The intersection of the Xiaojiang Fault and the Red River Fault at the southeastern margin of the Tibetan Plateau experiences intense tectonic activity, where repeated earthquakes cause variations in thermal spring hydrochemistry. This study applies Bayesian change point analysis and develops a multicomponent synergistic anomaly detection model, using monitoring data from the Qujiang (5 years, 2019–2024) and Wana (2.5 years, 2021–2024) springs in this region to facilitate the real-time forecasting of the timing for M≥4⁺, Ca²⁺, Cl^−, SO${}_{\mathrm{4}}^{\mathrm{2}-}$, δD, and δ¹⁸O, with TS ≥0.50. These components can serve as sensitive indicators for strong earthquake forecasting. The multicomponent synergistic alarm mechanism overcomes the limitations of single-parameter methods, where the number of hydrochemical components with synchronous anomalies serves as a reliable criterion for forecasting. A higher count of anomalous components typically correlates with larger earthquake magnitudes or shorter epicentral distances. This model has the potential to be applied to thermal spring monitoring across diverse active tectonic regions through targeted parameter optimisation, offering a valuable reference for earthquake forecasting.

To improve hydrological uncertainty estimation, recent studies have explored machine learning (ML)-based post-processing approaches that enable both enhanced predictive performance and hydrologically informed probabilistic streamflow predictions. Among these, random forests (RF) and their probabilistic extension, quantile random forests (QRF), are increasingly used for their balance between interpretability and performance. However, the application of QRF in regional post-processing settings remains unexplored. In this study, we develop a hydrologically informed QRF post-processor trained in a multi-site setting and compare its performance against a locally (at-site) trained QRF using probabilistic evaluation metrics. The QRF framework leverages simulations and state variables from the GR6J process-based hydrological model, along with readily available catchment descriptors, to predict daily streamflow uncertainty. Our results show that the regional QRF approach is beneficial for hydrological uncertainty estimation, particularly in catchments where local information is insufficient. The findings highlight that multi-site learning enables effective information transfer across hydrologically similar catchments and is especially advantageous for high-flow events. However, the selection of appropriate catchment descriptors is critical to achieving these benefits.

Effective discharge forecasts are essential in operational hydrology. The accuracy of such forecasts, particularly in short lead times, is generally increased through the integration of recent measurements of observed discharge; commonly known as discharge assimilation (DA). Recent studies have demonstrated the effectiveness of deep learning (DL) approaches for rainfall–runoff (RR) modeling, particularly Long Short-Term Memory (LSTM) networks, outperforming traditional approaches. However, most of these studies do not include DA procedures, which may limit their operational forecast performance. This study suggests and evaluates three DA strategies that incorporate discharge from either recent discharge measurements or forecasts from a pre-trained rainfall–runoff model. The proposed strategies, based on a Multilayer Perceptron (MLP) as orchestrator, include: (1) the integration of recently observed discharges, (2) the integration of both recent discharge observations and pre-trained model forecasts, and (3) the post-processing of model forecast errors. Experiments are implemented using two large datasets, CAMELS-US and CAMELS-FR, and two established benchmark models (BM): the trained LSTM model from Kratzert et al. (2019) and the conceptual Sacramento Soil Moisture Accounting (SAC-SMA) model from Newman et al. (2017), covering both deep learning and conceptual RR simulation approaches. The considered lead times range from 1 to 7 d, covering both short- and mid-term horizons. The approaches are evaluated within two forecast frameworks: (1) perfect meteorological forecasts over the forecasting lead time and (2) ensemble meteorological forecasts. The two frameworks yield contrasting outcomes. When evaluated under the perfect forecast framework, the application of DA leads to substantial improvements in forecast performance, although the magnitude of these gains depends on the initial performance of the benchmark models and the forecasting lead time. Improvements are consistently significant for the SAC-SMA cases, while for the LSTM cases, gains are observed mainly for basins where the LSTM initially underperforms. However, the ensemble forecast evaluation yields unexpected results: the performance ranking of the tested models changes markedly compared to the perfect forecast framework. The LSTM model, in particular, appears penalized by the under-dispersion of its forecast ensembles. Although this underdispersion could be partly attributable to the underdispersion of the forecast archives tested, it persists even when the model is driven by the high spread climatology-based ensemble. This finding underscores the importance of ensuring reliable ensemble dispersion for the efficient operational deployment of AI-based hydrological forecasts.

