Glacial meltwater constitutes a vital component of the water supply in arid and semi-arid areas. However, the influence of glacial melting on runoff and evapotranspiration under global warming remains insufficiently understood. Previous studies coupling the Soil and Water Assessment Tool (SWAT) model with glacier modules often failed to consider the spatial heterogeneity of temperature during glacial melting, potentially leading to biased estimates of meltwater volume. In this study, we developed a glacier-coupled SWAT (SWAT-glacier) model considering the digital elevation model (DEM) based temperature-driven glacial melt processes to elucidate the impact of glacial melting on hydrological processes across four river basins (Dongda, Xiying, Jinta, and Zamu) of the upper Shiyang River Basin (SYRB) in northwestern China from 1986 to 2021. Compared with the standard SWAT model, the proposed SWAT-glacier model significantly improved the simulation accuracy for both runoff and evapotranspiration. Specifically, in comparison with the standard SWAT model, the Nash-Sutcliffe efficiency of the SWAT-glacier model showed a relative improvement of approximately 0.42%-9.16% and 1.50%-10.15% for runoff and evapotranspiration, respectively, in the four river basins during the validation period. Annual glacial runoff occurred predominantly from May to October, whereas glacial melt-induced evapotranspiration peaked between June and August. From 1986 to 2021, the average contributions of glacial melt to runoff were 6.97% for Dongda, 3.06% for Xiying, 2.70% for Jinta, and 0.67% for Zamu, whereas its contributions to evapotranspiration were 9.06%, 5.14%, 3.21%, and 1.59%, respectively. This study presents a SWAT-glacier modeling framework that enhances the simulation of hydrological processes in cold regions. The proposed methodology can be extended to other glacierized basins to provide valuable insights into water resource management under climate change.

Water networks connect rivers, lakes, diversion projects, and storage works, and their scheduling-allocation must address structural and functional complexity. This study develops a water-society-ecology synergy-based integrated water-network scheduling-allocation (WSE-IWNSA) model and proposes a multi-objective evolutionary algorithm-ε-constraint-mathematical programming solver framework (MOEA-ε-MP) for large-scale multi-objective problems. Algorithmic performance of the framework is evaluated on two test problems by comparing two hybrid variants, H-Gurobi and H-COPT, with two standalone MOEAs. Practical applicability is further examined in the Shaanxi case using H-Gurobi. In the test problems, MOEA-ε-MP shows faster observed convergence than standalone MOEAs; more objectives tend to slow convergence and increase fluctuations; and H-Gurobi/H-COPT generally achieve better observed hypervolume, inverted generational distance, and inverted generational distance plus. In the case study, the WSE-IWNSA model delivers simultaneous improvements in water supply and ecological satisfaction. Overall, the proposed WSE-IWNSA model and MOEA-ε-MP framework provide an effective approach for large-scale multi-objective optimization of complex water networks.

Seasonal drought prediction at the basin scale remains constrained by the limited skill of dynamical and statistical methods at 1–6 month lead times—the horizon most critical for water resources operations. We examine atmospheric moist static energy (MSE) anomalies as drought precursors over the Han River basin, the water source area of China's South-to-North Water Diversion (Middle Route). MSE is decomposed into sensible heat, latent energy, and geopotential components. Using ERA5 monthly reanalysis at nine pressure levels (1981–2024), we construct 81 candidate features (3 components × 9 levels × 3 temporal variants). An L1-regularized logistic regression model is trained under leave-one-event-out cross-validation to predict basin-scale drought onset within a six-month window. The latent energy component dominates the retained feature set (43% of 42 non-zero coefficients), with the highest coefficient magnitudes and greatest cross-validation stability. This precursor signal concentrates at approximately 500 hPa across all three basin sub-regions. Episode-based verification yields a probability of detection of 0.946, a false alarm ratio of 0.054, and a critical success index of 0.897. The median pre-onset alarm episode spans 3.0 months, and the event-aligned probability trajectory first exceeds the discriminant threshold at t = -3. These results indicate that mid-tropospheric latent energy anomalies carry statistically robust drought-predictive information at seasonal lead times. The MSE decomposition framework thus offers a physically interpretable reference for drought early warning in monsoon-influenced water-source basins.

