and limited warning time. Flash Flood Early Warning Decision Support Systems (FF-EWDSS) have emerged as critical tools for disaster risk reduction, integrating meteorological forecasting, hydrological modeling, real-time monitoring, and communication technologies to provide timely alerts to vulnerable communities. This review examines the current state of global FF-EWDSS implementations, analyzing their key technical components, operational frameworks, and geographic applications. We synthesize evidence from 30 highly relevant studies spanning diverse hydroclimatic regions, from tropical Pacific islands to Mediterranean catchments and arid zones. The review identifies core system components including multi-sensor meteorological inputs (radar, satellite, numerical weather prediction), distributed hydrological models, real-time monitoring networks, and multi-channel communication systems. Major challenges include data scarcity in developing regions, model uncertainty in ungauged catchments, computational constraints for real-time operations, and the critical "last-mile" problem of translating technical warnings into protective community actions. Emerging technologies such as artificial intelligence, Internet of Things (IoT) sensors, cloud computing, and ensemble forecasting show promise for enhancing system performance and extending forecast lead times. This review provides a comprehensive foundation for researchers, practitioners, and policymakers working to advance flash flood early warning capabilities globally.

Abstract: Evaluating Sustainable Urban Built Morphology (SUBM) paĴerns at a metropolitan-wide scale obscures the localized trends in urban morphology across space and time. This research introduces Development Morphology Units (DMUs) as a micro-scale analytical concept for investigating the fine-grained spatio-temporal dynamics within urban morphology regimes. Based on 5,844 development tracts observed between 1990 and 2023 in Mecklenburg County, North Carolina, this study utilizes within-regime Spatio-Temporal Density-Based Spatial Clustering of Applications with Noise (ST-DBSCAN) on five previously identified regimes. These DMUs are further described according to four descriptors of temporal position, temporal span, spatial footprint, and spatial movement. The results detect substantial heterogeneity between DMUs in terms of spatio-temporal growth patterns. Peripheral and conventional suburban regimes account for large proportions of development tracts across the county, nonetheless, they revealed a limited DMUs formation ratio, with most of their tracts left unclustered, while the identified DMUs were predominantly small, localized, and short-lived. On the other hand, the accessibility-oriented regime demonstrated a cohesive DMU structure and contains sustained and spatially stable morphological units. Within the intermediate regimes, most DMUs characterize localized and episodic growth, alongside a small number of large-scale units. Also, the highest-achieving sustainability regime in the county exhibited recent, short-lived, spatially localized, and stationary units. Across all regimes, morphological growth paĴerns predominantly represent limited spatial movement, suggesting that developments sharing similar morphological characteristics tend to remain anchored to previously established areas. These paĴerns align with evolutionary urbanism, indicating that sustainability-oriented urban morphology evolves through localized, cumulative processes rather than spatially random expansion. The identified DUMs demonstrated that localized urban morphology patterns evolve distinctively across space and time. The findings inform urban sustainability practices by tracking the dynamics of sustainability-oriented urban morphology at local levels.

Abstract: Glacier forelands are well-suited for investigating vegetation development and plant colonization dynamics in previously uninhabited areas. The areas exposed by retreating glaciers represent sites without seed banks, thus plant colonization almost always represents a true primary succession. Vegetation dynamics relies on diaspores brought in from the surrounding. A common approach to study long-term vegetation developments is the employment of so-called chronosequences, deriving temporal sequences from spatially different sites of various site age. In glacier forelands, datable traces of former glacier extents are generally used for site age estimation. In 2009 and 2010, vegetation sampling for ten points in time since deglaciation were carried out in the foreland of Schwarzenbergferner, Ötz valley, Austria. The study area extends between the recent glacier terminus and the terminal moraine left by the Little Ice Age glacier during maximum extent (around 1850). The vegetation survey is supplemented by a one-year long measurement of soil temperature (September 1 2010 to August 31 2011), which allows not only for an assessment of the seasonal course of the topsoil thermal conditions across the chronosequence, but also, for the determination of snow cover duration and the length of the vegetation period. This article summarizes the vegetation development and microclimatic conditions along the chronosequence and discusses the results considering the interplay between the time since deglaciation and elevation being a microclimatological proxy.

Abstract: Land subsidence is a widespread geological hazard in urban areas, yet accurate prediction remains challenging due to the complex, non-linear spatio-temporal dynamics governing subsurface deformation. In this work, the PEGNet (Peridynamics-Emergent-GAT-GRU Network) method integrates Graph Attention Networks (GAT) with Gated Recurrent Units (GRU) under a three-level embedding strategy to predict land subsidence at monitoring points. Peridynamics gradient and emergent velocity features are fused with multi-source data at the feature level; multi-head graph attention captures spatial dependencies among monitoring points at the structure level; and a Peridynamics-based loss regularizer penalizes unrealistic spatial gradients while an attention entropy regularizer encourages focused attention patterns at the constraint level. Evaluated in Tongzhou District, Beijing over 60 months, the model achieves an RMSE of 5.16 mm, outperforming the GAT-GRU baseline by 25.1% and reducing RMSE by 51.7% relative to GCN-LSTM. Comprehensive ablation studies and baseline comparisons validate each component, showing that the dual-constraint framework effectively suppresses large-magnitude outliers. This work demonstrates that coupling macro-micro priors with spatio-temporal deep learning provides an effective paradigm for integrating physical and geoscience principles into data-driven subsidence prediction.

