Meteorological drought is a major climatic stressor that alters soil conditions and vegetation performance, affecting ecosystem sustainability and land management. However, the long-term spatial relationships among meteorological drought exposure vulnerability (MDEV), soil erosion susceptibility (SE), and vegetation health (VHI) remain poorly understood. This study investigates these relationships in the Mayurakshi River Basin, eastern India, using an integrated geospatial framework. MDEV was derived from long-term Standardized Precipitation Index (SPI)-based climatic information through an Analytic Hierarchy Process (AHP)-weighted overlay of eight conditioning factors. Soil erosion susceptibility was estimated using the Revised Universal Soil Loss Equation (RUSLE), while vegetation condition was assessed using the Vegetation Health Index (VHI) for 2010, 2015, and 2024. Pearson correlation, Ordinary Least Squares (OLS), Geographically Weighted Regression (GWR), and Kernel Density Estimation (KDE) were employed to evaluate spatial relationships.Results showed a significant positive association between MDEV and SE (r = 0.26, n = 2000, p < 0.01), indicating that greater drought exposure generally coincided with higher erosion susceptibility. Conversely, MDEV and SE exhibited significant negative relationships with VH (r = –0.21 to –0.45 and r = –0.39 to –0.42, respectively; p < 0.01), suggesting that greater drought exposure and erosion were associated with poorer vegetation condition. KDE and GWR revealed pronounced spatial heterogeneity. Overall, the study demonstrates the long-term spatial co-occurrence of drought exposure, soil erosion susceptibility, and vegetation degradation, providing a geospatial framework for identifying environmental stress hotspots and supporting integrated land, soil, and water resource management under increasing climatic variability.
Effective river restoration is vital to conserve ecosystems and biodiversity under increasing human and climatic pressures. However, the limited recovery of fish communities in many projects raises concerns about their effectiveness, underscoring the need for monitoring to support adaptive management. Recognizing the extensive restoration efforts in the Drawa catchment in Central Europe, we propose a comprehensive framework to assess restoration impacts on aquatic habitats and examine how climate change and anthropogenic pressures have influenced changes in fish communities within side channels between 2010 and 2024. The findings revealed a significant decline in diversity indices, along with a general reduction in total abundance and occupancy of indicator species in the side channels of the Drawa in 2024 compared to 2010, despite intensive restoration efforts. Land use and land cover (LULC) and most anthropogenic pressures in the catchment remained stable, except for human-induced morphological alterations and an increase in beaver dam construction. In contrast, climatic changes, particularly rising temperatures, reducing surface runoff and increased evaporation, have altered hydrological conditions, as evidenced by stream droughts in 2024 and a reduction in water bodies between 2017 and 2024. Therefore, the limited effectiveness of restoration efforts appears to be the result of multiple interacting factors, primarily climate change, beaver activity, and potential pollution indicated by increased conductivity. Overall, restoration can enhance ecological conditions; however, the recovery of fish communities necessitates comprehensive management, including pollution control, rewilding, river remeandering, and the establishment of refugia to address climate change and anthropogenic pressures, thereby safeguarding fish habitats.
Drought stress driven by intra-seasonal rainfall variability and rising evaporative demand constrains maize productivity in rainfed sub-humid systems of sub-Saharan Africa. This study evaluated the effects of Conservation Soil Water Management Practices (CSWMPs), including straw mulch, half-moon pits, and permanent planting basins, on seasonal, phenological, and depth-wise soil moisture dynamics and their implications for crop water use and productivity across three maize growing seasons. Volumetric soil moisture was monitored at 0-40 cm depth using FDR probes and linked to soil water storage, deficit to field capacity, available water, and soil physical properties. CSWMPs consistently maintained significantly higher soil moisture than the control (p ≤ 0.05) including during the critical reproductive stage of tasseling and silking, with half-moon pits showing the greatest retention. Grain yield increased from 2.7-4.9 t ha⁻¹ in the control to 7.0-8.2 t ha⁻¹ under straw mulch and permanent planting basins, while water use efficiency increased from 6-11 to 14-17.6 kg ha⁻¹ mm⁻¹ (p ≤ 0.05). Evapotranspiration remained statistically similar across treatments (p > 0.05), indicating that yield gains resulted from improved partitioning of water into productive transpiration rather than increased water consumption. Mixed-effects modeling identified crop growth stage and soil water storage as dominant controls of soil moisture, with significant phenological interactions. Overall, SWMPs enhanced soil moisture retention, improved soil structural functioning, and strengthened maize resilience by improving apparent water productivity and crop yield under variable climatic conditions.
