Olivia Miller

Excess riverine nutrients can limit water availability for humans and ecosystems. Nutrient monitoring on individual stream reaches provides important information about current and past conditions, yet most river reaches are unmonitored. Additionally, monitored data integrate effects of upstream sources, sinks, and pathways by which nutrients get into streams, making explicit planning for effective nutrient reduction strategies challenging. We developed and applied dynamic SPAtially Referenced Regression On Watershed attributes (SPARROW) models to quantify total nitrogen (TN) and total phosphorus (TP) delivery and source‐specific loading in all stream reaches of the conterminous United States (CONUS) at a seasonal timestep from 2000 through 2020. Between one third and one half of non‐point source nutrients lagged more than one season in their delivery to CONUS streams over the study period. The largest sources of nutrients that were applied and delivered to streams within a season were agricultural fertilizers, treated wastewater, atmospheric deposition, and manure. Increases in loads were widespread across the CONUS and predominantly associated with increased loading from lagged delivery of non‐point source inputs and current season fertilizer application. Model results also identified extensive areas where predicted seasonal concentrations frequently exceeded the National Rivers and Streams Assessment “poor” criteria. Increased loading occurred in areas with high yields and delivery to downstream receiving waterbodies, particularly across the Midwest, raising the risk of negative impacts to downstream water availability. The modeling approach presented here provides a continental‐scale framework for identifying when, where, and why nutrients are generated and how that may ultimately impact receiving waterbodies.

The Upper Colorado River Basin (UCRB) faces substantial water availability limitations. Although most streamflow originates as snowmelt, the partitioning of snowmelt between surface runoff and groundwater recharge and subsequent groundwater discharge to streams is highly uncertain. On average, over half of the streamflow in the UCRB is estimated to originate from groundwater discharge to streams, highlighting the importance of baseflow in sustaining surface water. However, the historical patterns of baseflow and streamflow, along with their variability over space and time and their specific sources, remain unknown at the basin scale. This study addresses those gaps by characterizing the sources and transport pathways of both baseflow and streamflow in the UCRB at a seasonal timestep from 1986 to 2020, including the lagged delivery of subsurface water to streams beyond the current season, using coupled models of baseflow and streamflow. Between 1986 and 2020, on average 63% of UCRB streamflow originated from baseflow. About half of this baseflow took longer than one season to reach streams, and outside the snowmelt season, baseflow was the dominant source of streamflow. Snowmelt was a key source of both baseflow and streamflow. Current season snowmelt contributed 33% of streamflow via runoff, and 22% of the 29% of streamflow that originated as current season baseflow via subsurface flow to streams. Over the study period, baseflow index (BFI) declined in headwaters and increased at mid-elevations. Springtime increases in BFI demonstrate the increasingly important role baseflow plays in water supply. Identifying the sources, locations, and timing of water that contributed to the UCRB outlet can inform management of water resources in the basin.

Freshwater salinization affects the availability of water for human use and ecosystem needs worldwide. It has been estimated that total dissolved solids (TDS) in the Colorado River Basin cause $350 million/year in damages and substantial resources are devoted to reducing TDS loading to streams. This study describes the development and application of coupled watershed models that enable TDS source tracking through the subsurface and across the landscape at a seasonal timestep for 35 years in the Upper Colorado River Basin. Results indicate that, on average, 75% of TDS loading to streams originates as baseflow, and 50% of loading is lagged in delivery by longer than one season. Snowmelt was identified as a dominant process controlling the transport of lagged TDS to streams. This approach informs when and where TDS mitigation efforts may be effective in a watershed that serves as a critical water supply for the southwestern United States.

Consistent, large-scale estimates of water availability are needed to identify and avoid potential conflicts among human and ecosystem uses of water. We present an assessment of water limitation, defined as the monthly balance (difference) between water supply (ws) and human consumptive water use (wc), for the conterminous United States (CONUS) during water years 2010–2020. We estimate that 26.7 million Americans, 8% of CONUS population, live in areas with chronic high or severe water limitation. Although ws greatly exceeds wc at the CONUS scale, water is limited locally or regionally due to spatial and temporal patterns in climate and wc. Our water limitation metric, the monthly supply and use index (SUI), peaked in 2012 during a widespread drought when 38% of the CONUS land area experienced elevated water stress. The central and Southwestern U.S. experienced the highest SUI due to the combination of low ws and high wc, especially for irrigation. Spatial overlays of SUI and fish habitat ranges, including those of conservation concern, revealed that several species had notable proportions of their habitat exposed to high or severe water limitation during spawning season over the modeled time period, especially the Arkansas River shiner. ws was calculated from two CONUS, physically-based, hydrologic models while wc was calculated from three CONUS models of water use for crop irrigation, thermoelectric power generation, and public supply. The ws and wc values were routed through a stream network and compared to calculate water limitation and SUI for human populations and fish species at the scale of 12-digit hydrologic unit codes. Evaluation of water availability at higher spatial and temporal resolution promotes more comprehensive analyses of the drivers of water availability and can be combined with complementary studies of water quality and water limiting thresholds to better understand the limitations on water availability.

