Search

Accurate flood risk assessment requires a comprehensive understanding of flood sensitivity to regional drivers and climate factors. This paper presents the scaling of floods (duration, peak, volume) with geomorphologic characteristics of the basin (i.e., drainage area, slope, elevation) and precipitation patterns (rainfall accumulation, variability). Long‐term daily streamflow observations over the 20th and early 21st centuries from Hydro‐Climatic Data Network streamgages across the conterminous United States are used to create a flood event database based on their flood stage information. Antecedent daily rainfall accumulation and variability corresponding to these floods are computed using Global Historical Climatology Network daily data set. Two Bayesian scaling models are developed, and the spatial organization of scaling exponents is investigated. The baseline model quantifies the scaling of floods to geomorphologic characteristics. The dynamic model quantifies the scaling of floods to antecedent precipitation distribution which is further conditioned on geomorphologic characteristics. Results show that small and low‐elevation basins have a stronger response to antecedent rainfall distribution in amplifying flood peaks, while high‐elevation steeper basins have a lower response for flood duration and volume. The dynamic models demonstrate that there are significant variations in the flood scaling rates, with the largest rates up to 40% and 4.5% for flood duration, 64% and 44% for peak, and 98% and 40% for volume found across the Northeast, Coastal Southeast, and Northwest with intensifying rainfall accumulation and variability, respectively. This study advances flood predictions by better informing the flood attributes in the context of dynamical land‐atmosphere perturbations. Plain Language Summary: The relationships between floods, physical properties of river basins, and the spatio‐temporal distribution of precipitation define the regional flood scaling laws. Improved regional flood scaling models are required for updating flood risk scoring tools to help insurers and decision‐makers in the light of anthropogenic climate changes and natural variability of Earth systems. Two scaling models are developed here to examine how the physical properties of basins and precipitation patterns can inform flood attributes across the conterminous United States (US). Using long records of streamflow and precipitation observations, we identified different scaling relationships between the catchment factors, preceding precipitation, and the duration, peaks, and volume of floods. Larger basins located at higher elevations are found to have lower scaling rates with flood peaks when interacting with the preceding precipitation patterns. Overall, low‐elevation flat basins across the Eastern, Southern, and Northwestern US have a larger positive response to precipitation in amplifying flood attributes. These new scaling laws can help advance our understanding of controlling mechanisms for flash floods and mega‐floods, which is required for a dynamic flood risk assessment in a changing environment. Key Points: Two Bayesian scaling models are developed to understand the sensitivity of floods to geomorphologic characteristics and catchment rainfallLow‐elevation and flat basins with different sizes have greater scaling to rainfall in amplifying flood duration and volumesLow‐elevation and small basins with different steepness show larger sensitivity to rainfall distribution that enhances flood peaks [ABSTRACT FROM AUTHOR]

Despite long‐standing efforts, hydrologists still lack robust tools for calibrating land surface model (LSM) streamflow estimates within ungauged basins. Using surface soil moisture estimates from the Soil Moisture Active Passive Level 4 Soil Moisture (L4_SM) product, precipitation observations, and streamflow gauge measurements for 617 medium‐scale (200–10,000 km2) basins in the contiguous United States, we measure the temporal (Spearman) rank correlation between antecedent (i.e., pre‐storm) surface soil moisture (ASM) and the storm‐scale runoff coefficient (RC; the fraction of storm‐scale precipitation accumulation converted into streamflow). In humid and semi‐humid basins, this rank correlation is shown to be sufficiently strong to allow for the substitution of storm‐scale RC observations (available only in basins that are both lightly regulated and gauged) with high‐quality ASM values (available quasi‐globally from L4_SM) in streamflow calibration procedures. Using this principle, we define a new, basin‐wise LSM streamflow calibration approach based on L4_SM alone and successfully apply it to identify LSM configurations that produce a high rank correlation with observed RC. However, since the approach cannot detect RC bias, it is less successful in identifying LSM configurations with low mean‐absolute error. Plain Language Summary: Accurately forecasting the fraction of rainfall that runs off into streams, as opposed to infiltrates into the soil, is critical for flash‐flood prediction, water‐resource monitoring, and tracking the transport of nutrients from agricultural fields into local waterways. Such forecasting is typically performed by hydrologic models that attempt to represent the physical processes responsible for surface runoff generation. However, to provide accurate streamflow forecasts, these models typically need to be calibrated against actual streamflow observations. This is problematic given the relatively poor, and declining, global availability of stream gauges. This paper presents a novel model calibration strategy that uses soil moisture from remote sensing and numerical modeling in place of streamflow observations during calibration. This transition has significant practical advantages because, unlike streamflow observations, the soil moisture data are continuously available across space. Our results demonstrate that this new approach can significantly improve hydrologic models within humid and semi‐humid basins lacking sufficient ground‐based instrumentation for traditional streamflow calibration. Key Points: Hydrologic models still perform poorly in ungauged basinsIn relative humid areas, the correlation between pre‐storm soil moisture and the storm‐scale runoff coefficient approaches unitySuch strong correlation can be exploited to improve the precision of runoff coefficient estimates provided by models in ungaged basins [ABSTRACT FROM AUTHOR]

Inflow anomalies at varying temporal scales, seasonally varying storage mandates, and multipurpose allocation requirements contribute to reservoir operational decisions. The difficulty of capturing these constraints across many basins in a generalized framework has limited the accuracy of streamflow estimates in land‐surface models for locations downstream of reservoirs. We develop a Piecewise Linear Regression Tree to learn generalized daily operating policies from 76 reservoirs from four major basins across the coterminous US. Reservoir characteristics, such as residence time and maximum storage, and daily state variables, such as storage and inflow, are used to group similar observations across all reservoirs. Linear regression equations are then fit between daily state variables and release for each group. We recommend two models—Model 1 (M1) that performs the best when simulating untrained records but is complex and Model 2 (M2) that is nearly as performant as M1 but more parsimonious. The simulated release median root mean squared error is 49.7% (53.2%) of mean daily release with a median Nash‐Sutcliffe efficiency of 0.62 (0.52) for M1 (M2). Long‐term residence time is shown to be useful in grouping similar operating reservoirs. Release from low residence time reservoirs can be mostly described using inflow‐based variables. Operations at higher residence time reservoirs are more related to previous release variables or storage variables, depending on the current inflow. The ability of the models presented to capture operational dynamics of many types of reservoirs indicates their potential to be used for untrained and limited data reservoirs. Plain Language Summary: It is difficult to estimate releases from many different types of reservoirs using the same method. This means that they are often poorly represented in hydrologic models. To overcome this challenge, we use a statistical model that can group observations that are similar and fit a linear equation to them using only information that is widely available for many reservoirs, like storage and release records. The model can accurately predict release for many reservoirs across four different basins in the United States in addition to providing relationships between the state of a reservoir and its expected release that can be used to build better models in the future. Key Points: Piecewise Linear Regression Trees can accurately capture operational dynamics from many reservoirs in a single, generalized formReservoir residence time and storage capacity can be powerful variables for grouping similar reservoirs across basins and climates [ABSTRACT FROM AUTHOR]

We use the five landscape ecology metrics of aggregation index, percentage of like adjacencies, interspersion and juxtaposition index, patch density, and Shannon's evenness index to assess spatial heterogeneity at 15 floodplains in the continental United States. Assessments are based on floodplain classes and patches delineated remotely using topography and vegetation. Floodplain reaches examined here represent diverse drainage areas, flow regimes, valley geometries, channel planforms, and biomes. We selected sites with minimal direct human alteration. Our objectives are to quantify floodplain spatial heterogeneity; evaluate whether statistically significant patterns are present; and interpret the statistical analyses with respect to the influence of lateral channel mobility and valley‐floor space available. We develop a conceptual model of the influences on lateral mobility and space available, and then test specific hypotheses derived from this conceptual model. These natural floodplains have a median aggregation index of 58.8%, median percentage of like adjacencies of 58.5%, median interspersion and juxtaposition index of 74.9%, median density of 1,241 patches/ha, and median Shannon's evenness index of 0.934 (n = 15). In other words, natural floodplains have moderate aggregation of classes, high evenness and intermixing of classes, and a wide range of patch densities. Drainage area, the ratio of floodplain/channel width, elevation, precipitation, total sinuosity, large wood volume, planform, and flow regime emerge as important variables to floodplain heterogeneity. These results highlight the influence of biotic‐abiotic interactions in shaping floodplain heterogeneity across diverse river corridors. Plain Language Summary: As river channels move across their valley floors, they create diverse topography including natural levees, cutoff meanders, secondary channels, and floodplain wetlands. These features increase the spatial variability of the floodplain. Many studies indicate that this variability strongly influences habitat abundance and diversity, storage of flood waters and excess nitrate or phosphate, and other floodplain functions. However, no one has systematically measured floodplain variability across multiple rivers and different geographic regions. We measured the variability of natural floodplains at 15 sites in the continental United States using metrics developed by landscape ecologists. We find that natural floodplains have distinctive variability signatures that relate to drainage area, the size of the floodplain, and other characteristics. Key Points: Natural floodplains have moderate aggregation of classes, high evenness and intermixing of classes, and a wide range of patch densitiesFloodplain heterogeneity reflects space available on the valley floor and lateral channel mobility [ABSTRACT FROM AUTHOR]

Summer streamflow predictions are critical for managing water resources; however, warming‐induced shifts from snow to rain regimes impact low‐flow predictive models. Additionally, reductions in snowpack drive earlier peak flows and lower summer flows across the western United States increasing reliance on groundwater for maintaining summer streamflow. However, it remains poorly understood how groundwater contributions vary interannually. We quantify recession limb groundwater (RLGW), defined as the proportional groundwater contribution to the stream during the period between peak stream flow and low flow, to predict summer low flows across three diverse western US watersheds. We ask (a) how do snow and rain dynamics influence interannual variations of RLGW contributions and summer low flows?; (b) which watershed attributes impact the effectiveness of RLGW as a predictor of summer low flows? Linear models reveal that RLGW is a strong predictor of low flows across all sites and drastically improves low‐flow prediction compared to snow metrics at a rain‐dominated site. Results suggest that strength of RLGW control on summer low flows may be mediated by subsurface storage. Subsurface storage can be divided into dynamic (i.e., variability saturated) and deep (i.e., permanently saturated) components, and we hypothesize that interannual variability in dynamic storage contribution to streamflow drives RLGW variability. In systems with a higher proportion of dynamic storage, RLGW is a better predictor of summer low flow because the stream is more responsive to dynamic storage contributions compared to deep‐storage‐dominated systems. Overall, including RLGW improved low‐flow prediction across diverse watersheds. Plain Language Summary: Water managers across the western United States depend on accurate streamflow prediction models for water planning and allocation during summer months. Historically, these models use snow metrics to predict summer flows, but increasing temperatures across the western US are decreasing snow input and accumulation and increasing early snow melt rates leading to changes in streamflow generation mechanisms. Here, we seek to understand how stream flows will respond under warmer climates in three watersheds with distinct climate and underlying bedrock in the western US. We expanded upon commonly used low‐flow model snow metrics to include snow and streamflow metrics from the previous year as well as groundwater dynamics during the annual recession curve. We found that including variables outside of commonly used snow parameters in the models produced more accurate predictions of summer low flows. We provide a framework for future analysis of streamflow response to warming across the western US. Key Points: Recession limb groundwater proportion is a strong predictor of summer low flowsRecession limb dynamics are especially important for improving low‐flow prediction at rain‐dominated sitesRecession limb groundwater may be a better predictor of low flow at sites where dynamic storage exceeds deep storage [ABSTRACT FROM AUTHOR]

