The Ensemble Streamflow Prediction (ESP) framework combines a probabilistic forecast structure with process‐based models for water supply predictions. However, process‐based models require computationally intensive parameter estimation, increasing uncertainties and limiting usability. Motivated by the strong performance of deep learning models, we seek to assess whether the Long Short‐Term Memory (LSTM) model can provide skillful forecasts and replace process‐based models within the ESP framework. Given challenges in implicitly capturing snowpack dynamics within LSTMs for streamflow prediction, we also evaluated the added skill of explicitly incorporating snowpack information to improve hydrologic memory representation. LSTM‐ESPs were evaluated under four different scenarios: one excluding snow and three including snow with varied snowpack representations. The LSTM models were trained using information from 664 GAGES‐II basins during WY1983–2000. During a testing period, WY2001–2010, 80% of basins exhibited Nash‐Sutcliffe Efficiency (NSE) above 0.5 with a median NSE of around 0.70, indicating satisfactory utility in simulating seasonal water supply. LSTM‐ESP forecasts were then tested during WY2011–2020 over 76 western US basins with operational Natural Resources Conservation Services (NRCS) forecasts. A key finding is that in high snow regions, LSTM‐ESP forecasts using simplified ablation assumptions performed worse than those excluding snow, highlighting that snow data do not consistently improve LSTM‐ESP performance. However, LSTM‐ESP forecasts that explicitly incorporated past years' snow accumulation and ablation performed comparably to NRCS forecasts and better than forecasts excluding snow entirely. Overall, integrating deep learning within an ESP framework shows promise and highlights important considerations for including snowpack information in forecasting.

More than half of annual runoff across the montane regions of the western US and Europe originates as snowmelt, making knowledge of snowpack crucial to the quality of water supply forecasts. However, ongoing and projected warming is expected to reduce snow water equivalent (SWE) and alter snowmelt timing, thus impacting forecast skill and uncertainty. Rising temperatures are anticipated to reduce the fraction of future precipitation falling as snow by up to 30% in intermountain and continental regions of the western US. This will fundamentally alter the regional water cycle, and we posit that this will increase forecast errors and uncertainty, ultimately impacting the quality of decision-making that relies on water supply information. This research assesses the Relative Economic Value (REV) of water supply forecasts under changing snowpack regimes to understand the impact of forecast uncertainty on economic outcomes. Forecast errors will be rigorously estimated using statistical, physical, and machine learning models applied to 76 western US basins. Historical and projected future hydrology (2025-2050) will serve as a test bed for the analysis. Preliminary results over these basins suggest forecast errors on the order of +/- 25%, corresponding with changes in economic outcomes of up to +/- 15%. With findings from the proposed research, we hope to aid water entities by assessing the economic outcomes based on the skill of water supply forecasts, thereby exploring how forecast users can maximize productivity in response to changing climate conditions.

A mystery has emerged as to why patterns of increasing extreme rainfall have not been accompanied by similar levels of flooding, garnering growing attention given concerns over future flood risks. Antecedent moisture conditions have been proposed as the missing explanatory factor. Yet, reasons for moisture variability prior to flooding remain largely unstudied. Here, we evaluate the potential utility of precipitation intermittency, defined as the dry spell length prior to a flood, to explain the variability of flooding over 108 watersheds from 1950 to 2022. Flood magnitude is shown to be sensitive to intermittency, particularly in arid and semi-arid regions (PET/P > 0.84) and for basins with low soil field capacity (<0.31 m3/m3). Following extended dry spells >20 days, floods are only possible from the most intense storms, whereas a wider range of storms can produce flooding for shorter intermittency. The flood probability decreases by approximately 0.5 % for each additional day of dry spell, with overall flood probabilities being up to 30 % lower following extended dry periods. These results underscore the potential utility of precipitation intermittency for diagnosing current and future flood risks.

Understanding the relationship between remotely sensed snow disappearance and seasonal water supply may become vital in coming years to supplement limited ground based, in situ measurements of snow in a changing climate. For the period 2001–2019, we investigated the relationship between satellite derived Day of Snow Disappearance (DSD)—the date at which snow has completely disappeared—and the seasonal water supply, i.e., the April—July total streamflow volume, for 15 snow dominated basins across the western U.S. A Monte Carlo framework was applied, using linear regression models to evaluate the predictive skill—defined here as a model’s ability to accurately predict seasonal flow volumes—of varied predictors, including DSD and in situ snow water equivalent (SWE), across a range of spring forecast dates. In all basins there is a statistically significant relationship between mean DSD and seasonal water supply (p ≤ 0.05), with mean DSD explaining roughly half of the variance. Satellite-based model skill improves later in the forecast season, surpassing the skill of in-situ-based (SWE) models in skill in 10 of the 15 basins by the latest forecast date. We found little to no correlation between model error and basin characteristics such as elevation and the ratio of snow water equivalent to total precipitation. Despite a relatively short data record, this exploratory analysis shows promise for improving seasonal water supply prediction, in particular for snow dominated basins lacking in situ observations.