Seasonal hydrological regulation in the Three Gorges Reservoir (TGR) profoundly alters sediment transport and the behavior of heavy metals in suspended sediments (SS). However, the seasonal variation patterns, key controlling factors, and associated ecological risks of SS-bound heavy metals remain incompletely understood. Four sampling sites were established at the confluence of the Yangtze River mainstem and a typical tributary (Ruxi River) in the Zhongxian reach of the TGR. SS samples were collected over two hydrological years. We measured concentrations of six heavy metals (Cd, Cr, Cu, Ni, Zn, and Pb) to evaluate contamination status and associated ecological risks. Concurrently, principal component analysis and random forest models were applied to identify interrelationships among heavy metals, controlling factors, and potential sources. Results showed that SS exhibited finer particle-sizes and higher nutrient enrichment during the dry season. Heavy metal concentrations displayed significant seasonal variations: in the mainstem, Cd, Cu, Ni, Zn, and Pb were markedly higher during the dry season; in the tributary, similar patterns were observed for Cu, Zn, and Pb. The geo-accumulation index and pollution load index indicated slight to moderate contamination, while the potential ecological risk index showed that ecological risks peaked in the mainstem during the dry season, reaching a moderate risk level with Cd as the dominant contributor. Total organic carbon and silt content were the most influential variables, and seasonal changes modulated Cd, Cu, and Zn variations. Heavy metals were inferred to be mainly associated with industrial wastewater and agricultural activities from multivariate analyses and regional land-use characteristics.
Agricultural land use fundamentally alters stream ecosystems, but previous studies show inconsistent effects on ecosystem respiration (ER), with little attention to autotrophic versus heterotrophic components. This study aimed to analyze how agricultural intensity affects gross primary production (GPP) and the balance between autotrophic respiration (AR) and heterotrophic respiration (HR). We used data from measurements of GPP and ecosystem respiration (ER) using open-water metabolism measurements in 25 lowland streams spanning an agricultural gradient (0-84.5% coverage) in the Nordic region. We separated ER into AR and HR using the Hall and Beaulieu partitioning framework. Contrary to many studies, GPP did not increase with agricultural land cover or nutrients but instead correlated with stream depth and plant coverage. However, the heterotrophic fraction of respiration (HR/ER) increased significantly with agricultural intensity, with agriculture explaining 56% of the variation (HR/ER increased 0.004 per 1% agriculture increase). Heterotrophic respiration comprised up to 94% of total respiration in the most agricultural catchments (median: 84%). This shift occurred despite wide variation in stream characteristics, indicating that catchment land use, not in-stream conditions, drives the metabolic regime. We propose that tile drainage and manure applications increase the quantity and lability of terrestrial organic carbon entering streams, fueling heterotrophic metabolism. This shift has critical implications for CO2 emissions, as heterotrophic respiration represents net conversion of terrestrial carbon to atmospheric CO2. Our findings suggest HR/ER could serve as a functional indicator of agricultural impacts and that agricultural carbon accounting should include stream-mediated CO2 losses.
Ecological water requirements (EWRs) are fundamental for wetland conservation and restoration. However, existing studies often fail to simultaneously capture differences among EWRs levels, seasonal dynamics, and responses to extreme hydrological conditions. To address these limitations, this study developed a remote sensing-based framework for the multi-level assessment of seasonal EWRs and applied it to 27 important wetlands in Northeast China. The results showed that larger wetlands generally exhibited higher ecological water requirements. Regarding functional water allocation, biodiversity conservation accounted for the largest proportion of EWRs (reaching 95–99% in some wetlands), followed by wetland extent maintenance, whereas ecosystem structural stability contributed the smallest share (≤3%). At the annual scale, no significant differences were detected among the three EWR levels (p > 0.05). At the seasonal scale, EWRs followed the order of overwintering period < breeding period < flood season. For all three EWRs levels, significant differences were observed between the overwintering period and both the breeding period and flood season (p < 0.05), whereas no significant difference was found between the breeding period and flood season (p > 0.05). Under extreme hydrological conditions, wet years were generally able to satisfy the maximum EWRs level of most wetlands, whereas dry years could only meet the minimum EWRs level. The proposed framework provides a novel approach for quantifying seasonal EWRs across multiple levels, offering a scientific basis for wetland ecological restoration and adaptive water resources management under climate change.