Abstract: This research aims to analyse the dynamic of land use structure in Fălticeni municipality over a four-decade period, examining the impact of major political, socio-economic and demographic transitions between 1985, 2006 and 2025. The shift from the centrally planned system to a free market economy, the installation of a democratic regime, the subsequent integration of Romania into the European Union, and the major adjustments in migration and natural population growth patterns created a context for largely unregulated urban sprawl, mainly to the detriment of agricultural land and natural landscapes. Moreover, the recent expansion of the city encompassed terrains with varying degrees of geomorphological vulnerability, some of which are affected by active slope processes and increased susceptibility to landsliding. Alongside geodynamic pressures, the urban body has long faced internal functional challenges, including the presence of extensive abandoned industrial perimeters, and a severe deficit of forested land and proper green areas. By correlating the recent urban spatial and functional evolution with physical vulnerability indicators, the study demonstrates the need for integrated planning and seeks to provide a framework for the management of urban growth tailored for medium-sized cities. The proposed solutions are rooted in the results yielded by the analyses of land use dynamics and susceptibility to natural risks, and focus on ecological zoning and controlled expansion on secure terrains; the functional redevelopment of brownfields into residential, commercial and recreational areas; the establishment of peri-urban forests fulfilling multiple functions, including risk mitigation and enhanced benefits for local communities; and correlating the design and execution of essential infrastructure such as the urban bypass with rigorous risk assessments, thus providing a sound methodological approach for sustainable urban planning which can be replicated for cities facing similar challenges.

Abstract: Despite the global crisis of mangrove deforestation, certain sediment-dominated estuaries exhibit remarkable morphodynamic resilience. This study investigates the spatiotemporal trajectory of the Ujung Pangkah estuary (1995–2025) to quantify natural progradation against anthropogenic pressures. Utilizing Google Earth Engine (GEE), Landsat archives, and a Random Forest classifier enhanced with advanced spectral indices (NDVI, mNDWI, EVI, MVI), we rigorously mapped the estuarine landscape. Accuracy assessment using independent historical validation yielded an exceptional overall accuracy of 97.30% for 2025. The change dynamics analysis revealed an explosive natural recovery. While the ecosystem suffered a gross historical loss of 574.02 ha primarily due to aquaculture conversion, this was vastly offset by a massive seaward gain of 1,245.15 ha on accreted mudflats. The total mangrove extent expanded from 613.88 ha in 1995 to 1,285.01 ha in 2025. Non-parametric statistical evaluation confirmed a highly significant, continuous median expansion rate of 27.65 ha/year. Crucially, hydrodynamic driver analysis using a red-band proxy for Total Suspended Solids (TSS) empirically validated that this progradation is intrinsically sediment-driven, fueled by the hyper-concentrated monsoonal discharge of the Bengawan Solo River. Safeguarding these dynamic frontiers requires urgent policy frameworks to legally protect newly accreted estuarine zones from future land-use conversion.

Abstract: Cancer incidence exhibits substantial spatial disparities linked to environmental, behavioral, built-environment, healthcare-access, and socioeconomic conditions, yet the spatial scales at which these relationships operate remain insufficiently understood. This study develops an explainable spatial epidemiology workflow that integrates Random Forest, SHapley Additive exPlanations (SHAP), Ordinary Least Squares (OLS), Geographically Weighted Regression (GWR), and Multiscale Geographically Weighted Regression (MGWR) to examine county-level incidence for all-site, colon, breast, and skin cancers across Texas, USA. Random Forest and SHAP were used to identify outcome-specific nonlinear predictor relevance, and OLS, GWR, and MGWR were used to compare global, local, and multiscale spatial associations before and after RF-SHAP feature screening. MGWR generally achieved higher model fit than OLS and GWR. Before RF-SHAP screening, MGWR R² values were 0.701 for all-site cancer, 0.516 for colon cancer, 0.499 for breast cancer, and 0.694 for skin cancer, compared with OLS R² values of 0.343, 0.338, 0.257, and 0.368. RF-SHAP reduced predictors by about one-half and consistently improved AICc. The results show that environmental exposures, activity-related conditions, transportation access, screening, food insecurity, and chronic health indicators contribute to spatial differences in cancer incidence. The framework links nonlinear machine-learning evidence with spatially explicit interpretation for transferable epidemiological analysis.