Extreme precipitation is projected to increase in many parts of the world, leading to concern over flooding and associated hazards. Despite the intuitive causal link between extreme precipitation and extreme flooding, observations and models suggest that other factors, in particular, low antecedent moisture may obscure or offset this relationship. Low antecedent soil moisture can arise through evaporation—driven by atmospheric warming that is projected to continue—but moisture is also reduced through an increase in the length of the time between storms, i.e., ‘storm intermittency’, which has been less studied and which can be derived directly from precipitation data. In this presentation, we focus on the modulating role of storm intermittency on the relationship between extreme precipitation and both flood potential, i.e., both flood peaks and volumes. We create an observation-based historical baseline of storm intermittency, 1950-2015, from a set of case-study basins to understand the relationship across a range of hydrometeorological settings. Next, the storm intermittency from a set of 16 CMIP6 climate models is evaluated relative to the baseline, and the evaluation is used to project future intermittency and likely outcomes on flood potential for the period 2016-2100. This projected flood potential is compared with hydrologic simulations from downscaled land surface models forced by the same CMIP6 models and differences between the projected and modeled outcomes are assessed in the context of simulated soil moisture and other hydrologic factors. Overall, we seek to understand the utility of storm intermittency as a predictor of flood potential and to understand the impacts of projected intermittency on future flood hazards.

In many parts of the world including the western United States, the allocation of water is governed by complex water laws that dictate who receives water, how much they receive, and when. Because these rules are generally based on the seniority of water rights, they are not necessarily focused on maximizing economic value across the entire economy. The maximization of value from water use economy-wide is a complex optimization problem that must explicitly consider each user’s water demand, willingness to pay (WTP) function, and the feedbacks among users in a coupled natural-human system model. In this study, we distill these complexities into a simple MATLAB® model developed to represent a two-user economy with water-dependent sectors representative of agriculture and industry. We feed the model with realistic values of relative water use, relative willingness to pay, and return flows to explore the relationships among these factors in water-limited systems. We find that the total economic value generated from water-dependent users depends primarily on the total water available in the system. However, for a given volume of water available, economic value is not necessarily maximized when all the water is appropriated to the user with the highest WTP. Rather, total economic value depends on the amount of water available, the relative WTP between the two users, and on the return flows generated from each sector’s water use. While our simple two-user model is a significant abstraction of the complexities inherent in natural systems, our study provides important insights into the coupled natural-human system dynamics of water allocation and use in water-limited environments.

Snowpack provides the majority of predictive information for water supply forecasts (WSFs) in snow-dominated basins across the western United States. Drought conditions typically accompany decreased snowpack and lowered runoff efficiency, negatively impacting WSFs. Here, we investigate the relationship between snow water equivalent (SWE) and April-July streamflow volume (AMJJ-V) during drought in small headwater catchments, using observations from 31 USGS streamflow gauges and 54 SNOTEL stations. A linear regression approach is used to evaluate forecast skill under different historical climatologies used for model fitting, as well as with different forecast dates. Experiments are constructed in which extreme hydrological drought years are withheld from model training, that is, years with AMJJ-V below the 15th percentile. Subsets of the remaining years are used for model fitting to understand how the climatology of different training subsets impacts forecasts of extreme drought years. We generally report overprediction in drought years. However, training the forecast model on drier years, that is, below-median years (P-15, P-57.5], minimizes residuals by an average of 10% in drought year forecasts, relative to a baseline case, with the highest median skill obtained in mid- to late April for colder regions. We report similar findings using a modified National Resources Conservation Service (NRCS) procedure in nine large Upper Colorado River basin (UCRB) basins, highlighting the importance of the snowpack-streamflow relationship in streamflow predictability. We propose an "adaptive sampling" approach of dynamically selecting training years based on antecedent SWE conditions, showing error reductions of up to 20% in historical drought years relative to the period of record. These alternate training protocols provide opportunities for addressing the challenges of future drought risk to water supply planning. Significance StatementSeasonal water supply forecasts based on the relationship between peak snowpack and water supply exhibit unique errors in drought years due to low snow and streamflow variability, presenting a major challenge for water supply prediction. Here, we assess the reliability of snow-based streamflow predictability in drought years using a fixed forecast date or fixed model training period. We critically evaluate different training protocols that evaluate predictive performance and identify sources of error during historical drought years. We also propose and test an "adaptive sampling" application that dynamically selects training years based on antecedent SWE conditions providing to overcome persistent errors and provide new insights and strategies for snow-guided forecasts.