Future hydrological droughts in reservoir-regulated regions remain uncertain due to the complex interactions between climate change and reservoir operation. Existing studies usually rely on simplified empirical representations of historical reservoir operations and rarely consider the role of optimal reservoir operation policies. Here, we used the upper Hanjiang River basin (UHRB) in China as a case study to project its future hydrological drought evolution using standard streamflow indices (i.e., SSI-1, SSI-3, and SSI-12) and to quantify the roles of climate change and reservoir operation. A long short-term memory (LSTM)-based hydrological model, coupled with a physics-informed LSTM reservoir model, was developed and driven by bias-corrected climate outputs from five global climate models to project future drought conditions under three scenarios (SSP126, SSP370, and SSP585). The results indicate that future climate change over the UHRB is projected to reduce natural streamflow and exacerbate hydrological droughts, with the most severe impacts projected in the far-future period (2071–2100) under SSP585. The traditional Ankang Reservoir operation reduces the frequency, duration and severity of short-term hydrological droughts (SSI-1 and SSI-3) under all scenarios, but shows limited effectiveness for long-term droughts (SSI-12). Importantly, optimal reservoir operating policies that aim to maximize hydropower generation and power generation guarantee rate reveal clear trade-offs between hydrological drought risk and hydropower benefits, thereby underscoring the importance of enhancing reservoir operation strategies for future drought management in reservoir-regulated basins.

Identifying the driver(s) of a process or phenomenon is central to understanding and predicting its future state. In complex hydrometeorological systems, a process can have multiple drivers dynamically coupled to the system across timescales. Thus, a robust method to identify drivers is imperative. In hydrological sciences, methods like multivariate regression and, more recently, Big Data machine-learning approaches rely on finding a co-relation between variables, rather than identifying cause-effect relations. This study evaluates cause-effect discovery (Causal Discovery or CD) algorithms in hydrometeorological systems. Although earlier studies have made important contributions to exploring CD methods, they have primarily focused on bivariate methods in simple synthetic environments. Specifically, we evaluate the following four theoretically distinct multivariate CD algorithms, (i) TCDF, (ii) VARLiNGAM, (iii) PCMCI+, and (iv) DYNOTEARS. We evaluate these algorithms within a large, complex simulated environment of the Global Land Data Assimilation System (GLDAS) where the drivers, reference truth, are known perfectly. We evaluate the drivers identified by CD methods against this reference truth and also contrast its results with the widely used method of co-relation identification, Pearson’s Correlation Coefficient (PCC). The results show that CD methods identify fewer false drivers compared to PCC, across a range of Köppen-Geiger climate types. For example, PCC failed to distinguish true drivers from instantaneous and lagged cross-correlations, typically present in hydrometeorological systems. Whereas, CD methods eliminate a higher number of false instantaneous and lagged drivers. Thus, though PCC identifies the highest number of true drivers, it suffers from high false drivers. Overall, CD methods perform similar to or better than PCC, while PCMCI+ and DYNOTEARS performed the best. Further, we test whether time-series prediction models perform better when predictors are limited to those identified as causal by CD methods. Evaluation of surface soil moisture predictions during drought shows that CD-based models outperform PCC-based models and are more parsimonious. Thus, we demonstrate the effectiveness of using causal discovery to eliminate spurious relations and obtain a robust set of drivers for prediction and process understanding across different climate conditions. This study overviews, demonstrates and tests efficacy of CD methods in studying cause-effect relations in hydrometeorological systems. By exposing their capabilities and differences in a simulated environment, we hope to encourage their use in the real world and move beyond co-relation.

Large-domain hydrologic modeling studies are becoming increasingly common. The evaluation of the resulting models is however often limited to the use of aggregated performance scores that show where model accuracy is higher and lower. Moreover, the inherent uncertainty in such scores (i.e., the sampling uncertainty), stemming from the choice of time periods used for their calculation, often remains unaccounted for. Here we use a collection of simple benchmarks whilst accounting for this sampling uncertainty to provide context for the performance scores of a large-domain hydrologic model. These benchmarks are simple ways of predicting the variable of interest and include, for example, the long-term daily mean flow, daily precipitation scaled by the average rainfall-runoff ratio, and a basic 2-parameter model that represents a catchment's diffusive response to precipitation inputs. Our test case consists of simulations from the National Water Model v3.0 for approximately 4900 basins across the United States. The benchmarks suggest that there are considerable constraints on the model's performance in approximately one-third of the basins used for model calibration and in approximately half of the basins where model parameters are regionalized. Sampling uncertainty has limited impact: in most basins the model is either clearly better or worse than the benchmarks, though numerous cases remain where sampling uncertainty makes it difficult to clearly distinguish between model and benchmark performance. The areas where the benchmarks outperform the model only partially overlap with areas where the model achieves lower performance scores, and this suggests that improvements may be possible in more regions than a first glance at model performance values may indicate. A key advantage of using these benchmarks is that they are easy and fast to compute, particularly compared to the cost of configuring and running the model. This makes benchmarking a valuable tool that can complement more detailed model evaluation techniques by quickly identifying areas that should be investigated more thoroughly.