Accurate runoff forecasting helps mitigate flooding and drought risks and ensure water security under changing conditions. Compared to deterministic prediction models, interval prediction can more effectively quantify uncertainty, enhancing practical applicability. However, the Mixture Density Network (MDN) model-a state-of-the-art probabilistic modeling approach in hydrology-is susceptible to bias from distributional misspecification, and its prediction intervals are often overly wide, reducing practical utility. We therefore innovatively incorporated the Weighted Conformal Inference (WCI) strategy, which accounts for distributional shifts in runoff sequences, and integrated it with MDN to develop the WCI-MDN model for runoff interval prediction. To validate the effectiveness of the WCI strategy, we constructed six models in total: MDNs and WCI-MDNs under three distributions-Gaussian Mixture (GMM), Laplace Mixture (LMM), and Countable Mixtures of Asymmetric Laplacians (CMAL)-and evaluated their accuracy and robustness using data from 222 basins in the CAMELS-AUS data set. Results indicated that among the three MDN models, the LMM distribution achieved the best interval prediction performance, followed by the CMAL and GMM distributions. After introducing the WCI strategy, the coverage width-based criterion (CWC) for GMM, LMM, and CMAL distributions decreased by approximately 61.1%, 48.7%, and 54.3%, respectively, across all basins, demonstrating that the WCI-MDNs achieved higher prediction reliability. Furthermore, compared to the MDNs, the standard deviation of the CWC for the WCI-MDNs was reduced by 66.7%-81.8%, indicating higher robustness. Thus, the study improved the existing MDNs, providing a promising new approach for runoff interval prediction.

The increase in frequency, severity, and destructiveness of droughts under global climate change, especially of megadroughts, has devastating impacts on agricultural production, economic development, and ecological protection. However, critical gaps remain in accurately identifying and quantifying megadroughts, which significantly impede effective preparedness, timely response strategies, and long-term mitigation efforts. Therefore, it is imperative to develop a robust method for identifying megadroughts in order to improve management strategies and enhance predictive capabilities for mitigating their severe impacts. Here, the standardized moisture anomaly index (SZI) was leveraged to monitor droughts in Shaanxi Province, China. Then, drought characteristics, namely duration (D), severity (S), and intensity (I), were extracted using the three-threshold run theory. Accordingly, the drought-affected area (A) was accurately quantified. The copula function was applied to analyze the three-dimensional joint return period of the D-S-A relationship. An identification method for megadroughts was developed, based on the functional relationship between the joint return periods and drought loss rate. The identification criteria for megadroughts were listed as follows: the drought characteristic values corresponding to the joint return period (Tor) were D, or S or A exceeding 12.5 months, or 8.5, or 21 × 104 km2, respectively; and for the co-occurrence return period (Tand), the criteria for D, S, and A were greater than 16.5 months, 5, and 20 × 104 km2, respectively. For the period 1961–2022 in Shaanxi Province, the Tor-based and Tand-based criteria identified seven and one megadroughts, respectively. These findings provide an effective reference for future identification and early warning of megadroughts.

Drought exerts catastrophic impacts on agricultural production, natural ecosystems, and social stability. Clarifying the spatiotemporal dynamics and causal mechanisms of drought risk is therefore critical for disaster prevention and mitigation. However, current drought risk assessment predominantly focuses on risk mapping and lacks in-depth causal analysis, resulting in weak linkages between risk assessment outcomes and adaptive strategy formulation. This study conducted a dynamic drought risk assessment by integrating hazard, exposure, vulnerability, and adaptation within a new weighted framework for the period 1990-2022 in Shaanxi province. Using a geographic explainable machine learning method, we then comprehensively evaluated the drivers of drought risk through three perspectives: global influence, marginal effect, and spatial pattern. The results revealed that drought risk in Shaanxi exhibited a temporal pattern of low in the early years, followed by a mid-period fluctuating increase, and a subsequent oscillating decrease thereafter. Significant spatial heterogeneity of drought risk was observed, with Yulin city and the middle region of central Shaanxi demonstrating the highest risk. Spatial effect emerged as the dominant driver of drought risk, accounting for 46.5% of total relative importance, suggesting that geographical context strongly modulates drought risk across Shaanxi. These findings enhance the understanding of drought causation and provide a scientific basis for developing targeted, regionally differentiated mitigation strategies.