Abstract: Population data at small area scales are essential for effective decision-making, influencing public health, disaster response, and resource allocation, amongst others. While national censuses remain the cornerstone of population data, they are often constrained by high costs, infrequent collection cycles, and coverage gaps, which can hinder timely data availability. To address these challenges, geospatial statistical approaches using limited microcensus surveys have been demonstrated as a reliable source, but the field has advanced substantially in recent years, with significant developments in both data sources and modelling methodologies. New approaches now leverage routine health intervention campaign data, satellite-derived settlement maps, and bespoke modelling approaches to produce reliable small area population estimates where enumeration is difficult or outdated. Various countries are applying these techniques to support census operations, health program planning, and humanitarian response. This manuscript reviews recent advances in ‘bottom-up’ population mapping approaches, highlighting innovations in input data, modelling methods, and validation techniques. We examine ongoing challenges, including partial observation of buildings under forest canopy, population displacement, and institutional uptake. Finally, we discuss emerging opportunities to enhance these approaches through better integration with traditional data ecosystems, capacity strengthening, and co-production with national institutions.

Abstract: Monocular 3D lane detection is challenged by the loss of directional detail in distant markings during feature downsampling and by geometric ambiguity between lateral curvature and road elevation under single-view projection. R-A3D addresses both issues through frequency-aware Riemannian anchor learning in an anchor-regression pipeline. A Haar-based feature pyramid retains four frequency subbands before learned channel projection, strengthening thin and low-contrast lane responses. Each evolving 3D anchor is summarized in the lateral–elevation plane by Gaussian mean and covariance statistics, embedded in a symmetric positive-definite matrix, and mapped to a tangent-space feature with the Log-Euclidean metric. This Riemannian anchor feature is fused residually with anchor-aligned visual evidence and recomputed after each cascade stage. On OpenLane, the ResNet-50 model achieves 64.9% F1, 94.1% category accuracy, and far-range lateral and vertical errors of 0.215 m and 0.081 m at 23.5 frames/s; the ResNet-18 model reaches 63.0% F1 at 53.2 frames/s. On ApolloSim Visual Variations, R-A3D achieves 96.1% F1 and reduces the two far-range errors by 31.0% and 30.3% relative to GLane3D. These results demonstrate complementary benefits from frequency-preserving visual evidence and proposal-dependent Riemannian geometry.

Abstract: Compound coastal hazards flooding, land subsidence, storm surge, and salinity intrusion, impose accelerating risks on deltaic communities. No published study has simultaneously mapped all four hazard types using a comparable algorithm diversity under spatially honest validation. This study presents a multi-hazard susceptibility mapping framework applied to Terrebonne Parish, coastal Louisiana. Eight algorithms were compared: XGBoost, Random Forest (RF), Gradient Boosting Machine (GBM), Support Vector Ma-chine (SVM), Multilayer Perceptron (MLP), 1-Dimensional Convolutional Neural Network (1D-CNN), Long Short-Term Memory (LSTM), and a hybrid deep learning architecture (CNN-LSTM); a Spatially Aware Ensemble Meta-Learner (SAEML) trained on spatially separated out-of-fold (OOF) predictions was additionally evaluated. Sixty-five condition-ing factors spanning ten thematic categories were compiled at 30-m resolution; features involved in hazard label generation were excluded to prevent circular reasoning. Model generalizability was assessed through conventional holdout testing and 5-fold spatial block cross-validation (5-km blocks). The 1D-CNN achieved the highest holdout F1-macro for flood susceptibility (0.923), XGBoost for land subsidence (0.919), and GBM for storm surge (0.865) and salinity intrusion (0.765). Spatial cross-validation revealed a 53- mean percentage-point performance degradation (F1-macro: 0.22–0.39 vs. holdout 0.72–0.92), demonstrating that conventional holdout metrics substantially overestimate geographic generalizability. The SAEML’s near-zero holdout-to-spatial-Cross-Validation gap (~0.005) arises by design from its OOF training procedure; under spatially honest comparison, SAEML spatial CV F1-macro (0.310–0.364) fell below the best base-model scores on all four hazards, establishing that ensemble stacking does not automatically improve upon CV-guided individual model selection. Five-class susceptibility maps show that over 81% of Terrebonne Parish is classified as High or Very High flood susceptibility. The composite Multi-Hazard Susceptibility Index (mean = 0.574; range = 0.125–0.938) identifies southern coastal areas as the highest-priority zone for integrated risk reduction, with implications for land-use planning, emergency management, and climate adaptation policy.

Abstract: River avulsion, the sudden relocation of a river channel to a new course from the parent channel, is a geomorphic process with direct implications for floodplain evolution, ecosystem dynamics, and infrastructure vulnerability. This review article discusses how sandbar migration acts as a precursor to avulsion by altering hydraulic geometry, redirecting flow paths, modifying sediment transport patterns, and affecting the development of incipient channels. The morphodynamical evolution of sandbars, influenced by sediment supply, flow regime, vegetation, and anthropogenic influences such as dams and sand mining, plays a central role in creating avulsion. Different methods, such as field measurements, remote sensing imagery (including multispectral, SAR, LiDAR and UAV), physics-based numerical models, machine learning, and deep learning techniques, which are used to evaluate river sandbar and river avulsion, are also thoroughly evaluated for efficacy and fit for purpose.