Numerous gridded precipitation (P) datasets have been developed to address a variety of needs and challenges. However, selecting the most suitable and reliable dataset remains difficult for users. We conducted the most comprehensive global evaluation to date of gridded (sub-)daily P<10&#8201;000&#8201;km²) worldwide, using each PPPP+&#8201;SM2RAIN) dataset performed best (median KGE of 0.64). The Global Data Assimilation System (GDAS) analysis ranked highest among the (re)analyses (median KGE of 0.72), slightly outperforming the widely used European Centre for Medium-range Weather Forecasts ReAnalysis&#160;5 (ERA5; median KGE of 0.71). Performance varied across K&#246;ppen-Geiger climate zones, with the highest scores in polar (E) regions (median KGE of 0.76 across datasets) and the lowest in arid (B) regions (median KGE of 0.53 across datasets). Spatial correlation analysis between catchment attributes and KGE scores identified aridity index, potential evaporation, and P Copernicus Electronic Production Support Office 2026-06-03T10:47:22+02:00 2026-06-03T10:47:22+02:00 https://doi.org/10.5194/hess-30-3425-2026

Understanding how permanent gullies regulate the transport of dissolved ammonium (NH${}_{\mathrm{4}}^{+}$), nitrate (NO${}_{\mathrm{3}}^{-}$), and phosphorus (P) in runoff delivered from agricultural hillslopes under different rainfall types is essential for controlling non-point source pollution in agroecosystems. In this study, we selected two agricultural catchments, each containing a single permanent gully, and monitored runoff at the gully head and the gully outlet during the rainy seasons of 2022 and 2023. Runoff samples were filtered through 0.45&#8201;&#181;m membrane filters and analyzed for dissolved NH${}_{\mathrm{4}}^{+}$, NO${}_{\mathrm{3}}^{-}$, and P concentrations, and the corresponding nutrient transport fluxes were then calculated. Based on event-scale rainfall characteristics, including rainfall depth, duration, average intensity, maximum 30&#8201;min intensity, and erosivity, rainfall events were classified using the k-means method to examine how different rainfall types influenced the role of gullies in the transport of dissolved NH${}_{\mathrm{4}}^{+}$, NO${}_{\mathrm{3}}^{-}$, and P. The results showed that: (1) Gullies significantly enhanced runoff generation, contributing 36.1&#8201;% of total runoff despite occupying only 12.4&#8201;% of the catchment area. This contribution varied across rainfall types (Type A: frequent, low-depth, low-erosivity; Type B: short-duration, high-intensity; Type C: long-duration, high-erosivity) and was highest under Type A (43.2&#8201;%) and lowest under Type C (33.8&#8201;%). (2) Gullies exerted a pronounced dilution effect on dissolved NH${}_{\mathrm{4}}^{+}$, NO${}_{\mathrm{3}}^{-}$, and P concentrations, particularly on dissolved NO${}_{\mathrm{3}}^{-}$${}_{\mathrm{4}}^{+}$, NO${}_{\mathrm{3}}^{-}$, and P transport fluxes was lower than that to runoff volume, accounting for 31.4&#8201;%, 22.4&#8201;%, and 31.1&#8201;% of dissolved NH${}_{\mathrm{4}}^{+}$, NO${}_{\mathrm{3}}^{-}$, and P transport fluxes at the outlet, respectively. (3) Type C rainfall dominated the transport of dissolved NH${}_{\mathrm{4}}^{+}$, NO${}_{\mathrm{3}}^{-}$, and P. Only 10.2&#8201;% of events contributed over 68&#8201;% of dissolved NH${}_{\mathrm{4}}^{+}$, NO${}_{\mathrm{3}}^{-}$, and P transport fluxes at the catchment scale and markedly increased their transport sensitivity to rainfall compared to Type A and Type B. These sensitivities were also intensified by gullies. These findings highlight the importance of prioritizing permanent gullies and high-erosivity rainfall events in strategies to reduce dissolved nutrient losses from agricultural catchments.