Abstract An increase in heavy precipitation associated with climate change has exacerbated flooding in the Eastern U.S. To assess regional flood risk with changing climatic conditions, we demonstrate the application of a novel hydrologic modeling framework that integrates climate projections with a coupled Noah‐MP land surface model and a two‐dimensional HEC‐RAS hydrodynamic model. We employ this framework along a 41 km reach of the Susquehanna River near Harrisburg, Pennsylvania, where recent flood damages exceeded $2 billion (2011 Irene and Lee floods). Historical and future 30‐year and 100‐year peak‐discharge estimates were compared to assess how flood risk might be altered due to climate change. Results indicate that precipitation increases from climate change do not always lead to increases in flood risk, because interplay of hydrological components in the watershed, which are considered by Noah‐MP, largely controls flooding severity. However, climate change is expected to increase the severity of extreme events; if a 50‐year flood (the recurrence interval of Tropical Storm Lee) occurred toward the end of the 21st century in the worst‐case emission scenario, then flood volume would increase by 40% and flood extent by 15%, due to an increase in soil moisture from a wetter overall climate.

This study assessed the impacts of climate change on terrestrial hydrological components and Crop Water Use (CWU) over the Chesapeake Bay watershed using a combination of Global Climate Models (GCMs) and a land surface model. To better understand the impacts of climate change on the hydrological cycle, long-term simulations of multiple earth system models from the Coupled Model Intercomparison Project (CMIP Phase 5) are statistically downscaled and bias-corrected using Multivariate Adaptive Constructed Analogs (MACA) scheme for use as model forcing. Precipitation indices from the twenty MACA-based GCMs are used to identify six best performing models. A mesoscale approach is developed, where CWU is estimated by accounting for the impacts of changing climate conditions and rising CO2 levels. Daily grid-based crop coefficients are derived from evapotranspiration data. The findings indicate a significant annual increase in precipitation (10 %) and temperature (+4.5 K) for the RCP 8.5 scenario towards the end of the 21st century. A significant reduction (13 % and 17 % respectively) in CWU is estimated for corn and soybeans, resulting from increased total precipitation and rising CO2 levels suppressing evapotranspiration. Our results indicate that even in a warmer regime, crop water use decreased due to rising CO2 concentrations due to climate change.

Seasonal water supply predictions offer critical information to aid in planning and mitigation of drought impacts. In many northern and montane systems, spring snow information has been shown to be the most important predictor of seasonal drought, since in these systems snow water storage can exceed that of man-made reservoirs. However, a warmer future portends for less precipitation falling in the form of snow, which challenges current prediction methods. This presentation focuses on evaluating physical and statistical techniques for seasonal water supply prediction in snow-fed systems under both historical and future climate conditions with the goal of identifying regions and methods where predictions are likely to remain skillful under future warming. Initial results using downscaled hydrologic projections over the western U.S. indicate that snow information contributes less predictive skill to drought forecasts over roughly two thirds of snow dominated regions by the middle of this century. Significantly greater resilience to warming is seen higher elevation zones (p<0.01) and for prediction methods that include non-snow predictors such as soil moisture. To understand the impact of non-stationary snow conditions on future drought predictions, we conduct a series of idealized experiments to evaluate the relative importance of secular trends versus changing variability of both snow and seasonal climate conditions. This presentation is part of a larger research effort seeking to identify alternatives to snow-based streamflow predictions to advance future drought predictability.

A mesoscale evaluation is performed to determine the impacts of climate change on terrestrial hydrological components and the Net Irrigation Water Requirement (NIWR) throughout the Chesapeake Bay watershed in the mid-Atlantic region of the United States. The Noah-MP land surface model is calibrated and evaluated against the observed datasets of United States Geological Survey (USGS) streamflow gages, actual evapotranspiration from USGS Simplified Surface Energy Balance (SSEBop) Model and soil moisture from Soil Analysis Climate Network (SCAN). Six best performing Global Climate Models (GCM) based on Multivariate Adaptive Constructed Analogs (MACA) scheme are included for two future scenarios (RCP 4.5 and RCP 8.5), to assess the change in water balance components, change in NIWR for two dominant crops (corn and soybeans) and uncertainty in GCM projections. Using these long-term simulations, the flood inundation maps are developed for future scenarios along the Susquehanna River including the City of Harrisburg in Pennsylvania. The HEC-RAS 2D model is calibrated and evaluated against the high-water marks from major historical flood events and the stage-discharge relationship of the available USGS streamgages. Finally, the impacts of climate change are assessed on flood inundation depth and extent by comparing a 30-yr and 100-yr flood event based on the historical and future (scenario-based) peak discharge estimates at the USGS streamgages. Interestingly, flood inundation extent and severity predicted by the model along the Susquehanna River near Harrisburg is expected to rise in the future climate scenarios due to the greater frequency of extreme events increasing total precipitation. Evaluating Changes in Terrestrial Hydrological Components Due to Climate Change in the Chesapeake Bay Watershed