Abstract Climate change impacts on water resources and drought characteristics will significantly affect regional agricultural water security. To clarify the response mechanism of water consumption and drought, this study coupled meteorological data from CMIP6 under two shared socioeconomic pathways in China. We used the 6‐month timescale standardized precipitation evapotranspiration index (SPEI‐6) and the AquaCrop model to quantify drought and the water footprint of maize production (WFprod), respectively. The study explored the response of WFprod to drought and its dominant processes from 2030 to 2099 over China. Results indicated that most regions exhibited intensifying drought trends, while limited areas showed wetting tendencies, with more pronounced trend magnitudes under SSP245. WFprod showed a more significant trend of change under SSP585, particularly a decreasing trend, with a broader affected area compared to SSP245. WFprod exhibited a negative correlation with the SPEI‐6 in most provinces, especially under SSP585. Under SSP245, the response of WFprod to drought in northern regions was primarily governed by physiological processes represented by yield, while in southern regions, it was mainly governed by physical processes related to crop water requirement. Under SSP585 scenario, the opposite was observed. Full irrigation could mitigate the negative impact of intensified drought on crop water use efficiency to some extent. However, in wetter regions, irrigation could increase possibility of higher water footprint. We revealed the response of WFprod to drought, providing a new perspective for water management under climate change.

Drought propagation is a key process linking meteorological anomalies to agricultural impacts within the hydrological cycle. Under climate warming, the superimposition of long-term increasing temperature trends ("Press") and short-term extreme drought events ("Pulse") fundamentally alters the propagation dynamics from meteorological drought to agricultural drought. Therefore, it is important to elucidate the driving mechanisms and impact patterns of this superposition effect on the drought propagation process. Using multi-source hydrometeorological datasets, we developed a Copula-based "Press-Pulse" framework to quantify meteorologicalto-agricultural drought propagation (MTAD) in the Yellow River Basin during 1961-2021, with a focus on propagation probabilities, propagation thresholds, and temperature-regulation effects,with a focus on propagation probabilities, propagation thresholds, and the temperature-regulation effects on MTAD. The results show that: (1) Propagation exhibits strong spatiotemporal heterogeneity, peaking in early summer (June) with basinaveraged probabilities exceeding 0.6 and response areas covering similar to 23% of the basin, before attenuating to similar to 0.3-0.4 by August due to precipitation replenishment. (2) The superimposition of high-temperature 'Press' significantly amplifies this risk, increasing agricultural drought probabilities by 10-25% and systematically deepening the triggering SPEI thresholds (e.g., from - 0.5 to -1.0), particularly in the water-limited middle reaches. (3) Identification of a critical Press Tipping Point reveals a distinct spatial divergence: while an intensified temperature press exacerbates drought susceptibility across the semi-arid Loess Plateau by accelerating soil moisture depletion, it conversely exerts a localized buffering effect in the upstream high-altitude regions, where the press-induced snowmelt recharge offsets pulse (precipitation) deficits. Overall, warming systematically lowers the barriers for drought propagation, underscoring the necessity of incorporating temperature-dependent dynamic thresholds into drought early warning and adaptation strategies.

Accurate streamflow forecasting is critical for water-resources management and disaster early warning. Combining seasonal–trend decomposition (STD) preprocessing with hydrological prediction models has become a prevailing approach. However, mainstream models such as LSTM, CNN, and GRU still struggle to capture the nonlinear dynamics of streamflow, and the effect of STD techniques on their predictive performance remains insufficiently explored. To address this gap, we introduce three cutting-edge architectures—FITS, FGN and PatchTST—into hydrology for the first time and benchmark them against the traditional LSTM, CNN and GRU baselines. Each of these six model classes is then paired with four STD techniques (MOV, LD, EXP and DFT-MOV), producing 24 hybrid models. We systematically evaluate all models at four streamflow gauging stations along the Jialing River. Without STD preprocessing, the FITS model achieved the highest predictive accuracy across the four stations (mean Nash–Sutcliffe efficiency, NSE = 0.986), followed by FGN (mean NSE = 0.984) and PatchTST (mean NSE = 0.981); all three also exhibited markedly greater stability than traditional models. Notably, FITS delivered these advantages at the lowest computational cost. Under a controlled computational budget, STD substantially improved the performance of LSTM, CNN, and GRU models (mean NSE increases of 0.12