Global flood models integrate flood maps of constant probability in space, ignoring the correlation between sites and thus potentially misestimating the risk posed by extreme events. Stochastic flood models alleviate this issue through the simulation of flood events with a realistic spatial structure, yet their proliferation at large scales has historically been inhibited by data quality and computer availability. In this paper, we show, for the first time, the efficacy of modeled river discharge reanalyses in the characterization of flood spatial dependence in the absence of a dense stream gauge network. While global hydrological models may show poor correspondence with absolute observed river flows, we find that the rate at which they can simulate the joint occurrence of relative flow exceedances at two given locations is broadly similar to when a gauge‐based statistical model is used. Evidenced over the United States, flood events simulated using observed gauge data from the U.S. Geological Survey versus those generated using modeled streamflows have similar (i) distributions of site‐to‐site correlation strength, (ii) relationships between event size and return period, and, importantly, (iii) loss distributions when incorporated into a continental‐scale flood risk model. Extremal dependence is generally quantified less accurately on larger rivers, in arid climates, in mountainous terrain, and for the rarest high‐magnitude events. However, local‐scale errors are shown to broadly cancel each other out when combined, producing an unbiased flood spatial dependence model. These findings suggest that building accurate stochastic flood models worldwide may no longer be a distant aspiration. Plain Language Summary: Global flood risk is commonly estimated through flood inundation maps with a defined probability of occurrence. These flood simulations have a key drawback in that they fail to capture the spatial patterns exhibited during real flood events, instead modeling the same probability of flooding on every river at once. Solutions which rely on networks of gauged river flow observations will necessarily break down in the majority of the world's regions which lack such a resource. In this paper, we use historic river flows simulated by global rainfall‐runoff models (rather than observed flows) into a statistical model which captures the spatial correlation of flow extremes. If we examine the relative flow exceedance probabilities from these hydrological models rather than the volumetric flow values, flood events are generated which exhibit similar characteristics to those when gauged flow observations are used. Crucially, the simulation‐ and observation‐generated flood events produce near‐identical losses to buildings in the United States. The implications of this are that true stochastic flood risk models, which account for spatial dependence, can proliferate globally via the generation of realistic flood event sets from hydrological models. Key Points: Large‐scale flood hazard models typically neglect to represent the spatial dependence of real flood eventsRelative flow exceedances simulated by the fusion of global hydrological with statistical models reproduce gauge‐driven flood event setsAt the continental scale, key characteristics of a flood risk model are indistinguishable when driven with observed versus modeled flow data [ABSTRACT FROM AUTHOR]

This paper is motivated to understand the multi‐temporal scale characteristics of Horton Index (HI), an indicator of catchment‐scale plant water use defined as the ratio of catchment vaporization (total evapotranspiration) to catchment wetting (precipitation minus surface runoff). Based on the generalized proportionality hypothesis, we first develop an analytical framework of HI applicable at multi‐temporal scales as a single function of ecological aridity index (EAI), defined as the ratio of potential evapotranspiration to catchment wetting. We successfully validate this framework at the long‐term, annual, and monthly scales at 343 natural catchments over the conterminous United States under varying climate and vegetation regimes. We show that vegetation plays an important role in regulating not only the partitioning of catchment wetting into vertical vaporization and horizontal baseflow, but also further partitioning of vaporization into initial and continuing components. Aided by this framework, we reveal emergent space‐time similarity patterns between HI's spatial (inter‐catchment, regional) and temporal (within‐catchment, inter and intra‐annual) variability. On the one hand, HI increases with EAI, both spatially and temporally, following similar trends predicted by the analytical framework. On the other hand, the slopes of HI ∼ EAI relationships, denoted as d (HI)/d (EAI), decreased with EAI both spatially and temporally and, again, following similar trends suggested by the analytical framework. These findings suggest the promising potential of Horton Index as a conceptual yet quantitative framework for improving understanding and modeling of multi‐scale ecological and hydrological processes and their interactions. Key Points: Derivation and validation of an analytical framework of Horton Index applicable over long‐term, annual, and monthly scales across biomesVegetation regulates catchment storage partitioning into vaporization and baseflow, then vaporization into initial and continuing componentsEmergent space‐time similarity patterns detected empirically and explained theoretically using the analytical framework [ABSTRACT FROM AUTHOR]

Many rivers and streams are ungauged or poorly gauged and predicting streamflow in such watersheds is challenging. Although streamflow signals result from processes with different frequencies, they can be "sparse" or have a "lower‐dimensional" representation in a transformed feature space. In such cases, if this appropriate feature space can be identified from streamflow data in gauged watersheds by dimensionality reduction, streamflow in poorly gauged watersheds can be predicted with a few measurements taken. This study utilized this framework, named data‐driven sparse sensing (DSS), to predict daily‐scale streamflow in 543 watersheds across the contiguous United States. A tailored library of features was extracted from streamflow training data in watersheds within the same climatic region, and this feature space was used to reconstruct streamflow in poorly gauged watersheds and identify the optimal timings for measurement. Among different regions, streamflow in snowmelt‐dominated and baseflow‐dominated watersheds (e.g., Rocky Mountains) was more effectively predicted with fewer streamflow measurements taken. The prediction efficiency in some rainfall‐dominated regions, for example, New England and the Pacific coast, increased significantly with an increasing number of measurements. The spatial variability of prediction efficiency can be attributed to the process‐driven mechanisms and the dimensionality of watershed dynamics. Storage‐dominated systems are lower‐dimensional and more predictable than rainfall‐dominated systems. Measurements taken during periods with large streamflow magnitudes and/or variances are more informative and lead to better predictions. This study demonstrates that DSS can be an especially useful technique to integrate ground‐based measurements with remotely sensed data for streamflow prediction, sensor placement, and watershed classification. Plain Language Summary: Many rivers and stream reaches are ungauged or poorly gauged because streamflow measurement is costly and resource intensive. Predicting the streamflow time‐series in these ungauged or poorly gauged watersheds is still challenging. Here, we use a signal processing technique called data‐driven sparse sensing on a national‐scale streamflow data set across the contiguous United States. We predict streamflow time‐series in each watershed based on existing streamflow data in watersheds nearby, and explore the best times during the year for measuring streamflow. Our analysis shows that data‐driven sparse sensing is an effective tool to predict streamflow time‐series in poorly gauged watersheds based on very few streamflow measurements. The streamflow in watersheds with high snowmelt and high baseflow can be more easily predicted than in other watersheds. Our analysis also shows that the streamflow measurements taken during periods with large streamflow peaks and variances contain more information and are beneficial for making predictions. We conclude that data‐driven sparse sensing can be further used to classify watersheds and to identify the best locations for streamflow gauging. Key Points: We utilize data‐driven sparse sensing to predict daily streamflow and identify the optimal times for streamflow measurement across the contiguous United StatesStreamflow was more effectively predicted in watersheds dominated by snowmelt and baseflow than those dominated by rainfall and quickflowThe optimal sampling times for streamflow prediction by data‐driven sparse sensing are periods with large flow magnitudes and variances [ABSTRACT FROM AUTHOR]

Multidecadal inland surface water dynamics are of increasing interest due to their importance to climate, ecology, and society, yet several key challenges impede long‐term monitoring of inland surface waters globally. This research investigates two novel methods, one addressing subhydroflattened surface estimate uncertainty issues, and a second addressing temporal resolution issues, using 46 water bodies across the western United States. First, low water level estimate uncertainty was reduced using multiple digital elevation models (ALOS, SRTM, and NED) to derive the hypsometric relationship for each lake from the digital elevation model with the lowest hydroflattened water surface. This technique reduced the number of images with subhydroflattened water surfaces by at least 549 over the best individual DEM resulting in higher water surface elevation estimate accuracy. Second, this paper introduces proportional hypsometry which dynamically generates surface area/elevation relationships for every image using clear pixels only by removing contamination from both the image and DEM. Proportional hypsometry was found to be ill‐suited for subhydroflattened water surface levels but produced comparable accuracy to clear images for above hydroflattened water levels. Overall, using the lowest hydroflattened surface along with proportional hypsometry improved temporal resolution enabling analysis of nearly 10,000 additional images while maintaining similar accuracy levels as images with <1% contamination (2.35 m RMSE vs. 2.17 m RMSE). This research increases lower water elevation estimate accuracy and temporal resolution and is scalable enabling regional and global water dynamic analysis. Key Points: Lowest hydroflattened surface models reduce low‐level water surface elevation estimate uncertaintyProportional hypsometry substantially increases the temporal resolution of the analysis while maintaining similar accuracy [ABSTRACT FROM AUTHOR]

The flow duration curve (FDC) is a hydrologically meaningful representation of the statistical distribution of daily streamflows. The complexity of processes contributing to the FDC introduces challenges for the direct exploration of physical controls on FDC. In this paper, the controls of climate and catchment characteristics on FDC are explored using a stochastic framework that enables construction of the FDC from three components of streamflow: fast and slow flow (during wet days) and slow flow during dry days. The FDC during wet days (FDCw) is computed as the statistical sum of the fast flow duration curve (FFDC) and the slow flow duration curve (SFDCw), considering their dependency. FDC is modeled as the mixture distribution of FDCw and the slow flow duration curve during dry days (SFDCd), by considering the fraction of wet days (δ) for perennial streams and both δ and the fraction of days of zero streamflow for ephemeral streams. The Kappa distribution is employed to fit the FFDC, SFDCw, and SFDCd for 300 catchments from Model Parameter Estimation Experiment (MOPEX) across the United States. Results show that the 0–20th percentile of FDC is controlled by FFDC and SFDCw, the 90–100th percentile of FDC is controlled by SFDCd, and the 20–90th percentile of FDC is controlled by three components. The relationships between estimated Kappa distribution parameters and climate and catchment characteristics reveal that the aridity index, the coefficient of variation of daily precipitation, timing of precipitation, time interval between storms, snow, topographic slope, and slope of recession slope curve are dominant controlling factors. Key Points: Flow duration curve (FDC) is modeled from three components: fast flow and slow flow during wet days and slow flow during dry daysThe duration curves for each component are modeled by Kappa distribution, and the climatic and landscape controls on FDCs are exploredAridity index, coefficient of variation of daily precipitation, slope, and recession slope are the dominant controls on FDCs [ABSTRACT FROM AUTHOR]

Drought monitoring and forecasting systems are used in the United States (U.S.) to inform drought management decisions. Drought forecasting efforts have often been conducted and evaluated at coarse spatial resolutions (i.e., >10‐km), which miss key local drought information at higher resolutions. Addressing the importance of forecasting drought at high resolutions, this study develops statistical models to evaluate 1‐ to 3‐month lead time predictability of meteorological and agricultural summer drought across the western U.S. at a 4‐km resolution. Our high‐resolution drought predictions have statistically significant skill (p ≤ 0.05) across 70%–100% of the western U.S., varying by evaluation metric and lead time. 1‐ to 3‐month lead time drought forecasts accurately represent monitored summer drought spatial patterns during major drought events, the interannual variability of drought area from 1982 to 2020 (r = 0.84–0.93), and drought trends (r = 0.94–0.97). 71% of western U.S summer drought area interannual variability can be explained by cold‐season (November–February) climate conditions alone allowing skillful 3‐month lead time predictions. Pre‐summer drought conditions (represented by drought indices) are the most important predictors for summer drought. Thus, the statistical models developed in this study heavily rely on the autocorrelation of chosen agricultural and meteorological drought indices which estimate land surface moisture memory. Indeed, prediction skill strongly correlates with persistence of drought conditions (r ≥ 0.73). This study is intended to support future development of operational drought early warning systems that inform drought management. Plain Language Summary: Droughts are complex and devastating natural disasters with severe implications in the western U.S. for energy production, agricultural yields, and wildfires, particularly during summer months. Drought monitoring and forecasting systems are used in the U.S. to inform drought management decisions. However, existing products and analyses are typically performed at coarse spatial resolutions (>10‐km). Because summer drought is spatially variable at finer resolutions than current drought monitoring and forecasting systems, it is valuable to develop seasonal drought forecasting systems at higher spatial resolutions. In this study we present a novel statistical modeling framework to forecast summer drought conditions at a high resolution (4‐km) at 1‐ to 3‐month lead times across the western U.S. Our models provide skillful predictions of drought across most of the western U.S., but there is degraded skill in areas where drought tends to be less persistent. Future research can expand on our work by predicting drought at finer temporal resolutions while maintaining the high spatial resolution used in this study's methodology. This research is intended to support development of future operational drought early warning systems that provide high spatial resolution forecasts. Key Points: Statistical models accurately predict summer meteorological and agricultural drought at 1‐, 2‐, and 3‐month lead times at a 4‐km resolutionMost of the interannual variability in summer drought area can be explained by preceding cold‐season climate conditionsStatistical models rely heavily on the autocorrelation of drought indices in drought forecasts [ABSTRACT FROM AUTHOR]