Historical estimates of seasonal snow mass are key to understanding snowmelt-driven streamflow and climate change impacts on mountain water resources. However, direct observations of snow mass are sparse in space and time, forcing most reconstructions to rely on snow models driven by uncertain meteorological inputs. While ground-based and satellite snow observations are commonly used to constrain these models, their potential is limited in data-scarce regions and before the onset of satellite monitoring. Here, we investigate the potential of streamflow observations as an additional source of information to improve historical snow mass reconstructions. We introduce an inverse hydrological modeling framework that selects realistic snow mass realizations based on the accuracy of their streamflow response. Before real-world application, we test the framework in two synthetic experiments. Our results demonstrate that streamflow has the potential to constrain snow mass reconstructions, but that non-uniqueness in the snow-streamflow relationship and uncertainties in the inverse modeling chain can easily stand in the way. We also show that streamflow is most helpful in constraining catchment-aggregated properties of snow mass reconstructions, in particular catchment-aggregated melt rates. Future work should assess the potential of streamflow to constrain snow mass reconstruction under real-world conditions and investigate the added value of streamflow when combined with other snow data sources.

Recent intensive research on the soil&#8211;plant&#8211;atmosphere continuum has introduced novel methodological approaches. These include new in-situ extraction techniques and the application of stable hydrogen and oxygen isotopes in&#160;water enabling to trace water movement and plant responses at much finer spatial and temporal scales. Such approaches provide detailed insights into soil water dynamics and plant adaptation to changing environmental conditions under climate change. This study aims to provide a comprehensive characterization of dynamics of distinct soil water pools &#8211; mobile versus tightly bound water &#8211; along an elevation gradient, while simultaneously assessing the impact of the absent snow accumulation in lowland areas on soil water distribution compared to higher elevations. In contrast to conventional bulk water sampling, a&#160;key innovation of this study lies in the experimental design across the elevation gradient combined with a novel extraction method that selectively isolates tightly bound soil water for isotopic analysis. The results indicate a prolonged residence time of winter-derived soil water in lowland sites, in contrast to a rapid turnover at the highest elevation, where the winter water signal dissipates shortly after snowmelt. Distinct isotopic compositions among water pools &#8211; mobile versus tightly bound water &#8211; were particularly evident in lowland areas at the edges of the growing season (up to 3&#8201;&#8240; and 21&#8201;&#8240; for &#948;¹⁸O&#948;²H, respectively), while tightly bound and bulk soil water exhibited &#8211; on average &#8211; only minor or no isotopic differences. In&#160;the&#160;context of the projected continued decline in snow cover at higher elevations in Central Europe, these findings are&#160;critical for improving predictions of soil water storage and, consequently, plant water availability under ongoing climate change.

Natural and pristine ecosystems, such as wetlands, are being either directly or indirectly threatened by a multiplicity of drivers, which include anthropogenic activities and the impacts they have on the use of natural resources. Strategies oriented to a sustainable management of natural resources (in particular, water) are therefore urgently needed, considering also the increasing effects of climate change. Despite their ecological importance, wetlands remain underrepresented in hydrological modelling studies, especially regarding their specific water needs under changing environmental conditions and different scenarios. This study aims to estimate the water requirements of a temporary wetland through a simple hydrological balance model, ultimately facilitating the identification of strategies for its long-term sustainable management. The pilot case study is the Do&#241;ana National Park, SW Spain, one of the case studies of the European project LENSES (PRIMA Call 2020). The model (&#8220;WetMAT&#8221;) is calibrated and validated using historical time series of key hydrological variables (Maximum Flooded AreaHydroperiod) taken from the literature, to describe the hydrological processes in the wetland. The model is then used for a scenario analysis focused on the assessment of climate change impacts on the state of the wetland and for assessing the ecological water demand of the wetland in a dynamic way, helping to quantify the water needs of such a fragile ecosystem. The results highlight the urgency and importance of developing tools that can help integrating environmental needs into water resources planning and management.