Millions of people in Indian sub-continent are affected by floods, which have increased during the past decades and likely to increase further in response to warming climate. However, changes in the frequency of floods and associated risks under the warming climate in Indian sub-continent remain largely unrecognized. Using the observations and simulations from Noah-MP land surface hydrological model, we examine the observed and projected changes in extreme precipitation and floods events. A majority of river basins in the Indian-subcontinent experienced a rise in extreme precipitation events during 1966–2005; however, these increasing trends are mostly insignificant at 5% level. Using downscaled and bias-corrected future and historic climate projections from Coupled Model Intercomparison Project (CMIP5) and the Noah-MP model, we show that the frequency of single and multi-day extreme precipitation and flood events are projected to increase substantially in the future over the Indian sub-continental river basins. A significant increase in multi-day (1–5 day) frequency of extreme precipitation and flood events, exceeding return levels based on the historic period (1966–2005), is projected under RCP 2.6 and RCP 8.5 emission scenarios. Our results show that the projected increase in flood frequency can be reduced to half under the low (RCP 2.6) emission scenario in comparison to the high emission scenario (RCP 8.5). More than half of the precipitation extremes result in flood events of 20, 50, and 100 year return periods in the majority of river basins in the Indian sub-continent. Moreover, the multi-day flood events are projected to increase with a faster rate in the future than the single day events, which can have strong implications for agriculture and infrastructure.

Despite the advances in climate change modeling, extreme events pose a challenge to develop approaches that are relevant for urban stormwater infrastructure designs and best management practices. The study first investigates the statistical methods applied to the land‐based daily precipitation series acquired from the Global Historical Climatology Network‐Daily (GHCN‐D). Additional analysis was carried out on the simulated Multivariate Adaptive Constructed Analogs (MACA)‐based downscaled daily extreme precipitation of 15 General Circulation Models and Weather Research and Forecasting‐based hourly extreme precipitation of North American Regional Reanalysis to discern the return period of 24‐hr and 48‐hr events. We infer that the GHCN‐D and MACA‐based precipitation reveals increasing trends in annual and seasonal extreme daily precipitation. Both BCC‐CSM1‐1‐m and GFDL‐ESM2M models revealed that the magnitude and frequency of extreme precipitation events are projected to increase between 2016 and 2099. We conclude that the future scenarios show an increase in magnitudes of extreme precipitation up to three times across southeastern Virginia resulting in increased discharge rates at selected gauge locations. The depth‐duration‐frequency curve predicted an increase of 2–3 times in 24‐ and 48‐h precipitation intensity, higher peaks, and indicated an increase of up to 50% in flood magnitude in future scenarios.

Abstract. India has witnessed some of the most severe historical droughts in the current decade, and severity, frequency, and areal extent of droughts have been increasing. As a large part of the population of India is dependent on agriculture, soil moisture drought affecting agricultural activities (crop yields) has significant impacts on socio-economic conditions. Due to limited observations, soil moisture is generally simulated using land-surface hydrological models (LSMs); however, these LSM outputs have uncertainty due to many factors, including errors in forcing data and model parameterization. Here we reconstruct agricultural drought events over India during the period of 1951–2015 based on simulated soil moisture from three LSMs, the Variable Infiltration Capacity (VIC), the Noah, and the Community Land Model (CLM). Based on simulations from the three LSMs, we find that major drought events occurred in 1987, 2002, and 2015 during the monsoon season (June through September). During the Rabi season (November through February), major soil moisture droughts occurred in 1966, 1973, 2001, and 2003. Soil moisture droughts estimated from the three LSMs are comparable in terms of their spatial coverage; however, differences are found in drought severity. Moreover, we find a higher uncertainty in simulated drought characteristics over a large part of India during the major crop-growing season (Rabi season, November to February: NDJF) compared to those of the monsoon season (June to September: JJAS). Furthermore, uncertainty in drought estimates is higher for severe and localized droughts. Higher uncertainty in the soil moisture droughts is largely due to the difference in model parameterizations (especially soil depth), resulting in different persistence of soil moisture simulated by the three LSMs. Our study highlights the importance of accounting for the LSMs' uncertainty and consideration of the multi-model ensemble system for the real-time monitoring and prediction of drought over India.