Trends in flood magnitudes vary across the conterminous USA (CONUS). There have been attempts to identify what controls these regionally varying trends, but these attempts were limited to certain—for example, climatic—variables or to smaller regions, using different methods and datasets each time. Here we attribute the trends in annual maximum streamflow for 4,390 gauging stations across the CONUS in the period 1960–2010, while using a novel combination of methods and an unprecedented variety of potential controlling variables to allow large‐scale comparisons and minimize biases. Using process‐based flood classification and complex networks, we find 10 distinct clusters of catchments with similar flood behavior. We compile a set of 31 hydro‐climatological and land use variables as predictors for 10 separate Random Forest models, allowing us to find the main controls the flood magnitude trends for each cluster. By using Accumulated Local Effect plots, we can understand how these controls influence the trends in the flood magnitude. We show that hydro‐climatologic changes and land use are of similar importance for flood magnitude trends across the CONUS. Static land use variables are more important than their trends, suggesting that land use is able to attenuate (forested areas) or amplify (urbanized areas) the effects of climatic changes on flood magnitudes. For some variables, we find opposing effects in different regions, showing that flood trend controls are highly dependent on regional characteristics and that our novel approach is necessary to attribute flood magnitude trends reliably at the continental scale while maintaining sensitivity to regional controls. Key Points: A wide variety of controls are necessary to explain flood magnitude trends across the United States between 1960 and 2010Climatic changes and land cover conditions are of similar importance for flood magnitude trends at the regional scaleControls on flood trends can have highly nonlinear effects and can have opposing effects in different hydro‐climatological subregions [ABSTRACT FROM AUTHOR]

Numerous studies have evaluated the reliability and hydrologic utility of various rainfall data sets through hydrological modeling. However, the calibration of hydrological models compensates for errors in rainfall inputs. The drivers, conditions, and factors affecting the calibration of hydrological models given the accuracy of rainfall inputs are not well understood. Here, we explore hydrological model adaptability to rainfall inputs of varied quality and its potential mechanisms. Twenty‐eight rainfall products from multiple sources are collected for a headwater catchment in the Southern United States. These rainfall data sets include measurements from rain gauges, weather radars, satellites, reanalysis products, and Weather Research and Forecasting model simulations. Such rainfall data sets with varied errors are used to independently calibrate a widely used conceptual Xin'anjiang (XAJ) hydrological model. Results suggest that the hydrological model can often adapt well to two scenarios of inaccurate rainfall inputs producing high‐performance streamflow simulations. This adaptive ability is controlled by an adaptable threshold of the overall bias of the rainfall inputs. Moreover, hydrological model adaptability to rainfall inputs is further influenced by how event‐based rainfall bias shapes the overall rainfall bias, especially from those of heavy rainstorms. The hydrological model can adapt to those rainfall inputs that contain important information content for model calibration. Notably, the adaptability to rainfall inputs of the XAJ model is mainly controlled by a bias reduction through adjustment of evapotranspiration and soil moisture storage, yielding satisfactory effective rainfall. The study quantitatively sheds new light on hydrological model adaptability to rainfall input quality. Plain Language Summary: Accurate rainfall is an ideal input for the hydrological model to simulate streamflow reliably. However, some inaccurate rainfall data sets can be compensated by hydrological model calibration to generate good streamflow. It is still unclear what kind of rainfall inputs the model calibration can and cannot adapt to. Here, we explore a large number of rainfall data sets from different sources and their corresponding streamflow simulation performance after independent model calibrations. It is found that the hydrological model can adapt well to two scenarios of inaccurate rainfall data. The adaptive ability of the hydrological model calibration to rainfall inputs is not only affected by the accuracy over the entire period but also by the accuracy of individual rainfall events, especially those of the most severe rainstorms. Therefore, a good hydrological simulation does not always mean its rainfall input data are reliable. This study enhances a quantitative understanding of how the model calibration adapts to the errors in rainfall input data. The findings are expected to provide quantitative guidance for bias correction and data fusion of rainfall products. Key Points: Two scenarios of inaccurate rainfall inputs and a threshold of overall rainfall bias are identified for hydrological model adaptabilityHydrological model adaptability is further influenced by how event‐based bias of individual rainstorms shapes the overall rainfall biasHydrological model adaptability is mainly controlled by a bias reduction through adjustment of evapotranspiration and soil moisture storage [ABSTRACT FROM AUTHOR]

Dissolved oxygen (DO) indicates the overall stream water quality and ecosystem health. We investigated emergent scaling of DO with the dominant environmental drivers in freshwater (non‐coastal) streams across the contiguous United States. Available data of monthly to quarterly sampling frequencies during 1998–2015 were obtained for 86 U.S. streams. Data analytics indicated water temperature (Tw) and pH (a proxy of carbon dioxide) dominating the key environmental process components of DO concentrations in the freshwater streams. The "climatic" process component (comprising Tw and net radiation) had, respectively, ∼3 and ∼9 times stronger control on DO than the "biogeochemical" (total nitrogen, total phosphorus, pH, and specific conductivity) and "hydro‐atmospheric" exchange (stream flow and atmospheric pressure) components. The predominant climatic control on stream DO was linked to the high extent of vegetated land (on average ∼53%) and steep slope (∼10%) in the draining watersheds, despite the notable presence of agricultural land (∼35%). An emergent power law scaling relationship was then developed to acceptably predict DO (mg/l) based on Tw (K) and pH, with the approximate exponents of −15/2 and 1/2, respectively (Nash‐Sutcliffe Efficiency = 0.72–0.73). The scaling law demonstrated the underlying organizing principles such as the depletion of stream DO due to reduced dissolution and increased metabolic respiration with the increasing temperature and nutrients. The scaling law was persistent across the various U.S. streams, representing gradients in climate, hydrology, biogeochemistry, and land use/cover. The findings would help understand and manage water quality and ecosystem health in freshwater streams across the United States and beyond. Plain Language Summary: Dissolved oxygen (DO) is an overall indicator of stream water quality and ecosystem health. We present here the environmental controls and scaling law of DO in 86 freshwater streams located across the United States. We found a dominant climatic control on stream DO, compared to the hydrologic and biogeochemical controls. The climatic control was linked to a high extent of watershed vegetated land and steep slope, despite a notable presence of agricultural land. A generalized scaling law was developed to predict stream DO (mg/l) based on water temperature (Kelvin) and pH (potential of hydrogen), with the approximate exponents of −15/2 and 1/2, respectively. The scaling law was applicable in diverse freshwater streams across United States. The findings would help understand, predict, and manage water quality and ecosystem health in freshwater streams around the world. Key Points: Climatic control on stream DO is 3 and 9 times stronger than the biogeochemical and hydro‐atmospheric controls, respectivelyHigh percentage of vegetated land and steep slope of draining watersheds contribute to the predominant climatic control of stream DOStream DO (mg/l) represents an emergent power law scaling with Tw (K) and pH having approximate exponents of −15/2 and 1/2, respectively [ABSTRACT FROM AUTHOR]

In this paper we seek to understand the nature of flood spatial dependence over the conterminous United States. We extend an existing conditional multivariate statistical model to enable its application to this large and heterogenous region and apply it to a 40‐year data set of ~2,400 U.S. Geological Survey gauge series records to simulate 1,000 years of U.S. flooding comprising more than 63,000 individual events with realistic spatial dependence. A continental‐scale hydrodynamic model at 30 m resolution is then used to calculate the economic loss arising from each of these events. From this we are able to compute the probability that different values of U.S. annual total economic loss due to flooding are exceeded (i.e., a loss‐exceedance curve). Comparing these data to an observed flood loss‐exceedance curve for the period 1988–2017 shows a reasonable match for annual losses with probability below 10% (e.g., >1 in 10‐year return period). This analysis suggests that there is a 1% chance of U.S. annual fluvial flood losses exceeding $78Bn in any given year, and a 0.1% chance of them exceeding $136Bn. Analysis of the set of stochastic events and losses yields new insights into the nature of flooding and flood risk in the United States. In particular, we confirm the strong relationship between flood affected area and event peak magnitude, but show considerable variability in this relationship between adjacent U.S. regions. The analysis provides a significant advance over previous national flood risk analyses as it gives the full loss‐exceedance curve instead of simply the average annual loss. Plain Language Summary: Traditional flood risk analyses make the assumption that flow probability (the chance that a given river discharge is exceeded) does not vary within river catchments within an event. Real floods, however, do not look like this: In some places flooding is more severe than in others. Over a few tens of kilometers of river assuming the same event return period everywhere is perfectly fine, but over larger areas it breaks down. At national scales traditional risk analyses can only estimate the average annual loss. To estimate the total annual losses that might occur in more extreme flooding years the risk analysis needs to be based on more realistic spatial patterns of flooding. In this paper we use a sophisticated statistical model, based on U.S. Geological Survey river flow data, to simulate 1,000 years of spatially realistic U.S. flooding comprising more than 63,000 individual events. By calculating the damage for each event as a dollar value, we are able to estimate the probability of the United States experiencing particular levels of annual flood losses. We show that there is a 1% chance of U.S. annual fluvial flood losses exceeding $78Bn in any given year, and a 0.1% chance of them exceeding $136Bn. Key Points: 1,000 years of realistic U.S. flood patterns, comprising >63,000 individual events, are simulated using a statistical modelMonetary losses for each event are calculated using a continental hydrodynamic model at 30 m resolutionThe analysis suggests that there is a 1% chance of U.S. annual fluvial flood losses exceeding $78Bn in any given year [ABSTRACT FROM AUTHOR]

In the US, the cost of water and wastewater services is rising three‐times faster than inflation. Over the next 20–25 years, required investments in water infrastructure are estimated to exceed $1 trillion, further increasing service costs. Combined with stagnating income levels, especially for poor households, increased costs will likely aggravate water affordability issues. Here, we document the extent of water affordability concerns in the US across income, geography, and race. We find that 10% of households face water affordability concerns, defined as expenditures on essential water and sewer services greater than 4.5% of annual household income. Households in the lowest income decile pay on average 6.8% of their annual income on water and sewer service. Our estimates are based on a large‐scale data set on water and sewer rates matched with Census block‐group‐level socioeconomic characteristics and covering approximately 45% of the US population. We demonstrate that using median household income at the county level drastically understates the extent of the water affordability problem. Additionally, we find that the number of households facing affordability concerns is positively associated with the proportion of black residents and negatively associated with Hispanic residents even after conditioning on prices and poverty levels. Lastly, we show that self‐sufficient water affordability policies that provide a lump‐sum rebate to low‐income households and are paid for by income taxes redistribute the burden borne by low‐income customers with fewer unintended consequences for non‐essential consumption than policies that change marginal incentives for water and sewer consumption. Plain Language Summary: Provision of affordable water and sewer service is a growing concern in the United States, although the extent of the problem is not known and the effectiveness of different policy options are underexplored. We compile a database of water and sewer prices for approximately 45% of the United States population to estimate annual expenditures on water and sewer service. We find that nearly one in ten households spend more than 4.5% of their annual household income on water and sewer service, and that affordability concerns are correlated with race after conditioning on poverty levels. Our results have implications for how to characterize water affordability and for how corrective distributional policies can be designed most effectively. Key Points: One in ten households have essential water and sewer expenditures exceeding 4.5% of their incomeThose in the lowest decile of income spend on average 6.8% of their income on essential water and sewer serviceAffordability policies that provide lump‐sum rebates to low‐income households are less distortionary than lowering marginal water prices [ABSTRACT FROM AUTHOR]

In mountains, the precipitation phase greatly varies in space and time and affects the evolution of the snow cover. Snowpack models usually rely on precipitation‐phase partitioning methods (PPMs) that use near‐surface variables. These PPMs ignore conditions above the surface thus limiting their ability to predict the precipitation phase at the surface. In this study, the impact on snowpack simulations of atmospheric‐based PPMs, incorporating upper atmospheric information, is tested using the snowpack scheme Crocus. Crocus is run at 2.5‐km grid spacing over the mountains of southwestern Canada and northwestern United States and is driven by meteorological fields from an atmospheric model at the same resolution. Two atmospheric‐based PPMs were considered from the atmospheric model: the output from a detailed microphysics scheme and a post‐processing algorithm determining the snow level and the associated precipitation phase. Two ground‐based PPMs were also included as lower and upper benchmarks: a single air temperature threshold at 0°C and a PPM using wet‐bulb temperature. Compared to the upper benchmark, the snow‐level based PPM improved the estimation of snowfall occurrence by 5% and the simulation of snow water equivalent (SWE) by 9% during the snow melting season. In contrast, due to missing processes, the microphysics scheme decreased performances in phase estimate and SWE simulations compared to the upper benchmark. These results highlight the need for detailed evaluation of the precipitation phase from atmospheric models and the benefit for mountain snow hydrology of the post‐processed snow level. The limitations to drive snowpack models at slope scale are also discussed. Plain Language Summary: The partitioning of precipitation between rainfall and snowfall is a crucial component of the evolution of the snowpack in mountains. Most snowpack models use the air temperature and humidity near the surface to derive the precipitation phase. However, the phase at the surface is strongly influenced by processes such as melting and refreezing of falling hydrometeors that occur above the surface. Atmospheric models simulate these processes and the corresponding phase at the surface. However, snowpack models rarely use this information. In this study, we considered two estimates of precipitation phase from an atmospheric model and tested them with a physically‐based snow model over the mountains of southwestern Canada and northwestern United States. The results were compared with traditional approaches using the air temperature and humidity near the surface to derive the precipitation phase. Our results showed that the precipitation phase associated with the snow level obtained from the atmospheric model improved snowfall estimate and snowpack prediction compared to the traditional approaches. In contrast, the cloud/precipitation scheme of the atmospheric model decreased performance in phase estimate and snow simulations due to missing physical processes. Our study highlights that snowpack predictions in the mountains can be improved if valuable information is obtained from atmospheric models. Key Points: Estimates of precipitation phase from an atmospheric model were used to drive snow simulations with a detailed snowpack modelSnowfall prediction and snowpack modeling are improved by using the snow level from post‐processing of the atmospheric modelDirect precipitation phase from the microphysics scheme does not improve snow simulations compared to simpler rain‐snow partitioning schemes [ABSTRACT FROM AUTHOR]