Evapotranspiration (ET) is a key hydrological and meteorological variable, serving as the critical nexus between water and energy exchanges. However, accurate estimation of global ET remains a challenging task, as process-based ET algorithms are often inadequate to capture the nonlinear relationship among environmental factors, and the application of data-driven ET algorithms is hindered by sparse and uncertain ET observations. In this study, we developed a novel ensemble framework that integrates three existing ET models (process-based algorithm, machine learning-based ET model, and hybrid model), aiming to provide high-precision terrestrial ET estimates. The framework is guided by an additional classifier that can achieve dynamic per-pixel model selection, thus fully utilizing the spatiotemporal dynamics of each model's distinct advantages in mapping global ET and avoiding the typical underestimation of high values by ensemble methods. Comprehensive validation of the model was carried out using in situ ET observations from the FLUXNET2015 dataset, catchment-scale water balance ET dataset, and six global-scale ET products, including comparisons to individual base models and another Attention-Based ensemble model. The quantitative comparisons across statistical metrics (RMSE, MAE, R², KGE) indicate that our ensemble model outperforms other evaluated models, especially in extreme samples. Meanwhile, the introduction of classifier can not only significantly enhance the algorithmic robustness and generalizability, but also allow us to gain a basic understanding of the mechanisms behind model selection by interpretability analysis. The study demonstrated the effectiveness of the proposed framework in enhancing ET estimation robustness, thereby providing a valuable reference for the estimation of other similar variables.

Throughfall in forests is spatially highly heterogeneous creating distinct patterns that persist over time and propagate into the soil. Despite its importance for forest ecohydrological processes, experimentally derived high-quality datasets describing spatio-temporal throughfall dynamics at fine temporal and spatial resolution are still scarce. The majority of studies were unable to measure throughfall at high temporal and/or spatial resolution because of extensive sampling efforts, especially in forests with complex structures. We present a novel, innovative and modular throughfall monitoring system for continuous, automated measurement of throughfall either as isolated canopy throughfall and as integrated throughfall (total throughfall reduced by litter interception). Without removing the water, the system allows to quantify the spatio-temporal throughfall variability at both intra-event and intra-stand levels. The network captures spatial throughfall patterns and their temporal persistence across rainfall events of varying size during leafed and non-leafed periods. The throughfall monitoring network features 60 self-built, cost effective throughfall samplers, with four throughfall collection compartments and tipping bucket units each connected to a newly developed microcontroller board enabling fully automated, low-maintenance operation during rainfall events. The network, collecting data since the winter of 2024/2025, is setup in a stratified sampling pattern among four forest plots of Beech, Douglas fir, Silver fir, and mixed trees in a mature temperate forest in Germany. Data from a four-week observation period in the spring of 2025 are included in this study to showcase the potential of this approach. The data support the networks' ability to capture small-range spatio-temporal throughfall patterns across the study area.

Terrestrial water storage anomalies (TWSA), jointly influenced by climatic variability and human activities, exhibits pronounced fluctuations across multiple temporal scales. A substantial portion of the fluctuations is attributed to climatic variability, like the El Ni&#241;o&#8211;Southern Oscillation (ENSO). Empirical reconstruction of climate-driven water storage based on relationships between GRACE satellite gravity observations and meteorological forcing data has become a common approach; however, existing models often neglect the regulating role of temperature in the transformation of precipitation into water storage. In this study, we propose a linear, four-parameter coupled recursive model that explicitly incorporates temperature effects on both the conversion and dissipation efficiency of water storage. Using GRACE/GRACE-FO satellite observations and meteorological forcing data, we reconstructed climate-driven TWSA over the global land grid (excluding Antarctica) at a monthly temporal resolution and 0.5&#176; spatial resolution for the period 2002 to 2021. For 116 major global river basins, we further derived basin-scale TWSA reconstructions and quantitatively evaluated the fraction of precipitation converted into TWSA. Compared with existing statistical reconstruction products, the results indicate that: (1) the proposed method achieves substantially faster parameter convergence, improving computational efficiency by several tens of times during the TWSA reconstruction process; (2) the proposed model demonstrates superior performance in approximately 89&#8201;% of river basins and 62&#8201;% of global land grid cells. Additional comparisons with the physically based Catchment Land Surface Model (CLSM) product from NASA's Global Land Data Assimilation System (GLDAS) show that the proposed method better captures the temporal variability of GRACE TWSA in most basins. At the daily scale, the reconstructed TWSA agrees well overall with the ITSG-Grace2018 daily solution and GLDAS-2.2. This study enhances the understanding of the mechanisms governing terrestrial water storage variations at both global and regional scales, provides a quantitative assessment of climate-driven water storage changes, and offers a solid foundation for disentangling the respective impacts of climatic variability and human activities on water resources.