Tropical urban estuaries are environments of high ecological, socioeconomic and sanitary relevance, acting as transition zones between continental and marine systems and concentrating multiple water uses, tourism, fisheries, urban occupation, wastewater discharges and sanitation-related pressures. Intensified coastal urbanization, hydrological variability, altered discharge regimes and inputs of organic matter, nutrients, fine sediments, suspended solids and fecal-indicator microorganisms have increased the need for tools capable of representing pollutant transport, dispersion, dilution, retention and transformation. In this context, numerical hydrodynamic and water quality modeling has become a strategic approach for understanding estuarine circulation, coastal plume dynamics, water renewal, ecological responses and the effects of pollution sources on environmental quality. This article reviews literature published between 2022 and 2026 on the application of numerical models to assess hydrodynamics, water quality and pollution in tropical urban estuaries and associated coastal environments. Scopus, ScienceDirect, SciELO and Wiley Online Library were consulted, and 40 studies were selected for the final review matrix. The revised analysis expands the discussion of widely used platforms, including Delft3D-WAQ, WASP, EFDC, TELEMAC, FVCOM, SCHISM, MOHID, CE-QUAL-ICM/CE-QUAL-W2, SLIM and MIKE-Eco Lab, and compares their strengths and limitations for estuarine applications. The review highlights that practical reliability is constrained mainly by limited field data for calibration and validation, especially for biogeochemical, microbiological and sediment-related variables. The findings show that modeling is most useful when hydrodynamics, pollutant fate, fine-sediment dynamics, ecological indicators and sanitation scenarios are integrated into monitoring and decision-support frameworks.
Mediterranean rivers combine strong hydrological variability with increasing human pressures, which can significantly affect aquatic insect communities. This study aims to assess the spatiotemporal dynamics of Simuliidae assemblages in the Martil River basin (northwestern Morocco) in relation to physicochemical and hydromorphological gradients, with particular focus on the effects of flow regulation as a key ecohydrological driver. The basin hosts four dams whose impoundment structures alter flow continuity and longitudinal connectivity across the river network. A total of 2,258 larvae and pupal exuviae representing eight species and three genera were collected across 16 sites from summer 2017 to spring 2018. Species richness and abundance were higher in upstream and mid-altitude reaches, whereas downstream sections exhibited reduced diversity and dominance by disturbance-tolerant taxa, particularly Simulium intermedium. Cold-adapted Prosimulium species were confined to cool, oxygen-rich headwaters. Environmental variables followed a longitudinal gradient, with increasing temperature and conductivity and decreasing dissolved oxygen downstream. Multivariate analyses revealed significant differences in assemblage composition between upstream and downstream stations (PERMANOVA: F = 3.18, R² = 0.19, p = 0.042), confirmed by NMDS ordination. Hydroclimatic variability, combined with flow regulation, appears to shape Simuliidae communities through intensified environmental filtering and reduced niche differentiation, promoting tolerant species in altered downstream habitats. These findings underscore the need for maintaining minimum ecological flows and restoring longitudinal connectivity as priority management actions in regulated North African Mediterranean rivers.
The ongoing acceleration of global warming poses an unprecedented threat to freshwater ecosystems. In Central Europe, responses of diverse freshwater fish to climate warming have not been well evaluated. Based on long-term river water temperature data from 59 Polish rivers spanning 1966–2020, this study establishes Poland's first nationwide three-tiered temperature stress index system encompassing optimal growth days, thermal stress days, and critical stress days. It provides a quantitative assessment of the spatio-temporal evolution of thermal habitat quality for cold-water salmonids and warm-water cyprinids. The study of 35 rivers hosting warm-water cyprinids reveals a growing trend in optimal growth days per year, increasing at +0.45 d/yr, with 32 rivers exhibiting significant growth (p < 0.01). Conversely, there is no significant shift in thermal stress or critical stress days annually, suggesting that current warming has primarily extended the favorable growth period, although continued warming may progressively increase exposure to thermal stress of cyprinids. In stark contrast, among 37 rivers inhabited by cold-water salmonids, optimal growth days declined by −0.07 d/yr, with 12 rivers displaying significant decrease (p < 0.01). Concurrently, thermal stress days increased at +0.22 d/yr (p < 0.01) across 32 rivers. Moreover, critical stress days rose significantly in 22 rivers, indicating expanding risk of extreme temperatures. Spatial analysis reveals notable geographical differentiation in heat stress risk for salmonids. The southern montane region, upper Odra River, and northern Baltic Sea coastal basin were identified as hotspots of thermal risk aggregation. This study advances adaptive management recommendations concerning zoning and classification.