This study explores multi‐scale variability in the relationship between turbidity (Tn) and flow (Q) in 162 watersheds across the contiguous United States. Sites are selected where Tn acts as a good surrogate for suspended sediment concentration. We use dynamic linear models (DLMs) to infer time‐varying parameters of Tn‐Q rating curves at each site, and calibrate a hyper‐parameter of the DLM model (δ $\delta $) to quantify the degree of dynamicity in the rating curve relationship. The DLM can capture dynamics in the Tn‐Q relationship at the resolution of the data (daily in this study), enabling an analysis of the dynamics across time scales. Regional multivariate regressions are used to identify physiographic features that relate to the magnitude of δ $\delta $ and spectral signatures in the DLM parameters across sites. Results show that watersheds in the Midwest and Pacific Northwest tend to exhibit more variable Tn‐Q relationships, while these relationships are more stable in watersheds in the humid east and Lower Mississippi River basin. Stream network complexity, soil composition, perennial snow coverage, saturation‐excess overland flow, and modifications to the stream network are associated with the dynamicity in the Tn‐Q relationship. DLM parameters exhibit cyclic behavior at sub‐monthly, sub‐annual, and annual time scales at sites across the country, with annual cycling associated with basin features that reflect watershed sediment availability and the erosive power of rivers. Overall, our analysis highlights significant multi‐scale variability in Tn‐Q relationships across the nation, with important implications for how sediment dynamics should be measured and managed at the watershed‐scale. Plain Language Summary: Effective sediment management often requires accurate sediment yield predictions and forecasts. However, the predictive skill of various models often declines in regions where the sediment transport processes are variable across timescales. This study aims to improve our understanding of variability in the relationship between turbidity (Tn) and flow (Q) through time, focusing on sites where turbidity is a good proxy for suspended sediment concentration. We use regression models with parameters that can vary over time to quantify the variability in the Tn‐Q relationship and to highlight the presence of seasonal patterns in transport processes. We then explore how watershed characteristics influence the variability and seasonality in the Tn‐Q relationship. Our results suggest that stream network complexity, soil composition, perennial snow coverage, saturation‐excess overland flow, and modifications to the stream network influence the overall variability in the Tn‐Q relationship, while the seasonality is mainly associated with basin features that reflect watershed sediment availability and the erosive power of rivers. The results of the study have important implications for how sediment dynamics should be addressed in watershed management studies. Key Points: The dynamicity of sediment transport processes is quantified using dynamic linear rating curve models of turbidity and flowThe degree of rating curve dynamicity varies spatially and is partially influenced by natural and anthropogenic physiographic featuresDynamic rating curves exhibit cyclic behavior at various timescales across the US, which is also partially explained by basin features [ABSTRACT FROM AUTHOR]

Stochastic precipitation generators (SPGs) are often used to produce synthetic precipitation series for water resource management. Typically, an SPG assumes a stationary climate. We present an hourly precipitation generation algorithm for nonstationary conditions informed by the global climate model (GCM) forecasted average monthly temperature (AMT). The physical basis for precipitation formation is considered explicitly in the design of the algorithm using hourly pressure change events (PCE) to define the relationship between hourly precipitation and AMT. The algorithm consists of a multivariable Markov Chain and a moving window driven by time, temperature, and pressure change. We demonstrate the methodology by generating a 100‐year, continuous, synthetic hourly precipitation time series using GCM AMT projections for the Northeast United States. When compared with historical observations, the synthetic results suggest that future precipitation in this region will be more variable, with more frequent mild events and fewer but intensified extremes, especially in warm seasons. The synthetic time series suggests that there will be less precipitation in the summers, while winters will be wetter, consistent with other research on climate change projections for the Northeast United States. This SPG provides physically plausible weather ensembles for water resource studies involving climate change. Key Points: This article introduces a nonparametric stochastic process in generating synthetic precipitations for nonstationary climateThe basis of this algorithm is highly related to the physical mechanism of precipitation formation defined by pressure change events [ABSTRACT FROM AUTHOR]

The aim of this paper is to improve semiseasonal forecast of groundwater availability in response to climate variables, surface water availability, groundwater level variations, and human water management using a two‐step data‐driven modeling approach. First, we implement an ensemble of artificial neural networks (ANNs) for the 300 wells across the High Plains aquifer (USA). The modeling framework includes a method to choose the most relevant input variables and time lags; an assessment of the effect of exogenous variables on the predictive capabilities of models; and the estimation of the forecast skill based on the Nash‐Sutcliffe efficiency (NSE) index, the normalized root mean square error, and the coefficient of determination (R2). Then, for the ANNs with low‐ accuracy, a MultiModel Combination (MuMoC) based on a hybrid of ANN and an instance‐based learning method is applied. MuMoC uses forecasts from neighboring wells to improve the accuracy of ANNs. An exhaustive‐search optimization algorithm is employed to select the best neighboring wells based on the cross correlation and predictive accuracy criteria. The results show high average ANN forecasting skills across the aquifer (average NSE > 0.9). Spatially distributed metrics of performance showed also higher error in areas of strong interaction between hydrometeorological forcings, irrigation intensity, and the aquifer. In those areas, the integration of the spatial information into MuMoC leads to an improvement of the model accuracy (NSE increased by 0.12), with peaks higher than 0.3 when the optimization objectives for selecting the neighbors were maximized.tT Key Points: Artificial neural networks can accurately forecast semiseasonal groundwater level changesMuMoC improved the groundwater well‐level forecasting skill for 1‐ to 4‐month lead times with respect to a single ANN model by 25% in NSEThe implementation of MuMoC is recommended in case of densely gauged areas [ABSTRACT FROM AUTHOR]

Household willingness to pay (WTP) for reductions in water hardness was estimated using a defensive investment framework. Using total dissolved solids data observed in municipal water supplies in combination with product‐by‐store level point‐of‐sale scanner data on consumable water softening product sales, we reveal preference estimates of WTP for water hardness reduction and quality control. Using instrumental variable regressions, we show that household marginal WTP increases as the observed total dissolved solids in municipal water increases. Estimations show that on average a household is willing to pay approximately $12 annually to reduce municipal water hardness, which is around 4% of the average annual water utility bill in the United States. Aggregated county WTP estimates vary geographically. Households have a nonnegligible WTP for water hardness reduction, which has important policy implications for optimal water hardness management by municipal water authorities and those policies aimed to target salinity management within surface and subsurface water supplies. Particularly, this paper informs municipal water managers of the demand side of water hardness issues and that desalination can be a solution to solve water hardness issues in some areas but not everywhere. Key Points: Based on a random utility framework, we show household willingness to pay for reductions in municipal water hardnessResults contribute to the literature on point‐of‐use water treatment and our understanding of desalinationOur estimates provide a better understanding of the demand for water softness and desalination management [ABSTRACT FROM AUTHOR]

Streamflow drought is a recurring challenge, and understanding spatiotemporal patterns of past droughts is needed to manage future water resources. We examined regional patterns in streamflow drought metrics and compared these metrics to low flow timing and magnitude using long‐term daily records for 555 minimally disturbed watersheds. For each streamgage, we calculated streamflow drought duration (number of days) and deficit (flow volume below a specified threshold) for each climate year (April 1–March 31). We identified drought using five thresholds (2%–30%) and two approaches: variable thresholds with unique values for each day of the year, and a fixed threshold based on all period‐of‐record flows. We then analyzed drought trends using the Mann‐Kendall test with persistence adjustment for 1921–2020, 1951–2020, and 1981–2020, and computed correlations between annual streamflow drought metrics and climate metrics using values from a monthly water balance model. Spatial patterns in drought metrics were consistent between variable and fixed approaches, though fixed threshold durations were typically longer and variable threshold deficits larger. High interannual variability in drought duration emerged in the central, interior west, and southwestern U.S., with high deficit variability in the interior west. Drought metrics were weakly correlated with low flow magnitude and timing, providing unique information. Drought duration and deficit increased in the southern and western U.S. for both 1951–2020 and 1981–2020, particularly using fixed thresholds, and paralleled trends in aridity. Projections of continued aridification for the southern and western U.S. may increase drought durations and deficits and intensify water availability impacts. Key Points: Streamflow drought metrics show similar spatial patterns using fixed and variable thresholds, with unique information versus 7‐day low flowStreamflow drought duration and deficit increased in the south and west U.S., with more increases when using fixed drought thresholdsStreamflow drought trends were correlated to aridity trends; continued west U.S. aridification may increase streamflow drought impacts [ABSTRACT FROM AUTHOR]

Accurate baseflow estimation is essential for ecological protection and water resources management. Past studies have used environmental predictors to extend baseflow from gauged basins to ungauged basins, publishing several regional or global datasets on mean annual baseflow. However, time series datasets of baseflow are still lacking due to the complexity of baseflow generation processes. Here, we developed a monthly baseflow data set using a Deep learning model called the long short‐term memory (LSTM) networks. To better train the networks across basins, we compared the standard LSTM architecture using 8 time series as inputs with four variant architectures using 16 additional static properties as inputs. Dividing the contiguous United States into nine ecoregions, we used baseflow calculated from 1,604 gauged basins as training targets to calibrate the five LSTM architectures for each ecoregion separately. Results show that three variant architectures (Joint, Front, and Entity‐Aware‐LSTM) perform better than the standard LSTM, with median Kling‐Gupta Efficiencies across basins greater than 0.85. Based on Front LSTM, the monthly baseflow data set with 0.25° spatial resolution across the contiguous United States from 1981 to 2020 was obtained. Potential applications of the data set include analyzing baseflow trends under global change and estimating large‐scale groundwater recharge. Plain Language Summary: After precipitation has ceased for some time, upstream banks slowly release groundwater into streams through delayed pathways. This type of flow is known as baseflow. Baseflow is important because it maintains streamflow during prolonged droughts, providing water supplies for food production and fish habitat. Several methods have been developed to estimate baseflow from streamflow. However, due to the complex processes of baseflow generation, it is challenging to estimate baseflow for a basin where streamflow is unavailable. This study introduced a deep learning model called Long Short‐Term Memory (LSTM) network to bridge the gap. To better train across basins, we compared the standard LSTM with four variant architectures that incorporate additional static properties as inputs. Results show that three variants (Joint, Front, and Entity‐Aware‐LSTM) perform better than the standard LSTM, with median Kling‐Gupta Efficiencies across basins greater than 0.85. Training the Front LSTM separately in nine ecoregions of the contiguous United States, we identified the relationship between 24 spatiotemporal features and the baseflow calculated from streamflow in 1,604 basins, thereby deriving a monthly baseflow data set at 0.25° spatial resolution across the contiguous United States. This data set provides an opportunity to analyze large‐scale baseflow trends and is beneficial for groundwater estimation. Key Points: We develop a monthly baseflow data set at 0.25° resolution using Deep learning networks (Long Short‐Term Memory, LSTM) with 8 dynamic time series as inputsThree LSTM variants using 16 static properties as additional inputs outperform the standard LSTM, with median Kling‐Gupta Efficiencies across 1,604 basins greater than 0.85Dynamic inputs interpret 77.8% in baseflow simulation, with P and potential evapotranspiration contributing the most, while slope dominates the static inputs [ABSTRACT FROM AUTHOR]