Canopy interception alters the kinetic energy of raindrops reaching the ground, which has important implications for soil erosion, water conservation, and ecosystem functioning. A novel estimation model for the kinetic energy of rainfall under the canopy is developed by stratifying the canopy using parameters such as leaf area index and leaf inclination angle, explicitly distinguishing between canopy-dripped and splashed raindrops. The efficacy of the model is subsequently assessed and analyzed through a comprehensive examination of nine field datasets encompassing LiDAR and raindrop spectrum observations. The simulated under-canopy total kinetic energy, splashing drop kinetic energy, and dripping drop kinetic energy showed total R²$\mathrm{J}{\mathrm{m}}^{-\mathrm{2}}{\mathrm{h}}^{-\mathrm{1}}$, with measurement including uncertainty of 54.1&#8201;&#177;&#8201;12.4, 3.7&#8201;&#177;&#8201;0.150.4&#8201;&#177;&#8201;12.4&#8201;$\mathrm{J}{\mathrm{m}}^{-\mathrm{2}}{\mathrm{h}}^{-\mathrm{1}}$, respectively. Simulations indicate that the under-canopy raindrop spectrum and kinetic energy are primarily controlled by canopy structure and vary less than above-canopy rainfall properties across the observed events. Sensitivity analysis shows that the model is generally robust, with rainfall intensity, the pinning proportion coefficient, LAI and surface contact angle exerting the greatest influence, while other factors have limited impact. Remaining limitations, including simplified branch-drip representation, component-partitioning assumptions and measurement uncertainties, highlight the need for improved parameterization and broader observations.

Arctic rivers represent important components of the Arctic and global hydrological and climate systems, serving as dynamic conduits between terrestrial and marine environments in some rapidly changing regions. They transport freshwater, sediments, nutrients, and carbon from vast watersheds to the Arctic Ocean and affect ocean circulation patterns and regional climate dynamics. Despite their importance, modeling Arctic rivers remains challenging because of sparse data networks, unique cryospheric dynamics, and complex responses to hydrometeorological variables. In this study, a novel hybrid deep learning model is developed to address these challenges and predict Arctic river discharge by incorporating Kolmogorov&#8211;Arnold Networks (KAN), Long Short-Term Memory, and the attention mechanism with seasonal trigonometry encoding and physics-based constraints. It integrates several novel components: (1)&#160;A KAN-based deep learning component learns and captures intricate temporal patterns from nonlinear hydrometeorological data; (2)&#160;Explicit physical constraints designed for the characteristics of permafrost-dominated watersheds govern snow accumulation and melt processes through the architectural design and loss function; (3)&#160;Seasonal variations are accounted for using trigonometry functions to represent cyclical patterns; (4)&#160;A residual compensation structure allows the proposed model to revisit systematic errors in initial predictions and helps capture complex nonlinear processes that are not fully represented. The Kolyma River, which is dominated by permafrost, is adopted to test the performance of the newly developed model. It obtains more robust and accurate predictive performance compared to baseline models. The role of physical constraints, the residual compensated architecture, and the trigonometry encoding are assessed by ablation analysis. The results indicate that these components improve the predictive performance. This novel approach offers a new pathway for addressing key challenges of hydrological forecasting in cold, permafrost-dominated regions and provides a robust framework for improving Arctic river discharge prediction.

Agricultural areas often experience green water scarcity &#8211; the limitation of crop growth by soil moisture supplied through rainfall and snowmelt &#8211; due to e.g. unfavourable soil texture, high potential evapotranspiration rates, poor or inefficient crop management, and fluctuations in meteorological conditions. Driven by the growing effects of climate change and the rising water and food demands of an increasing world population, agricultural green water scarcity is becoming an increasingly important phenomenon. In this global modelling study, a plant-physiology based indicator of green water stress is applied, that quantifies the ratio between soil moisture limitation and atmospheric water demand on agricultural areas. Results show that currently (2015&#8211;2019 average) 44&#8201;% (686&#8201;Mha) of the global agricultural area is green water stressed. Hotspots are characterized by a high seasonal variability in stress conditions, and are mainly located in India and Pakistan, northern Sub-Saharan Africa, North Africa and southwestern Asia. Using an analogous blue water stress indicator &#8211; which relates human water use for households, industry and agriculture to available blue water resources &#8211; current irrigation is shown to alleviate plant water stress in agricultural areas by compensating for green water scarcity on 8&#8201;% (140&#8201;Mha) of the total agricultural area, but simultaneously increases the share of areas experiencing blue water stress by 6&#8201;% (96&#8201;Mha). This shift in water stress types highlights the importance of jointly considering the interconnected green and blue water resources and stresses in pathways towards sustainable water use in agriculture.