Reservoir operation is a major driver of hydrological alteration in regulated rivers and substantially affects fish spawning habitats, particularly during dry-season reproductive periods. This study quantified the impacts of reservoir regulation on the spawning habitat of Procypris rabaudi downstream of the Fu Shou Yan Reservoir in China. An integrated hydrology-hydraulics-habitat framework was developed by coupling a monthly waterbalance calculation of downstream releases, one-dimensional steady-flow hydraulic simulations, and an integral habitat suitability method to estimate weighted usable area (WUA) under three ecological flow release scenarios. Results show that reservoir regulation markedly redistributes seasonal discharge by elevating dryseason flows and attenuating flood-season peaks, resulting in a pronounced compression of intra-annual variability. During the spawning season, regulated releases were substantially less variable than natural inflow and failed to reproduce the natural April discharge increase under all scenarios. Consequently, hydraulic variability in discharge, velocity, and water depth was significantly reduced at four downstream cross-sections, with the strongest alteration near the dam and partial recovery further downstream. Spawning habitat availability increased monotonically from Scenario 1 to Scenario 3 at both spawning grounds. Under the highest release scenario, WUA during February-March exceeded pre-construction levels, indicating effective mitigation of lowflow habitat constraints. However, the natural April peak in spawning habitat was not recovered under any regulated scenario. These results indicate that increasing base releases can enhance early-season spawning habitat but is insufficient to restore peak-season habitat associated with natural high-flow conditions, highlighting the need for seasonally structured and dynamically varying ecological flow strategies.
Few evaluation frameworks investigate the mechanisms causing runoff alterations by quantifying the causes of runoff alterations across different time scales (wet/normal/dry seasons and months) and in-depth analysis of each meteorological indicator's contribution to runoff change. This study quantitatively evaluated the hydrological regime of the Jialing River before and after the abrupt change predicated on the Indicators of Hydrologic Alteration and Range of Variability Approach (IHA-RVA) and the Gini coefficient. Through the differential equation of runoff characteristics, and separated in detail the contribution of each meteorological indicator to runoff alterations. Additionally, the ABCD monthly water balance model (ABCD) expanded and validated the results of the differential equations on the runoff contribution on a time scale. The findings demonstrate that the overall hydrological regime changed moderately in the River (48.63%). Of the 14 meteorological indicators separated by differential equation, the wet season precipitation contributed the most to the runoff alterations, with a contribution rate of -178.12% of the runoff changes driven by all the meteorological indicators, and the coefficient of variation of the annual precipitation contributed the least, with a contribution rate of 2.16%; use ABCD model reconstruction of natural runoff found significant differences in the contribution of drivers at different time scales.
Freshwater-dependent birds are among the most threatened vertebrate groups globally, yet their spatial distribution remains poorly understood in large river basins. The Indian Ganga Basin (IGB), despite being rich in biodiversity, lacks basin-scale assessments of habitat suitability for threatened wetland birds. We conducted this work to map the spatial distribution and habitat suitability of threatened freshwater-dependent bird species across the IGB and evaluate the alignment of suitable habitats with existing protected areas using a digital citizen-science approach. We extracted occurrence records (2021-2025) for 14 threatened birds (4 Endangered, 10 Vulnerable) from the eBird database. We performed species distribution models using Random Forest algorithm with selected bioclimatic and topographic predictors. We classified habitat suitability into low, moderate, and high categories and overlaid with protected-area boundaries. We observed that highly and moderately suitable habitats were concentrated mainly in the north-western and western IGB, particularly the Shivalik hills and the Yamuna-Chambal river system. Collectively, only 24.93% of the basin was highly suitable, while merely 1.67% of this area fell within protected areas, indicating a substantial spatial mismatch between conservation priority and formal protection. Most climatically and topographically suitable habitats for threatened freshwater birds lie outside the current protected-area network and often span multiple states, which may limit the effectiveness of existing conservation frameworks. Integrating citizen science with spatial modelling reveals conservation gaps in the IGB and highlights the need to expand protected areas, strengthening inter-state coordination, and engaging local communities through participatory monitoring to safeguard freshwater avifauna in changing riverine landscapes.