Satellite imagery is commonly used to map surface water extents over time, but many approaches yield discontinuous records resulting from cloud obstruction or image archive gaps. We applied the Dynamic Surface Water Extent (DSWE) model to downscaled (250‐m) daily Moderate Resolution Imaging Spectroradiometer (MODIS) data in Google Earth Engine to generate monthly surface water maps for the conterminous United States (US) from 2003 through 2019. The aggregation of daily observations to monthly maps of maximum water extent produced records with diminished cloud and cloud shadow effects across most of the country. We used the continuous monthly record to analyze spatiotemporal surface water trends stratified within Environmental Protection Agency Ecoregions. Although not all ecoregion trends were significant (p < 0.05), results indicate that much of the western and eastern US underwent a decline in surface water over the 17‐year period, while many ecoregions in the Great Plains had positive trends. Trends were also generated from monthly streamgage discharge records and compared to surface water trends from the same ecoregion. These approaches agreed on the directionality of trend detected for 54 of 85 ecoregions, particularly across the Great Plains and portions of the western US, whereas trends were not congruent in select western deserts, the Great Lakes region, and the southeastern US. By describing the geographic distribution of surface water over time and comparing these records to instrumented discharge data across the conterminous US, our findings demonstrate the efficacy of using satellite imagery to monitor surface water dynamics and supplement traditional instrumented monitoring. Plain Language Summary: Daily satellite images can be used to create cloud‐free maps of surface water for most of the United States (US). An analysis of these maps over multiple decades reveals surface water change in large‐scale water features such as lakes, reservoirs, and major rivers. Results show that increasing surface water trends are more common across the central US while decreasing trends are present across much of the western and eastern regions over the past 17 years. More specifically, we observed certain seasonal trends including increasing spring flooding along the Mississippi River. Comparing these data to trends in streamgage discharge provides additional insights into certain regions where surface water changes are more directly related to dynamic surface water flows. Results from this study identify national and regional surface water variability over space and time that can contribute to more effective water resource management. Key Points: Monthly Moderate Resolution Imaging Spectroradiometer maps can be used to calculate and compare surface water area trends across US ecoregionsThe central United States (US) underwent an increase in surface water while most of the western/eastern US underwent a decline from 2003 through 2019Monthly trends in surface water were congruent with monthly trends in discharge for the majority of Level III Ecoregions in the US [ABSTRACT FROM AUTHOR]

Precipitation prediction at seasonal timescales is important for planning and management of water resources as well as preparedness for hazards such as floods, droughts and wildfires. Quantifying predictability is quite challenging as a consequence of a large number of potential drivers, varying antecedent conditions, and small sample size of high‐quality observations available at seasonal timescales, that in turn, increases prediction uncertainty and the risk of model overfitting. Here, we introduce a generalized probabilistic framework to account for these issues and assess predictability under uncertainty. We focus on prediction of winter (Nov–Mar) precipitation across the contiguous United States, using sea surface temperature‐derived indices (averaged in Aug–Oct) as predictors. In our analysis we identify "predictability hotspots," which we define as regions where precipitation is inherently more predictable. Our framework estimates the entire predictive distribution of precipitation using copulas and quantifies prediction uncertainties, while employing principal component analysis for dimensionality reduction and a cross validation technique to avoid overfitting. We also evaluate how predictability changes across different quantiles of the precipitation distribution (dry, normal, wet amounts) using a multi‐category 3 × 3 contingency table. Our results indicate that well‐defined predictability hotspots occur in the Southwest and Southeast. Moreover, extreme dry and wet conditions are shown to be relatively more predictable compared to normal conditions. Our study may help with water resources management in several subregions of the United States and can be used to assess the fidelity of earth system models in successfully representing teleconnections and predictability. Key Points: We present a rigorous probabilistic framework and novel statistical metrics to assess the predictability of seasonal precipitation totalsHotspots of winter precipitation predictability occur in the Southwest and Southeast of contiguous United States (CONUS)Sea Surface Temperature (SST) information may help with predicting dry or wet conditions in some areas, while normal conditions are difficult to predict [ABSTRACT FROM AUTHOR]

Clean water from forests is commonly used to supply drinking water to communities both within and outside basin boundaries through inter‐basin transfers (IBTs). Here, we modified the Water Supply Stress Index (WaSSI) model to provide estimates of mean water yield and the proportion of mean flow originating on forested lands at the 12‐Digit Hydrologic Unit Code scale across the conterminous United States (CONUS). We accounted for the benefits of forests for drinking water supply and receiving populations through IBTs by incorporating a new IBT database, surface intake location information for public drinking water systems, and modeled water yield from forests. We compiled the new database of 594 IBTs ranging from 0.01 million m3 yr−1 to 8,900 million m3 yr−1, for a total transferred volume of 116,894 million m3 yr−1. According to our results, forested lands comprised 28.7% of the total land area across CONUS, but contributed 46% of the total surface water yield. Approximately 125.5 million people derived more than 10% of their surface drinking water supply from forested lands, and 83.1 million people received more than 50% of their surface drinking water supply from forested lands. Of those 83.1 million people receiving more than 50% of their surface drinking water supply from forested lands, 19.4 million people obtained some (≥0.01%) of that water through IBTs. We conclude that accounting for IBTs is critical to accurately assess the contribution of forested watersheds for surface drinking water supply. Hydrologic models for assessment and decision making must include IBTs to fully account for the effects of climate change and human population dynamics on water resource availability at watershed to regional scales. Results from this study can aid water resource and forest managers in developing integrated watershed management plans at a time when climate change, population growth, and land use change threaten water supplies. Key Points: National Forest System and other forested lands comprised 28.7% of the total land area but contributed 46.0% of the total water yieldApproximately 125.5 million and 83.1 million people received more than 10% and 50% of their surface drinking water supply from forested lands, respectivelyWithout considering inter‐basin transfer, benefits of forests to water supply would be underestimated [ABSTRACT FROM AUTHOR]

Artificial levees are anthropogenic structures designed to hydrologically disconnect rivers from floodplains. The extent of artificial levees in the contiguous United States (CONUS) is unknown. To better estimate the distribution of artificial levees, we tested several different geomorphic, land cover, and spatial variables developed from the National Elevation Dataset, the National Land Cover database, and the National Hydrology Dataset HR Plus. We used known levee locations from the National Levee Database as training data. We tested machine learning and general logistic models' ability to detect artificial levees in a 100‐year hydrogeomorphic floodplain of seven geographically diverse 8‐digit HUC basins. Random forest models outperformed other models in predicting the location of levees using variables representing geomorphic attributes, land cover, and distance from streams ranging in size between stream order one through six. To demonstrate the ability of our approach to detect unknown levees, we conducted a leave‐one‐out cross‐validation in the lower Mississippi Basin using approximately 1,100 artificial levees. This approach detected known levees constituting 94% of the total levee length in the basin. Scaling up to the CONUS, we applied a high performing (overall accuracy of 97%) random forest model using land cover and stream order variables. We detected 182,213 km of potential levees, mostly along streams of order 2–6 in the Mississippi and Missouri River Basins, indicating that the national levee database contains 20.4% of levee length. Potential levees and those documented in the national levee database modify 2% of the total length of streams in the contiguous United States. Plain Language Summary: There has been a lot of research exploring how humans have impacted river systems to include impacts from dams and roads at national and global scales. However, the study of artificial levees has been limited to local studies due to incomplete databases of levee location. Artificial levees are linear dirt mounds, built next to rivers to stop flooding of property built on floodplains. Because the location of every artificial levee is not known in the US, we do not know to what extent floodplains have been separated from their rivers. To ameliorate this, our study explores different methods to detect unknown levees and applies an algorithm that is 97% accurate to the contiguous U.S. Most levees are built within floodplains, so we limited our study to the 100‐year floodplain. We tested different types of algorithms in a limited study with variables explaining levee shape, human land cover, and distance from streams of different sizes. Then, we applied the most effective models to the contiguous U.S. We located over 182,000 km of potential levees, potential areas with artificial levees not identified in the NLD, most of which were built on smaller streams and concentrated in the Mississippi and Missouri River basins. Key Points: An incomplete database of artificial levees inhibits the study of artificial levee impacts to floodplain extent in the contiguous USADifferent methods of detecting artificial levees are tested in case study basins and in the contiguous USAThis study detects 182,000 km of potential levees, along smaller streams in the Mississippi and Missouri Basins [ABSTRACT FROM AUTHOR]

Subsurface tile drainage (TD) is a dominant agriculture water management practice in the United States (US) to enhance crop production in poorly drained soils. Assessments of field‐level or watershed‐level (<50 km2) hydrologic impacts of TD are becoming common; however, a major gap exists in our understanding of regional (>105 km2) impacts of TD on hydrology. The National Water Model (NWM) is a distributed 1‐km resolution hydrological model designed to provide accurate streamflow forecasts at 2.7 million reaches across the US. The current NWM lacks TD representation which adds considerable uncertainty to streamflow forecasts in heavily tile‐drained areas. In this study, we quantify the performance of the NWM with a newly incorporated tile‐drainage scheme over the heavily tile‐drained Midwestern US. Employing a TD scheme enhanced the uncalibrated NWM performance by about 20–50% of the fully calibrated NWM (Calib). The calibrated NWM with tile drainage (CalibTD) showed enhanced accuracy with higher event hit rates and lower false alarm rates than Calib. CalibTD showed better performance in high‐flow estimations as TD increased streamflow peaks (14%), volume (2.3%), and baseflow (11%). Regional water balance analysis indicated that TD significantly reduced surface runoff (−7% to −29%), groundwater recharge (−43% to −50%), evapotranspiration (−7% to −13%), and soil moisture content (−2% to −3%). However, TD significantly increased soil profile lateral flow (27.7%) along with infiltration and soil water storage potential. Overall, our findings highlight the importance of incorporating the TD process into the operational configuration of the NWM. Key Points: A new subsurface tile drainage module is incorporated into the National Water Model (NWM) to predict streamflow over the tile‐drained areasNWM with a tile drainage module can predict high‐flows and streamflow peaks better than the original NWM over heavily tile‐drained areasIncorporating tile drainage into the NWM considerably enhanced the streamflow event hit rates and reduced false alarm rates [ABSTRACT FROM AUTHOR]

Deep learning (DL) models can accurately predict many hydrologic variables including streamflow and water temperature; however, these models have typically predicted hydrologic variables independently. This study explored the benefits of modeling two interdependent variables, daily average streamflow and daily average stream water temperature, together using multi‐task DL. A multi‐task scaling factor controlled the relative contribution of the auxiliary variable's error to the overall loss during training. Our experiments examined the improvement in prediction accuracy of the multi‐task approach using paired streamflow and water temperature data from sites across the conterminous United States. Our results showed that for 56 out of 101 sites, the best performing multi‐task models performed better overall than the single‐task models in terms of Nash‐Sutcliffe efficiency for predicting streamflow with single‐site models. For 43 sites, the best multi‐task, single‐site models made no significant difference in predicting streamflow. The multi‐task approach had a smaller effect when applied to a model trained with data from 101 sites together, significantly improving performance for only 17 sites. The multi‐task scaling factor was consequential in determining to what extent the multi‐task approach was beneficial. A naïve selection of this factor led to significantly worse‐performing models for 3 of 101 sites when predicting streamflow as the primary variable, and 47 of 53 sites when predicting stream temperature as the primary variable. We conclude that a multi‐task approach can make more accurate predictions by leveraging information from interdependent hydrologic variables, but only for some sites, variables, and model configurations. Key Points: A single deep learning model was used to predict both water temperature and streamflowThe best configured single‐site multi‐task models improved streamflow predictions for most sites testedA naïve implementation of multi‐task learning was detrimental to water temperature predictions [ABSTRACT FROM AUTHOR]

Access to credible estimates of water use is critical for making optimal operational decisions and investment plans to ensure reliable and affordable provisioning of water. Furthermore, identifying the key predictors of water use is important for regulators to promote sustainable development policies to reduce water use. In this paper, we propose a data‐driven framework, grounded in statistical learning theory, to develop a rigorously evaluated predictive model of state‐level, per capita water use in the United States as a function of various geographic, climatic, and socioeconomic variables. Specifically, we compare the accuracy of various statistical methods in predicting the state‐level, per capita water use and find that the model based on the random forest algorithm outperforms all other models. We then leverage the random forest model to identify key factors associated with high water‐usage intensity among different sectors in the United States. More specifically, irrigated farming, thermoelectric energy generation, and urbanization were identified as the most water‐intensive anthropogenic activities, on a per capita basis. Among the climate factors, precipitation was found to be a key predictor of per capita water use, with drier conditions associated with higher water usage. Overall, our study highlights the utility of leveraging data‐driven modeling to gain valuable insights related to the water use patterns across expansive geographical areas. [ABSTRACT FROM AUTHOR]