Accurate streamflow forecasting underpins robust water-resources planning and flood-risk mitigation. Datadriven hydrological models-such as Long Short-Term Memory (LSTM) networks-capture temporal dependencies but seldom represent explicit interactions among hydro-meteorological variables. Hybrid approaches that embed convolutional neural networks (CNNs), graph neural networks (GNNs), or attention mechanisms (AMs) mitigate this limitation, yet they introduce substantial computational overhead and lack comprehensive benchmarking. We therefore proposed a lightweight frequency-domain multilayer perceptron (FDMLP) that transforms multivariate inputs into complex spectra along the feature dimension and applies the compact MLP to model their real and imaginary parts. Multi-step forecasts with 1-, 3-, and 5-day lead times were performed at three hydrometric stations located at the upper, middle, and lower reaches of the Yellow River. Four hybrids-AM-LSTM, CNN-LSTM, GNN-LSTM, and FDMLP-LSTM-were systematically evaluated against a plain LSTM baseline. All hybrid models outperformed the baseline. FDMLP-LSTM was the top performer, followed by CNN-LSTM. Moreover, FDMLP-LSTM consistently achieved the highest performance under extreme flow conditions. Despite this, the FDMLP layers increased training and inference times by only about 40 %, markedly less than the overhead introduced by AM and GNN layers. These findings highlight the pivotal role of modeling crossfeature dependencies and demonstrate that FDMLP can deliver superior predictive skill with a favorable accuracy-efficiency trade-off.

Drought-flood abrupt alternation (DFAA) events, characterized by rapid shifts between drought and flood conditions, are receiving increased attention due to their exacerbated impacts on human societies and ecosystems. This review systematically summarizes recent advances in the definitions, identifications, characterizations, attributions, effects, and future projections of DFAA events, along with current challenges and future research perspectives. Despite a significant increase in flood-related disasters since 2000, compounded by a notable rise in drought frequency, a unified definition of DFAA remains elusive. We explore meteorological, hydrological, and agricultural DFAA indices, highlighting limitations of threshold-based methods and indicator approaches. The synergistic interaction between anthropogenic climate change, driving a ∼7 % increase in atmospheric water vapor per degree Celsius, and natural climate variability exacerbates DFAA events. Climate projections indicate a potential increase in DFAA frequency, with global warming projected to raise the risk of DFAA, disproportionately affecting poorer populations. The review underscores the detrimental impacts of DFAA on crop yields and ecosystems, increasing soil erosion and wildfire risks. Future research should focus on standardized methodologies, integrated datasets, uncertainty analyses, and non-stationary fitting to improve the impact assessment and risk management of DFAA.

Ecosystem resilience is essential for sustaining ecosystems in the face of increasingly extreme climate conditions. However, current research faces challenges in improving the accuracy of identifying ecological drought events, enhancing temporal resolution, and refining model fitting for resilience quantification. In this study, a novel, daily-scale Ecosystem Service Supply (ESS) index is introduced, integrating water yield, carbon storage, and habitat quality into a unified measure weighted by ecosystem service equivalency factors. Baiyangdian wetland in North China was chosen as the study area, where the ESS index and the Standardized Ecological Water Deficit Index (SEWDI) were combined to jointly identify ecological drought events that directly alter ecosystem functioning. To characterize post-drought recovery dynamics, Bayesian non-parametric quantile regression model is employed to construct resilience curves, which capture the relationship between drought intensity and ecosystem recovery time at multiple quantiles. The results reveal that forests exhibit greater resilience under mild drought conditions, while grasslands demonstrate superior recovery capacity during more severe droughts. Due to differences in ecosystem structure and vegetation characteristics, forest ecosystems experience a significant decline in resilience under extreme drought conditions. In contrast, grassland ecosystems generally maintain stable or even slightly improved recovery resilience. By improving the temporal resolution of ecosystem assessments and introducing a flexible method to quantify resilience trajectories, this approach offers valuable insights for evaluating drought resilience and guiding ecosystem management in water-limited environments.

The soaring ecoclimatic events have undermined ecosystem diversity, vegetation health, and crop yield. However, ecological drought, as a vivid manifestation of an ecoclimatic event, is not adequately understood with respect to its spatiotemporal evolutions and risk assessment in a warming climate. Leveraging the simulations from the Coupled Model Intercomparison Project Phase 6 (CMIP6) under the Shared Socioeconomic Pathways with modest and higher emission scenarios (i.e., SSP2-4.5 and SSP5-8.5), we used the standardized ecological water shortage index (SEWDI), based on the ecological water deficit to characterize the ecological drought over Northwestern China (NWC). For assessing the change in risk of ecological drought under the SSP2-4.5 and SSP5-8.5 scenarios over NWC, the characteristics of ecological drought were extracted via the three thresholds in run theory, and then the corresponding risk indicators were calculated. Results showed that SEWDI can be utilized to effectively characterize the ecological drought over NWC. The decrement rate of severity for the ecological drought exceeded –0.02/10a over NWC in the middle future (i.e., the period 2051-2075) under the SSP2-4.5 and SSP5-8.5 scenarios. The regions with higher (lower) risk of ecological drought were mainly located in the western and central (eastern) parts of NWC under the SSP2-4.5 and SSP5-8.5 scenarios, especially a larger ecological drought risk would occur under the SSP5-8.5 scenario. In comparison with the historical period 1982–2014, the projected ecological drought risk over NWC pronounced increased under these two projected scenarios. These findings have implications for coping with the potential ecoclimatic risk of droughts.