Lake Toba, the world's largest volcanic caldera and a designated UNESCO Global Geopark, is facing significant threats, yet a definitive ichthyofaunal census remains unavailable through traditional methods. This study aimed to assess the fish species composition and status, particularly focusing on the prevalence of non-native species, using environmental DNA (eDNA) metabarcoding. Water samples were collected from five strategic locations in Lake Toba during August-September 2025, and eDNA was analyzed using the MiFish primers targeting the mitochondrial 12S rRNA gene. The eDNA analysis successfully identified 23 fish species, categorizing 19 species as confirmed detection (82.61%) and 4 as unconfirmed taxa, 17.39%). The results revealed a dominance in relative read abundance by 11 introduced species against only 8 native species. The study recorded two new introduced fish species for the lake: Gambusia affinis and Trichopodus pectoralis. Furthermore, the analysis provided taxonomic clarity for the Mystacoleucus genus, confirming the conspecific nature of M. padangensis and M. marginatus. While the high proportion of confirmed detections supports eDNA as a reliable monitoring tool, the occurrence of unconfirmed taxa, including the repeated detection of Rhinogobius similis, highlights the need for careful bioinformatic filtering and further ecological validation. In conclusion, this eDNA study provides the first molecular inventory of Lake Toba's fish fauna, indicating that introduced species may be an important ecological factor influencing native fish populations. These findings highlight the potential need for conservation attention to support declining native fish populations and maintain the biodiversity of this globally significant geopark.
Recreational aquatic environments can serve as a vehicle for the transmission of parasites of human health concern. This study assessed the presence, diversity, and distribution of waterborne parasites in the water and sediments phases from two recreational environments in Salta, Argentina: the Wierna River (WR) and the General Belgrano Reservoir (GB). Eleven sampling campaigns (wet and dry seasons, 2015–2016) yielded 33 water and 33 sediment samples. Samples were processed by ultrafiltration (water) and desorption (sediments), and parasites were identified microscopically. Free-living amoebae, Giardia spp., Cryptosporidium spp., Acanthamoeba spp., hookworm larvae, Ascaris lumbricoides, Taenia or Echinococcus, and Hymenolepis nana were detected. Parasite frequency was higher in GB (100% in water, 89% in sediments) than in WR (60% in both phases). Sediments harbored higher parasite concentrations than water, whereas water exhibited greater species richness. There was no seasonal trend in the findings. Correlation analysis showed some associations between parasites and fecal indicator bacteria or physicochemical parameters, particularly in sediments. The water–sediment partition coefficient indicated that sediments acted as reservoir for most parasite taxa, with sporadic resuspension events likely influencing the detections in water. The pathogenic species, including first records of H. nana and Taenia or Echinococcus in these environments, underscores potential health risk despite fecal indicator bacteria levels met recreational water guidelines. These findings highlight the need to incorporate sediment analysis in monitoring programs to better understand sanitary status and to improve management strategies for recreational waters.
Monsoon-dominated river basins are increasingly exposed to compound climate risks arising from intensified hydrological extremes, environmental degradation, and socio-ecological vulnerability. This study assesses coupled flood hydrology, land degradation, and livelihood vulnerability in the Kulsi River Basin, Assam, India, using an integrated ecohydrological approach. Extreme rainfall and flood discharge were characterized through probabilistic frequency analysis based on 25 years (2000-2024) of annual maximum hydro-meteorological data, while environmental stress was quantified using rainfall erosivity and soil erosion assessment. Livelihood vulnerability was evaluated using an IPCC-based Livelihood Vulnerability Index derived from household-level indicators. The results indicate pronounced escalation of rainfall and flood magnitude with increasing return period, widespread soil erosion affecting approximately one-fifth of the basin, and moderate livelihood vulnerability characterized by high exposure and sensitivity but partially moderated by adaptive capacity driven by flood exposure, agricultural dependence, and limited adaptive capacity. Prolonged flood inundation further amplifies livelihood disruption. Integration of hydrological, environmental, climatic, and socio-economic indicators yielded a Hydro-Livelihood Risk Index (HLRI) value of 0.757, corresponding to the high-risk category and indicating substantial basin-scale vulnerability to compound hydro-climatic hazards. Uncertainty analysis further confirmed the robustness and stability of the estimated risk level. The study demonstrates that flood-induced livelihood risk in monsoon river basins is strongly governed by coupled hydrological and ecological processes rather than hydrological extremes alone. By linking flood dynamics, environmental degradation, and socio-economic vulnerability within a single assessment, the findings provide actionable insights for ecohydrology-informed adaptation and river basin management in data-constrained regions.