The location of a point on the landscape within a stream network (hydrologic position) can be an important predictive measure in hydrology. Hydrologic position is defined here by two metrics: lateral position and distance from stream to divide, both measured horizontally. Lateral position (dimensionless) is the relative position of a point between the stream and its watershed divide. Distance from stream to divide (units of length) is an indicator of position within a watershed: generally small near a confluence and generally large in headwater areas. Watersheds and watershed divides are defined here by Thiessen polygons rather than topographic divides. Lateral position and distance from stream to divide are also defined in the context of hydrologic order. Hydrologic order "n" is defined as the network of streams, and associated divides, of order n and higher. And given that a point can have different positions in different hydrologic orders the term multiorder hydrologic position (MOHP) is used to describe the ensemble of hydrologic positions. MOHP was mapped across the conterminous United States for nine hydrologic orders at a spatial resolution of 30 m (about 8.7 billion pixels). There are 18 metrics for each pixel. Four case studies are presented that use MOHP metrics as explanatory factors in random forest machine learning models. The case studies show that lower order MOHP metrics can serve as indicators of hydrologic process while higher‐order metrics serve as indicators of location. MOHP is shown to have utility as a predictor variable across a large range of scales (50,000 to 8,000,000 km2). Plain Language Summary: In hydrology, as in other endeavors, location matters. This study presents a new type of data that describes the location of a point on the landscape in the context of the network of streams that are present across the continental United States. The new data are presented as maps, and different patterns can be recognized in different areas of the United States. The patterns that can be seen also vary as one looks more or less closely at an area. The patterns that are present in these maps are shown to be useful for the purposes of mapping water resources. Key Points: Multiorder hydrologic position (lateral position and distance from stream to divide) quantifies the location of a point on the landscapeMultiorder hydrologic position is mapped for nine hydrologic orders at a 30‐m resolution for the conterminous United StatesMultiorder hydrologic position is used in machine learning models for the purposes of physiographic and hydrologic mapping [ABSTRACT FROM AUTHOR]

Using 2015 American Housing Survey data, this study generates the first national estimates of the use of drinking water alternatives among households who perceive their tap water to be unsafe. Multivariate regression models examine factors influencing both perception of tap water and the choice of tap alternatives. We find that minority households more commonly perceive their tap water to be unsafe and choose bottled water when they perceive the tap as unsafe. While perception of unsafe tap water is most prevalent among Hispanic households, among the population perceiving their tap to be unsafe, Black households more commonly buy bottled water. We estimate the minimum annual expenditure by all U.S. households who perceive the tap as unsafe and buy bottled water to meet necessary consumption standards at $5.65 billion annually. Novel secondary treatment interventions and education campaigns are needed to address the perception of tap water as unsafe and consequent alternative source reliance, while additional research is needed to document the active trade‐offs households make to manage their water sources, and to better understand determinants of the perception of tap water as unsafe within minority groups. Key Points: This paper uses data from the 2015 American Housing Survey to examine the prevalence of U.S. households' perception of tap drinking water as unsafe and the consequential choices of alternative primary drinking water sources by those householdsThe analysis confirms the results of previous research which showed that Hispanic households more commonly perceive their tap water to be unsafe, but also contributes the novel finding that among households who perceive tap water as unsafe, African American households more commonly choose bottled water as an alternative drinking water sourceThe results suggest that the economic consequences of household perception of drinking water as unsafe are borne disproportionately by already disadvantaged minority groups, and emphasize the need for targeted policy interventions [ABSTRACT FROM AUTHOR]

Surface water is the most readily accessible water resource and provides an array of ecosystem services, but its availability and access are stressed by changes in climate, land cover, and population size. Understanding drivers of surface water dynamics in space and time is key to better managing our water resources. However, few studies estimating changes in surface water account for climate and anthropogenic drivers both independently and together. We used 19 years (2000–2018) of the newly developed Dynamic Surface Water Extent Landsat Science Product in concert with time series of precipitation, temperature, land cover, and population size to statistically model maximum seasonal percent surface water area as a function of climate and anthropogenic drivers in the southeastern United States. We fitted three statistical models (linear mixed effects, random forests, and mixed effects random forests) and three groups of explanatory variables (climate, anthropogenic, and their combination) to assess the accuracy of estimating percent surface water area at the watershed scale with different drivers. We found that anthropogenic drivers accounted for approximately 37% more of the variance in the percent surface water area than the climate variables. The combination of variables in the mixed effects random forest model produced the smallest mean percent errors (mean −0.17%) and the highest explained variance (R2 0.99). Our results indicate that anthropogenic drivers have greater influence when estimating percent surface water area than climate drivers, suggesting that water management practices and land‐use policies can be highly effective tools in controlling surface water variations in the Southeast. Plain Language Summary: People and the environment rely on water to exist and thrive, especially water on the Earth's surface because that is the easiest place to get it. The amount of surface water and where it is located is changing with the climate and changes in people's water use, and our need for it is increasing. To plan ahead for future water needs, we need to better understand how the climate and people are changing surface water patterns both separately and together. To help improve our understanding of these changes, we modeled the amount of surface water in three different ways. First, we modeled based on climate data (like temperature and precipitation); second, based on human data (like land use and population); and third, based on both climate and human data together. We found that we could best model the amount of surface water if we used both climate and human data together, and that human data can explain a lot of the changes in the amount of surface water. These results mean that we can work to control changes in the amount of surface water by controlling human actions through planning and policies. Key Points: We developed top‐down data‐driven models to estimate percent surface water area based on Landsat imagery and climate and human driversCompared to estimates based on climate and human drivers independently, combining them reduces percent surface water area estimation errorForest‐dominated land cover was the most influential explanatory variable to estimate percent surface water area across all our variables [ABSTRACT FROM AUTHOR]

Model calibration and validation are critical in hydrological model robustness assessment. Unfortunately, the commonly used split‐sample test (SST) framework for data splitting requires modelers to make subjective decisions without clear guidelines. This large‐sample SST assessment study empirically assesses how different data splitting methods influence post‐validation model testing period performance, thereby identifying optimal data splitting methods under different conditions. This study investigates the performance of two lumped conceptual hydrological models calibrated and tested in 463 catchments across the United States using 50 different data splitting schemes. These schemes are established regarding the data availability, length and data recentness of continuous calibration sub‐periods (CSPs). A full‐period CSP is also included in the experiment, which skips model validation. The assessment approach is novel in multiple ways including how model building decisions are framed as a decision tree problem and viewing the model building process as a formal testing period classification problem, aiming to accurately predict model success/failure in the testing period. Results span different climate and catchment conditions across a 35‐year period with available data, making conclusions quite generalizable. Calibrating to older data and then validating models on newer data produces inferior model testing period performance in every single analysis conducted and should be avoided. Calibrating to the full available data and skipping model validation entirely is the most robust split‐sample decision. Experimental findings remain consistent no matter how model building factors (i.e., catchments, model types, data availability, and testing periods) are varied. Results strongly support revising the traditional split‐sample approach in hydrological modeling. Plain Language Summary: Hydrological model calibration is a critical model building process that infers key model parameter values from observed system response data. Conventionally, this process requires the historical period to be split into a calibration period for tuning parameters and a validation period for testing model robustness (i.e., the split‐sample). Unfortunately, there is a lack of empirical evidence supporting how exactly to define the split‐sample. We designed an exhaustive and novel experiment comparing the range of possible split‐sampling schemes, including calibrating to older/recent years, calibrating to a short/long period, and calibrating to the full period of available system response data. Each scheme was evaluated based on performance, assessed in three different ways, in numerous post‐validation model testing periods for each of the 926 calibration case studies (two different hydrological models applied in 463 catchments). Results show that using older data for model calibration and then using newer data for validation, which is the typical practice in the literature, is an inferior choice and should be avoided. The results also show that calibrating to the full historical data and skipping model validation entirely is the most robust choice. Therefore, the split‐sample approach applied in this community for decades should be revised. Key Points: A unique split‐sample experiment is performed across 463 catchments to provide guidance on split sample decision‐making in model calibrationCalibrating models to the full available data period and skipping model validation entirely is the most robust choiceCalibrating models to older data and then validating models on newer data, a very common approach in literature, is an inferior choice [ABSTRACT FROM AUTHOR]

Many studies investigate river floods by analyzing annual maximum series that record the largest flow of each year, including many within‐bank events inconsequential for human communities. Fewer focus on larger floods, especially at the continental scale. Using 473 streamgages across the conterminous United States with near‐natural flow from 1966 to 2015, we characterized the seasonality, occurrence, and climatic associations of the 10 largest and 2 largest floods at each site. These are often overbank events that have ecosystem functions and pose risks to humans. We grouped sites into 14 clusters corresponding to climatic and physiographic regions and characterized their flood seasonality at a monthly resolution using a probabilistic method. We then evaluated annual occurrence regionally and nationally, and seasonal occurrence regionally, by complementing a traditional approach to trend analyses with a novel method based on expected occurrence. Relationships between flood occurrence and climate indices were also investigated. Large floods have strong seasonality in some regions, but in areas with numerous flood‐generating mechanisms, seasonality is more complex. There is little evidence nationally that large riverine floods are more or less frequent than expected in recent years and only two regions show significant trends in annual counts; few show seasonal trends. We found some regional relationships between flood counts and climate indices, annually and seasonally; nationally the Pacific North American pattern is related to annual counts of the 2 largest floods. Large‐flood occurrence was generally stable across the United States in the last five decades; this may or may not continue with projected warming. Plain Language Summary: Flood studies usually focus on records of the largest streamflow of each year because they are widely available and provide relatively large samples. However, the largest flow of each year is not necessarily big enough to inundate the floodplain and pose risks to human communities. Our study investigated only the largest observed floods in the record, ones that occur roughly every five to fifty years. We used a set of 473 gages that measure near‐natural river flow across the United States. We studied many aspects of these consequential events at regional and national scales for the period 1966–2015, including seasonality, changes over time, and associations with known climate patterns. We found that the seasonality of large floods is strong in some regions, but weak or complex in others where floods have a variety of causes. There is little evidence that large floods have become more or less frequent with time, but we found some regional and national associations between large‐flood occurrence and known climate patterns such as the Pacific North American pattern. Large‐flood occurrence has generally been stable in the last five decades, but continued re‐evaluation of the historical record is important as new data are collected. Key Points: First broad characterization of large‐flood occurrence in the conterminous United States in the modern hydroclimate, including seasonalityBoth traditional and novel methods to assess temporal distribution of large floodsFew significant temporal trends, but significant relations to climate indices for some regions and seasons [ABSTRACT FROM AUTHOR]

Water availability in the dry western United States (US) under climate change and increasing water use demand has become a serious concern. Previous studies have projected future runoff changes across the western US but ignored the impacts of ecosystem response to elevated CO2 concentration. Here, we aim to understand the impacts of elevated CO2 on future runoff changes through ecosystem responses to both rising CO2 and associated warming using the Noah‐MP model with representations of vegetation dynamics and plant hydraulics. We first validated Noah‐MP against observed runoff, leaf area index (LAI), and terrestrial water storage anomaly from 1980 to 2015. We then projected future runoff with Noah‐MP under downscaled climates from three climate models under Representative Concentration Pathway 8.5. The projected runoff declines variably from the Pacific Northwest by −11% to the Lower Colorado River basin by −92% from 2016 to 2099. To discern the exact causes, we conducted an attribution analysis of the modeled evapotranspiration from two additional sensitivity experiments: one with constant CO2 and another one with static monthly LAI climatology. Results show that surface "greening" (due to the CO2 fertilization effect) and the stomatal closure effect are the second largest contributors to future runoff change, following the warming effect. These two counteracting CO2 effects are roughly compensatory, leaving the warming effect to remain the dominant contributor to the projected runoff declines at large river basin scales. This study suggests that both surface "greening" and stomatal closure effects are important factors and should be considered together in water resource projections. Plain Language Summary: Water shortage in the western United States (US) is becoming increasingly serious due to increasing socioeconomic demands and climate change. Although previous studies have projected various degrees of runoff changes, they neglect the impact of rising CO2 on runoff projections. To explore the possible role that CO2 may play in the hydrologic cycle, we conducted three experiments with the newly improved Noah‐MP land model including vegetation dynamics and plant hydraulics. Consistent with previous studies, the western US tends to be drier toward the end of the 21st Century. CO2‐induced leaf area index increases (surface "greening") contribute considerably to the projected widespread transpiration increases and runoff reductions; however, these changes are nearly compensated by the stomatal closure effect of CO2 on transpiration, leaving the warming effect to remain the major cause to these transpiration and runoff changes. Therefore, the dual roles of CO2 in the hydrologic cycle through interactions with vegetation processes need to be considered in water resource projections. Key Points: Annual runoff of the major western US rivers is projected to decline variably from −11% to −92% by 2099 under Representative Concentration Pathway 8.5The counteracting CO2 fertilization and stomatal closure effects on runoff are roughly compensatoryDue to the two offsetting CO2 effects, warming remains the dominant driver for the projected runoff decline at river basin scales [ABSTRACT FROM AUTHOR]