Study Region: The Upper Yellow River Basin (UYRB), China Study focus: The increasing frequency, spatial extent, and intensity of hydrological droughts pose devastating impacts on water security, ecosystem stability, and sustainable development. While deep learning models have demonstrated significant promise in drought forecasting, particularly in data-scarce basins, their inherent opacity hinders the understanding of drought mechanisms. Therefore, we proposed a Spatial Explainable Deep Learning (SEDL) framework suitable for ungauged basins that can be integrated with various deep learning models. This framework aims to quantitatively analyze the spatial driving mechanisms that govern hydrological drought occurrence and enhances the accuracy of categorical drought prediction. New hydrological insights for the region: This study quantitatively demonstrated that natural runoff (with the mean contribution of 42.45 %) was the primary driving factor of hydrological droughts, surpassing the effects of precipitation (26.90 %) and average temperature (31.05 %). Crucially, regional hydrological drought occurrence was spatially influenced by temperature-runoff interactions in upstream headwater catchments and local precipitation variations. The improved model based on the SEDL framework achieved significant improvements in performance, with maximum increments of 8.7 % in accuracy, 62.7 % in Kappa coefficient (K), and 10.3 % in F1-score. By integrating deep learning with explainable artificial intelligence, the SEDL framework revealed the spatial physical driving factors of hydrological droughts while achieving 85.2 % accuracy on the test set, thus establishing a new research paradigm for explainable drought prediction.

Agricultural drought poses significant challenges to food production and ecosystem sustainability. The evolution of drought and its recurrence period feature are important for drought mitigation and risk management. This study aims to evaluate agricultural drought using the Standardized Soil Moisture Index (SSMI) based on Global Land Data Assimilation System (GLDAS) products, and extract drought variables using a three-dimensional identification method in Northwestern China. Then, the spatiotemporal dynamics and recurrence characteristics of agricultural droughts were evaluated. The results showed that: (1) Study area experienced alternating phases of dry and wet, with intensification in the 1960s and 1990s, and significant humidification post-2000. All four seasons show a wetting trend, while spring exhibited notable humidification trends. (2) Drought intensity and frequency displayed regional variability, agricultural drought in the west part of the study area were high in frequency but low in intensity, while the opposite was true in the east. (3) Agricultural drought exhibited cyclical behavior with dominant periods of 3.5, 6.6, and 13.5 years, reflecting interannual and interdecadal fluctuations. (4) Multivariate joint recurrence periods highlighted significant correlations among drought characteristics, emphasizing the risk of underestimation when considering single variables. These results offer valuable insights for water resources allocation and drought mitigation.

Climate change has intensified droughts, severely reducing vegetation productivity and even shifting the ecosystem from a carbon sink to a carbon source. Thus, understanding the spatial and temporal variations in vegetation responses to droughts is increasingly important. This study conducts coincidence analysis to examine the vulnerability and response time of vegetation to summer droughts from 1982 to 2022 across the Northern Hemisphere (NH) and employs random forest and partial correlation methods to identify their underlying drivers. The results reveal that arid regions and grasslands exhibit higher coincidence rates and shorter response time. Grasslands have the highest coincidence rate (0.38) and shortest response time (23 days), followed by shrublands, savannas, deciduous forests, and evergreen forests. Trends indicate that vegetation coincidence rates increased significantly (0.1/decade from 1993 to 2013), while lagged days decreased (-7.8 days/decade from 1990 to 2005), showing greater vulnerability to droughts. Spring phenology and productivity influence coincidence rate variations in about 27 % of the study area. Higher latitudes and cold regions exhibit stronger correlations between the start of growing season dates (SOS) and coincidence rates, suggesting that earlier growing seasons may enhance resistance to summer droughts in boreal forests. Conversely, in grasslands, earlier SOS negatively correlates with coincidence rates, indicating that rapid vegetation growth increases drought-related losses. These findings highlight the need to consider vegetation phenology interactions in drought assessments to improve ecosystem resilience and predictability.