Land resources are increasingly degraded worldwide, posing a threat to ecological sustainability and agricultural productivity. Soil erosion is a major global challenge, driven by human activities. This study evaluates the effectiveness of selected best management practices (BMPs) in reducing sediment yield and surface runoff in the Ginchi catchment, Ethiopia. An integrated dataset, including streamflow, sediment yield, satellite imagery, and meteorological data, was used to parameterize and run the SWAT+ model. Eight BMP scenarios were selected based on existing practices and compatibility with site conditions. Model performance was evaluated using NSE, R2, and PBIAS, which indicated satisfactory agreement between simulated and observed values. The results of the baseline scenario indicated a mean sediment yield (40.7 t ha-1yr-1) exceeding the national soil loss tolerance threshold (11 t ha-1 yr-1). All combined BMP scenarios successfully reduced sediment yield below the threshold level and significantly decreased runoff. Among the BMPs, the combination of soil bunds, strip cropping, and reforestation demonstrated superior performance, reducing sediment yield and surface runoff by 87.2% and 22.3%, respectively, compared to the baseline scenario. Reforestation on steep slopes alone reduced sediment and runoff by 44.1% and 3.8%, respectively. Therefore, an integrated BMP implementation is highly effective in regulating sediment and runoff. However, environmental feasibility alone cannot guarantee sustainable land management. It is advisable to conduct an economic feasibility study, and the final implementation should be based on its ecological and economic performance. Policies that promote hillside reforestation and integrated land management should be strengthened to enhance agricultural productivity and ecological sustainability.
Villa Victoria Reservoir (VVR) is one of the most important bodies of water in central Mexico, being part of the Cutzamala System, which supplies drinking water to the population of Mexico City. In addition, fish species are farmed there for food purposes, including Cyprinus carpio, a species that has been stocked in the reservoir for aquaculture purposes and is now well established in this aquatic system. Among the contaminants present in VVR are heavy metals, pesticides, and pharmaceuticals, which previous studies have shown to be neurotoxic to fish. In this study, we evaluated different neurotoxicity biomarkers in juvenile C. carpio. The results showed a significant increase in lipid peroxidation (LPX) and hydroperoxide (HPX) levels, along with enhanced activity of the antioxidant enzymes superoxide dismutase (SOD) and catalase (CAT), indicating oxidative stress. In contrast, protein oxidation (POX) did not show any significant changes. The strongest biological effects were consistently observed at sites 3 and 5. Additionally, marked histological alterations were detected in the brain tissue of exposed fish. Changes were also observed in the expression of genes related to synapses, such as adenosine receptors, the enzyme acetylcholinesterase, and the calcium channels ip3r and ryr; in genes of the apoptotic pathway p53 and bcl2; and in the mt2 gene, which is related to heavy metal detoxification. All of the above were reflected in changes in the freezing parameters in the Novel Tank (NT) and Dark & Light behavioral tests. These behavioral alterations, particularly increased immobility, may reflect stress-related responses that could impair ecological performance under natural conditions. These findings highlight the potential ecological risk associated with exposure to environmental mixtures in the reservoir and underscore the need for further monitoring and risk assessment.
Accurate estimation of the potential evapotranspiration (PET) is crucial for hydrological modeling and water resource management. This study develops and compares two deep learning (DL) models, namely an indirect graph convolutional recurrent network (GCRN) and a direct GraphWaveNet model, to predict daily PET in the Nakdong River basin, South Korea. Reference PET was calculated from 52 years (1973-2024) of meteorological data using the FAO-56 Penman-Monteith (PM) method. The indirect framework predicts variables (net radiation, vapor pressure, and vapor-pressure deficit) prior to PET recomposition, whereas the direct framework predicts PET with reduced inputs on a data-scarce basis. Quantitative evaluation across 13 stations demonstrates that GraphWaveNet outperforms GCRN, with RMSE ranging from 0.52 to 0.68 mm/day and MAE values ranging from 0.40 to 0.50 mm/day, which are 50-65% lower than GCRN's (RMSE values ranging from 1.3 to 1.6 mm/day). The direct model also yielded a higher R-2 (0.85-0.92) than GCRN (0.14-0.48) and had a slight bias (<0.08). GraphWaveNet captured seasonal and short-term variations while maintaining variance; however, the indirect GCRN was underestimated under high-PET conditions due to errors propagated through the reconstructed PM components. Findings confirm that adaptive spatiotemporal graph-based DL has a robust, precise, and physically coherent approach to PET estimation.