Long short‐term memory (LSTM) networks represent one of the most prevalent deep learning (DL) architectures in current hydrological modeling, but they remain black boxes from which process understanding can hardly be obtained. This study aims to demonstrate the potential of interpretive DL in gaining scientific insights using flood prediction across the contiguous United States (CONUS) as a case study. Two interpretation methods were adopted to decipher the machine‐captured patterns and inner workings of LSTM networks. The DL interpretation by the expected gradients method revealed three distinct input‐output relationships learned by LSTM‐based runoff models in 160 individual catchments. These relationships correspond to three flood‐inducing mechanisms—snowmelt, recent rainfall, and historical rainfall—that account for 10.1%, 60.9%, and 29.0% of the 20,908 flow peaks identified from the data set, respectively. Single flooding mechanisms dominate 70.7% of the investigated catchments (11.9% snowmelt‐dominated, 34.4% recent rainfall‐dominated, and 24.4% historical rainfall‐dominated mechanisms), and the remaining 29.3% have mixed mechanisms. The spatial variability in the dominant mechanisms reflects the catchments' geographic and climatic conditions. Moreover, the additive decomposition method unveils how the LSTM network behaves differently in retaining and discarding information when emulating different types of floods. Information from inputs within previous time steps can be partially stored in the memory of LSTM networks to predict snowmelt‐induced and historical rainfall‐induced floods, while for recent rainfall‐induced floods, only recent information is retained. Overall, this study provides a new perspective for understanding hydrological processes and extremes and demonstrates the prospect of artificial intelligence‐assisted scientific discovery in the future. Key Points: The expected gradients method was used to retrieve flooding mechanisms learned by long short‐term memory networks in 160 catchments in the United StatesSnowmelt, recent rainfall, and historical rainfall induce 10.1%, 60.9%, and 29.0% of the 20,908 identified flow peaks, respectivelyThe additive decomposition method unveiled models' distinct behaviors in retaining and discarding information for different flood types [ABSTRACT FROM AUTHOR]

Specific yield (Sy) is one of the most important aquifer parameters in groundwater models with large storage changes. However, it has received limited attention from the modeling community and there are few, if any, approaches that can be widely applied for determination of Sy at the scale required for regional aquifer studies. Despite the limitations of parameter estimation methods and of direct measurements of aquifer storage properties in the field, the large uncertainty in model Sy values is typically ignored and Sy is incorrectly assumed to be a known quantity. These practices can introduce errors into model predictions of aquifer budget components, such as recharge and storage depletion. In turn, aquifer management and planning can be greatly impacted, as estimates of aquifer responses to pumping changes are highly dependent on those simulated budget components. In this work, we use a groundwater model of a portion of the High Plains aquifer in the central United States to illustrate the impacts of inappropriately large Sy values on model predictions and the related water management ramifications. Building on our recent work, we propose a data‐driven approach for determining specific yield for regional groundwater models. Results demonstrate that this new approach can improve model reliability and lead to more robust predictions of aquifer responses to management scenarios. Key Points: Specific yield values are uncertain in many models of unconsolidated aquifersModels with inappropriate specific yield values can produce unreliable aquifer budgets and responses under future management scenariosA new approach is proposed for estimating the spatial distribution of specific yield for regional groundwater models [ABSTRACT FROM AUTHOR]

Regional flood frequency analysis (RFFA) is widely used to quantify flood risk at ungauged and sparsely gauged locations. There are minimal attempts to use partial duration series (PDS) for RFFA, though the use of PDS instead of widely used annual maximum series (AMS) can offer some advantages. This article contributes two novel random/regression forests (RFs)‐based methodologies, namely generalized pareto distribution (GPD)‐based distributional RFs (DRFs) and multivariate RFs (MVRFs), for RFFA with PDS. The RFs facilitate modeling interactions between predictors and their complex relationships with the predictands without explicitly specifying them. The DRFs and MVRFs comprise an ensemble of corresponding regression trees, each constructed by recursive binary partitioning of the feature space into meaningful segments. The proposed DRFs account for the sampling uncertainty of PDS in the partitioning and parameter estimation. In DRFs (MVRFs), quantile estimates for an ungauged site are obtained using maximum likelihood estimates (expected values) of GPD parameters corresponding to the segments to which the site belongs. The potential of DRFs and MVRFs relative to two recently proposed techniques (univariate RFs‐based quantile regression, generalized additive model based on GPD) is demonstrated through Monte‐Carlo simulation experiments and a study on 1,031 watersheds in the United States. The key features influencing scale and shape parameters of GPD fitted to PDS of the watersheds are identified as drainage area and 24‐hr rainfall intensity corresponding to 2‐year return period, respectively. Those identified for shape parameter differ from key features known based on analysis with AMS and generalized extreme value distribution. Key Points: Two new methods based on random forests (RFs) are proposed for regional flood frequency analysis with partial duration series (PDS)Proposed distributional RFs (DRFs) can account for sampling uncertainty in PDS and capture complex predictor‐predictand relationshipsDRFs appear effective in Monte‐Carlo simulations and study on the United States and yielded key features influencing PDS of floods [ABSTRACT FROM AUTHOR]

This paper reports the development of a ∼30 m resolution two-dimensional hydrodynamic model of the conterminous U.S. using only publicly available data. The model employs a highly efficient numerical solution of the local inertial form of the shallow water equations which simulates fluvial flooding in catchments down to 50 km2 and pluvial flooding in all catchments. Importantly, we use the U.S. Geological Survey (USGS) National Elevation Dataset to determine topography; the U.S. Army Corps of Engineers National Levee Dataset to explicitly represent known flood defenses; and global regionalized flood frequency analysis to characterize return period flows and rainfalls. We validate these simulations against the complete catalogue of Federal Emergency Management Agency (FEMA) Special Flood Hazard Area (SFHA) maps and detailed local hydraulic models developed by the USGS. Where the FEMA SFHAs are based on high-quality local models, the continental-scale model attains a hit rate of 86%. This correspondence improves in temperate areas and for basins above 400 km2. Against the higher quality USGS data, the average hit rate reaches 92% for the 1 in 100 year flood, and 90% for all flood return periods. Given typical hydraulic modeling uncertainties in the FEMA maps and USGS model outputs (e.g., errors in estimating return period flows), it is probable that the continental-scale model can replicate both to within error. The results show that continental-scale models may now offer sufficient rigor to inform some decision-making needs with dramatically lower cost and greater coverage than approaches based on a patchwork of local studies. [ABSTRACT FROM AUTHOR]

In this paper, we describe a new method of modeling coastal inundation arising from storm surge by coupling a widely used storm surge model (ADCIRC) and an urban flood inundation model (FloodMap). This is the first time the coupling of such models is implemented and tested using real events. The method offers a flexible and efficient procedure for applying detailed ADCIRC storm surge modeling results along the coastal boundary (with typical resolution of ∼100 m) to FloodMap for fine resolution inundation modeling (<5 m). The coastal inundation during Hurricane Sandy was simulated at both the city (New York City) and subregional (lower Manhattan) scales with various resolutions. Results obtained from the ADCIRC and coupled ADCIRC-FloodMap simulations were compared with the recorded (high water marks) and derived (inundation extent based on the planar method) data from FEMA. At the city scale, coupled ADCIRC-FloodMap modeling demonstrates improved prediction over ADCIRC modeling alone for both the extent and depth of inundation. The advantage of the coupled model is further illustrated in the subregional modeling, using a mesh resolution of 3 m which is substantially finer than the inland mesh resolution used by ADCIRC (>70 m). In further testing, we explored the effects of mesh resolution and roughness specification. Results agree with previous studies that fine resolution is essential for capturing intricate flow paths and connectivity in urban topography. While the specification of roughness is more challenging for urban environments, it may be empirically optimized. The successful coupling of ADCIRC and FloodMap models for fine resolution coastal inundation modeling unlocks the potential for undertaking large numbers of probabilistically based synthetic surge events for street-level risk analysis. [ABSTRACT FROM AUTHOR]

Virtual water flows are used to map the indirect water consumption connections implied by the supply chain of a city, region, or country. This information can be used to manage supply chains to achieve environmental policy objectives and mitigate environmental risks to critical supply chains. A limitation of prior work is that these flows are typically analyzed using monolayer networks, which ignores crucial intersectoral or interlayer couplings. Here, we use a multilayer network to account for such couplings when analyzing blue virtual water flows in the United States. Our multilayer network consists of 115 different regions (nodes), covering the entire conterminous United States; 41 coupled economic sectors (layers); and ∼2 × 107 possible links. To analyze the multilayer network, we focus on three fundamental network properties: topological connectivity, mesoscale structure, and node centrality. The network has a high connectivity, with each node being on average connected to roughly 2/3 of the network's nodes. Interlayer flows are a major driver of connectivity, representing ∼54% of all the network's connections. Five different groups of tightly connected nodes (communities) characterize the network. Each community represents a preferred spatial mode of long‐range virtual water interaction within the United States. We find that large (populous) cities have a stronger influence than small ones on network functioning because they attract and recirculate more virtual water through their supply chains. Our results also highlight differences between the multilayer and monolayer virtual water network, which overall show that the former provides a more realistic representation of virtual water flows. Plain Language Summary: Water is a critical input in the production of any product. The total water used in the production of a consumer product, including the water used in the production of the product's raw materials and component parts, is known as the embedded or virtual water. By tracing the water used in the production of a product along all its supply‐chain stages, we map the flow of virtual water between regions for the entire product economy of the United States. This mapping results in a multilayer network of virtual water flows. By analyzing this network, we identify five dominant spatial patterns of virtual water interactions in the United States. Each pattern represents a group of distant regions that strongly depend on each other for the provisioning of virtual water. One pattern shows that a substantial amount of water withdrawn on the Western United States is used to meet cities' demand for products on the Eastern seaboard. The identification of these patterns is a critical step in the design of strategies for making supply chains more resilient to water stress and reducing the impact of economic activity on water‐scarce regions. Key Points: The United States has five distinctive interregional patterns or spatial modes of long‐range virtual water interactionsLarge cities have a greater influence than small ones on network functioning because they act as bottlenecks in the flow of virtual waterMultilayer network provides a more realistic representation of virtual water flows than single‐layer network [ABSTRACT FROM AUTHOR]

Understanding the dominant drivers of hydrological change is essential for water resources management. Watersheds in the United States are experiencing different types of changes (e.g., wet gets wetter and dry gets drier); however, few studies have analyzed what drivers are responsible for these changes, and how the dominant drivers vary over time and as a function of the climate/water regime and land cover. This study uses a time‐varying Budyko framework to quantify the relative importance of precipitation, potential evapotranspiration, and other factors (e.g., climate seasonality, agricultural drainage, and urbanization) in 889 watersheds in the contiguous United States from 1950 to 2009. Results show that watersheds that are getting wetter are primarily due to increases in precipitation. However, watersheds in dry climates that are getting drier are primarily due to other factors, while watersheds in wet climates that are getting drier are primarily due to precipitation. The drivers causing statistically significant streamflow trends vary depending on dominant land‐use types. Temporally, the increasing effects of other factors are more pronounced after the 1980s in the Midwest. The dominant drivers of streamflow in the United States are time‐varying instead of constant. This is consistent with non‐stationary patterns of streamflow. The time‐varying drivers provide information on the processes that are increasingly important and require the most attention in water resources management. Key Points: Dominant drivers of streamflow in the United States are shown to be time‐varying between 1950 and 2009Precipitation is the dominant driver in wet gets drier watersheds, while other factors are the dominant driver in dry gets drier watershedsPrecipitation is the dominant driver in reference watersheds, and other factors are the dominant driver in agricultural and urban watersheds [ABSTRACT FROM AUTHOR]