Agricultural drought-induced yield reductions threaten socioeconomic stability and food security. Reliable early warning systems are essential for mitigating these threats. However, agricultural drought early warning based on meteorological conditions remains poorly explored, and prevalent conditional probabilities for determining drought propagation thresholds are largely subjective. In this study, an agricultural drought early warning threshold (ADEWarT) model was developed, combining Copula and diminishing marginal benefit theory, leveraging inherent regional drought propagation characteristics to objectively determine the thresholds. The ADEWarT model was applied and evaluated in the rain-fed agricultural areas of the Yellow River basin (raYRB) during the crop-growing seasons. Results showed that in most raYRB during the crop-growing seasons, early warning indicators for agricultural drought were the preceding meteorological drought conditions, while in the central-western raYRB during spring, they were the preceding compound drought and hot events. The performance metric, Matthews correlation coefficient (MCC), demonstrated that the ADEWarT model performed well (MCC > 0.4) across the raYRB. Compared with subjectively determined propagation thresholds, the proposed model exhibited superior performance during summer and autumn. The eXtreme Gradient Boosting combined with SHapley Additive exPlanations (XGBoost-SHAP) method revealed that the impacts of vegetation on early warning thresholds were modulated by hydrothermal conditions. Areas with lower absolute early warning thresholds were generally located on steep slopes, shady aspects, and at low elevations across the raYRB, indicating higher agricultural drought risk. This study provides technical guidance for agricultural drought risk management in the raYRB and offers a transferable framework applicable to other regions.

Understanding the dynamics of terrestrial carbon cycling is imperative for mitigating climate change. However, conventional analyses of Net Ecosystem Productivity (NEP) often treat extreme events as mere fluctuations, obscuring the mechanistic linkage between short-term disturbances and long-term ecosystem stability. To address this gap, we developed a novel analytical framework integrating three-dimensional event identification, long-term trend mutation detection, and machine learning-based attribution to analyze NEP in the Yellow River Basin (YRB) from 1982 to 2022. We found that: (1) The YRB’s overall greening trend conceals a complex reality of widespread structural changes, with “accelerated growth” patterns coexisting alongside alarming “increase-to-decrease” reversals, revealing significant underlying risks. (2) Our three-dimensional analysis, validated with independent data, identified 58 extreme carbon source events and established them as the direct trigger for the most frequent long-term trend mutations. (3) Water availability is the absolute dominant factor, and its quantified critical threshold (e.g., < 189 mm annual precipitation) provide a unified mechanistic framework that explains the basin’s spatial vulnerabilities, trend reversals, and extreme events. By pioneering an event-based, three-dimensional perspective, our study offers a new paradigm for assessing ecosystem resilience. The established linkages among short-term events, long-term mutations, and their hydro-climatic thresholds provide critical insights for developing proactive ecological risk management strategies.

In arid areas, the intricate interconnections and competition among water, agriculture, and ecology are particularly pronounced. Enhancing the synergy within the water-agriculture-ecology (WAE) system, while seeking common ground of competing sectors, presents a formidable challenge in managing water and land resources. In this study, the synergy of the WAE system was assessed using a coordinated development degree function, which was developed by considering the coefficient of variation and spatial distance projection. We investigated the multi-factor dynamic regulation of the WAE system through a water-agriculture-ecology co-optimization (WAECO) modelling framework, which adheres to a regulatory model that follows global-to-local optimization and bottom-up feedback. Using this framework, key factors, such as reservoir water supply, groundwater exploitation, and planting structure in the Shiyang River Basin (SRB), a typical arid basin in northwest China, were regulated. Results indicated crop yields and economic benefits in the baseline year reflected increments of 1.2 % and 5.4 %, respectively, compared to the actual scenario, while simultaneously increasing ecological water satisfaction by 11.1 % post-co-optimization. Through bilateral regulations between supply and demand, the annual average water deficit of the WAE system notably decreased from 7.5 % to 3.4 % in the mixed irrigation area of Liuhe midstream. The WAECO model effectively reconciled competing sectoral interests and improved the synergy of the WAE system, as indicated by a 6.3 % improvement in the coordinated development degree over the static regulation model. The new framework integrates a broad spectrum of regulatory factors and provides decision-makers with thorough and practical information, thereby facilitating the integrated management of the WAE system in arid areas.