Lakes are often defined by seasonal cycles. The seasonal timing, or phenology, of many lake processes are changing in response to human activities. However, long‐term records exist for few lakes, and extrapolating patterns observed in these lakes to entire landscapes is exceedingly difficult using the limited number of available in situ observations. Limited landscape‐level observations mean we do not know how common shifts in lake phenology are at macroscales. Here, we use a new remote sensing data set, LimnoSat‐US, to analyze U.S. summer lake color phenology between 1984 and 2020 across more than 26,000 lakes. Our results show that summer lake color seasonality can be generalized into five distinct phenology groups that follow well‐known patterns of phytoplankton succession. The frequency with which lakes transition from one phenology group to another is tied to lake and landscape level characteristics. Lakes with high inflows and low variation in their seasonal surface area are generally more stable, while lakes in areas with high interannual variations in climate and catchment population density show less stability. Our results reveal previously unexamined spatiotemporal patterns in lake seasonality and demonstrate the utility of LimnoSat‐US, which, with over 22 million remote sensing observations of lakes, creates novel opportunities to examine changing lake ecosystems at a national scale. Plain Language Summary: Lakes have seasonal cycles that result in yearly peaks in algal growth. The size and timing of these peak periods depends on the amount of nutrients available and the timing of key events such as freezing and thawing. Bluer lakes with little algae typically have one peak in the spring, while greener, high algae lakes can have multiple peaks or longer duration peaks that span the summer months. The timing and duration of these peaks manifest in changes to overall lake color. Here, we look at how these seasonal cycles changed in over 26,000 lakes across the United States between 1984 and 2020. We show that seasonal cycles are changing with distinct regional patterns. Specifically, lakes are generally moving toward stable blue states in the Pacific Northwest, while high latitude lakes in the Northeast are increasingly showing seasonal cycles associated with high algae waterbodies. Lakes at high elevations and in catchments with large year‐to‐year fluctuations in temperature and population density are most prone to changes in seasonal cycles over time. Key Points: Summer lake color phenology can be generalized into five distinct seasonal patterns of greening and blueing eventsSince the mid‐1990s, the number of lakes with color patterns corresponding to eutrophic waterbodies has been increasingWe observe these patterns using a new U.S. lake remote sensing data set that contains over 22 million lake observations [ABSTRACT FROM AUTHOR]

As shallow aquifers become depleted or contaminated worldwide, use of deep aquifers will likely increase to meet growing water demands despite expensive drilling costs and well maintenance, and limited recharge to many of these deep systems. We discuss depletion of the Cambrian‐Ordovician sandstone aquifer, a deep bedrock aquifer that has been a major source of water in the Midwestern US for over 150 years. Using a participatory groundwater modeling approach with active stakeholder engagement, we developed a conceptual framework to assess risk to different sandstone units of the aquifer based on simulated heads and production well behavior (pumping levels and specific capacities). Over 320 m of static head decline has occurred since predevelopment in the deepest sandstone unit. Shallower units are actively being dewatered. Drawdown and head separation between sandstone units is even greater under pumping conditions and is most acute near a regional fault zone. Current depletion is prompting massive infrastructure changes to mitigate anticipated water supply shortages. Our results show that even with a substantial reduction in water use by 35% in 2030, many areas of the aquifer remain at High or Severe risk out to 2070. Head decline, as opposed to storage loss, is a critical metric for understanding depletion of deep bedrock aquifers. The participatory modeling process made clear that pumping levels at production wells should not be ignored in evaluations of aquifer sustainability and risk. Our conceptual risk framework could be applied to other deep bedrock aquifers undergoing depletion. Key Points: Over 320 m of head decline has occurred in the deep sandstone aquifer system underlying the United States' third largest metropolitan areaParticipatory modeling allows assessment of risk to groundwater supplies under pumping conditionsAssessment of deep aquifer sustainability requires understanding of multi‐aquifer wells, geologic barriers, and production well behavior [ABSTRACT FROM AUTHOR]

Salinity in the Colorado River Basin causes an estimated $300 to $400 million per year in economic damages in the United States. To inform and improve salinity‐control efforts, this study quantifies long‐term trends in salinity (dissolved solids) across the Upper Colorado River Basin (UCRB), including time periods prior to the construction of large dams and preceding the implementation of salinity‐control projects. Weighted Regressions on Time, Discharge, and Season was used with data sets of dissolved‐solids and specific‐conductance measurements, collected as early as 1929, to evaluate long‐term trends in dissolved‐solids loads and concentrations in streams from 1929 to 2019 (n = 14). Results indicate that large, widespread, and sustained downward trends in dissolved‐solids concentrations and loads occurred over the last 50–90 years. For 12 of the 14 stream sites with significant downward change, median declines of −38% (range of −14% to −57%) and −40% (range of −9 to −65%) were observed for flow‐normalized concentration and load, respectively. Steepest rates of decline occurred from 1980 to 2000, coincident with the initiation of salinity‐control efforts in the 1980s. However, there was a consistent slowing or reversing of downward trends after 2000 even though salinity‐control efforts continued. Significant decreases in salinity occurred as early as the 1940s at some streams, indicating that, in addition to salinity‐control projects, changes in land cover, land use, and/or climate substantially affect salinity transport in the UCRB. Observed dissolved‐solids trends are likely the result of changes to watershed‐related processes, not due to changes in the streamflow regime. Plain Language Summary: Salinity, or dissolved salt, in the Colorado River Basin causes an estimated $300 to $400 million per year in economic damages in the United States. The Colorado River Basin Salinity Control Program implements and manages projects to reduce salinity, investing millions of dollars per year in improved irrigation systems, vegetation recovery, and other mitigation strategies. To inform and improve mitigation efforts, there is a need to better understand changes in salinity that occurred prior to the implementation of salinity‐control projects in the 1980s. This study uses decades of water‐quality measurements, collected as early as 1929, to explore salinity trends in the Upper Colorado River Basin from 1929 to 2019. Findings indicate that large, widespread, and sustained downward trends in salinity occurred over the last 50–90 years. Further, the timing and amount of salinity reductions suggest changes in land cover, land use, and/or climate, in addition to salinity control, substantially affect how salinity is transported to streams in the basin. Identifying the causes of dropping salinity levels will be important for water managers in the basin so they can anticipate future changes in salinity, develop more efficient salinity‐control practices, and capitalize on natural processes that reduce salinity. Key Points: Substantial and widespread declines in dissolved‐solids loads and concentrations were observed across the Upper Colorado River Basin during the 20th centuryMajority of the downward trends are estimated to be the result of watershed‐related processes, not changes in the streamflow regimeDecreases in the 1940s indicate watershed processes other than salinity control substantially affected dissolved‐solids transport [ABSTRACT FROM AUTHOR]

Hydrometeorological flood generating processes (excess rain, short rain, long rain, snowmelt, and rain‐on‐snow) underpin our understanding of flood behavior. Knowledge about flood generating processes improves hydrological models, flood frequency analysis, estimation of climate change impact on floods, etc. Yet, not much is known about how climate and catchment attributes influence the spatial distribution of flood generating processes. This study aims to offer a comprehensive and structured approach to close this knowledge gap. We employ a large sample approach (671 catchments across the contiguous United States) and evaluate how catchment attributes and climate attributes influence the distribution of flood processes. We use two complementary approaches: A statistics‐based approach which compares attribute frequency distributions of different flood processes; and a random forest model in combination with an interpretable machine learning approach (accumulated local effects [ALE]). The ALE method has not been used often in hydrology, and it overcomes a significant obstacle in many statistical methods, the confounding effect of correlated catchment attributes. As expected, we find climate attributes (fraction of snow, aridity, precipitation seasonality, and mean precipitation) to be most influential on flood process distribution. However, the influence of catchment attributes varies both with flood generating process and climate type. We also find flood processes can be predicted for ungauged catchments with relatively high accuracy (R2 between 0.45 and 0.9). The implication of these findings is flood processes should be considered for future climate change impact studies, as the effect of changes in climate on flood characteristics varies between flood processes. Key Points: Flood generating processes are mostly influenced by climate attributesWhich attributes are influential varies between different flood processes and climate regionsMix of flood generating processes can be predicted for ungauged catchments with high accuracy [ABSTRACT FROM AUTHOR]

The paper 'Climate and agricultural land use change impacts on streamflow in the upper midwestern United States' by Satish C. Gupta, Andrew C. Kessler, Melinda K. Brown, and Francis Zvomuya (hereafter referred to as Gupta et al.) purports to evaluate 'the relative importance of changes in precipitation and LULC (land use, land cover) on streamflow in 29 Hydrologic Unit Code 008 watersheds in the Upper Midwestern United States.' However, as we report here, the approach used by Gupta et al. is wholly inadequate for making such an evaluation. Gupta et al. use strong language to criticize other studies and imply a level of certainty that goes well beyond, and in some cases is entirely unsupported by, the results they have presented. We take this opportunity to point out several critical flaws in their study. [ABSTRACT FROM AUTHOR]

Future changes in the occurrence of flood events can be estimated using multi‐model ensembles to inform adaption and mitigation strategies. In the near future, these estimates could be used to guide the updating of exceedance probabilities for flood control design and water resources management. However, the estimate of return levels from ensemble experiments represents a challenge: model runs are affected by biases and uncertainties and by inconsistencies in simulated peak flows when compared with observed data. Moreover, extreme value distributions are generally fit to ensemble members individually and then averaged to obtain the ensemble fit with loss of information. To overcome these limitations, we propose a Bayesian hierarchical model for assessing changes in future peak flows, and the uncertainty coming from global climate, global impact models and their interaction. The model we propose allows use of all members of the ensemble at once for estimating changes in the parameters of an extreme value distribution from historical to future peak flows. The approach is applied to a set of grid‐cells in the eastern United States to the full and to a constrained version of the ensemble. We find that, while the dominant source of uncertainty in the changes varies across the domain, there is a consensus on a decrease in flood magnitudes toward the south. We conclude that projecting future flood magnitude under climate change remains elusive due to large uncertainty mostly coming from global models and from the intrinsic uncertain nature of extreme values. Key Points: A Bayesian hierarchical model is developed to assess the changes in future flood magnitude and quantify uncertainty in a single stepFuture flood magnitude at selected sites over the eastern United States decreases in the south of the domain with varying uncertaintyA constrained ensemble based on how well model runs replicate timing of observed peak flows yields similar results to the full ensemble [ABSTRACT FROM AUTHOR]

Residence time distribution (RTD) is a critically important characteristic of groundwater flow systems; however, it cannot be measured directly. RTD can be inferred from tracer data with analytical models (few parameters) or with numerical models (many parameters). The second approach permits more variation in system properties but is used less frequently than the first because large‐scale numerical models can be resource intensive. Using a novel automated approach, a set of 115 inexpensive general simulation models (GSMs) was used to create RTD metrics (fraction of young groundwater, defined as <65 years old; mean travel time of young fraction; median travel time of old fraction; and mean path length). GSMs captured the general trends in measured tritium concentrations in 431 wells. Boosted Regression Tree metamodels were trained to predict these RTD metrics using available wall‐to‐wall hydrogeographic digital sets as explanatory features. The metamodels produced a three‐dimensional distribution of predictions throughout the glacial system that generally matched with the numerical model RTD metrics. In addition to the expected importance of aquifer thickness and recharge rate in predicting RTD metrics, two new data sets, Multi‐Order Hydrologic Position (MOHP) and hydrogeologic terrane were important predictors. These variables by themselves produced metamodels with Nash‐Sutcliffe efficiency close to the full metamodel. Metamodel predictions showed that the volume of young groundwater stored in the glaciated United States is about 6,000 km3, or about 0.5% of globally stored young groundwater. Plain Language Summary: Groundwater is an important source of water for people and ecosystems. The quality of groundwater depends partly on how long it has resided underground, but this is difficult to know. There are several methods to determine groundwater residence time, but they mostly do not apply over large regional aquifers. In this study, we used a novel method coupling statistical learning and numerical modeling to calculate groundwater residence time throughout the glaciated areas of the United States. The results reveal that most of the groundwater in the area is less than 65 years old and usually travels distances ranging from less than 1 km to about 30 km before discharging to streams, rivers, and lakes. Key Points: A novel method coupling statistical learning and numerical models is presentedThree‐dimensional continuous digital data sets of groundwater age metrics are presentedTwo novel data sets, Multi‐Order Hydrologic Position and hydrogeologic terrane, contain information about groundwater residence time [ABSTRACT FROM AUTHOR]