Water availability is fundamentally important to human well-being, economic vitality, and ecosystem health. The United States Geological Survey (USGS) recently completed a comprehensive assessment of water availability in the United States which included water supply, human water consumption, water quality, and ecological flows. The assessment relied upon national-scale models of natural and human processes including hydrologic conditions and human water consumption. Surface water total nitrogen and phosphorus concentrations were assessed along with groundwater nitrate and arsenic concentrations and ecologically relevant streamflow alteration. From 2010–2020, around 27 million people lived in areas where water consumption was > 80 % of water supply and therefore likely to experience regular water limitations. Water limitation was most severe in areas with high withdrawals for crop irrigation. The areal extent of potential water limitation was greatest in 2012–2013 during an unusually hot and dry period and coincided with elevated withdrawals for crop irrigation. Total nitrogen and phosphorus concentrations were elevated in surface water in many parts of the conterminous U.S. (CONUS), particularly agricultural areas. Regional comparisons showed that areas with the most severe water use imbalances also tended to have the highest concentrations of nutrients in surface waters and groundwater contaminants. The analysis highlights the multifaceted ways that excessive human water consumption can create water availability limitations.

Freshwater salinization impacts the availability of water for human use and ecosystem needs across the globe. In the southwestern United States, it has been estimated that total dissolved solids (TDS; a proxy for salinity) from the Upper Colorado River Basin cause upwards of $300 million per year in economic damages. Substantial resources are devoted to reducing TDS loading to streams and rivers in the basin, with targeted mitigation efforts that have been informed by long-term average models that have identified ubiquitous dissolution of minerals in soils and rocks and relatively small areas of irrigated lands as dominant sources of TDS. Recent work has demonstrated that between 65% and 80% of TDS loading to streams in the basin originates from baseflow, a proxy for groundwater discharge to streams, highlighting the importance of subsurface processes in TDS delivery to streams. This study describes the development and application of two coupled, temporally dynamic Spatially Referenced Regressions on Watershed attributes (SPARROW) models that estimate sources and processes influencing the transport of dissolved solids to streams in the basin at a seasonal time-step over a 35-year period. The key advance is using seasonal estimates of baseflow TDS loads from a baseflow-specific model as explicit time-varying inputs to a total in-stream TDS model, which allows for source tracking through the subsurface and across the landscape over a range of timescales. Results suggest that baseflow is a dominant source of TDS loading throughout the basin, and its relative source proportion varies spatially and across timescales as a function of climate, geology, and land use. Given the lagged delivery associated with baseflow contributions of constituents to streams, this approach of first quantifying baseflow and its drivers relative to all sources has implications for how, when, and where mitigation efforts aimed at reducing TDS loading to streams may be effective.

In the southwestern United States, the Upper Colorado River Basin (UCRB) faces substantial water availability challenges. Snowmelt dominates hydrology in the basin, with much of the streamflow originating as meltwater. Snowmelt either becomes surface runoff or groundwater recharge, but this partitioning is not well constrained. Furthermore, groundwater recharge from snowmelt can discharge back into streams becoming an important component of streamflow. On average, over half of the streamflow in the UCRB is estimated to originate from groundwater discharge to streams, highlighting the importance of baseflow in sustaining surface water. Yet we have not quantified past spatio-temporal variability of baseflow and its contributions to streamflow, nor do we understand variations in streamflow and baseflow sources under shifting hydroclimates. Here we describe the development and application of linked models of baseflow and streamflow to characterize sources and transport pathways of both baseflow and streamflow in the UCRB at a seasonal timestep from 1986-2020, including the lagged delivery of groundwater to streams over longer timescales. Results suggest that baseflow yields are greatest in headwater catchments during spring, and that a majority of baseflow is derived from water that takes over one season to move through the subsurface to streams. We also find that although streamflow and its sources vary seasonally, on average across the basin, about half of streamflow is from baseflow, which is particularly important for sustaining streams outside the snowmelt season and at lower elevations.

Recent and ongoing innovations in SPAtially Referenced Regression On Watershed attributes (SPARROW) modeling provide improved representation of nutrient dynamics and identify opportunities to optimize surface-water monitoring over large regional catchments. Efficient and effective monitoring, management, and restoration of surface-water quality requires adequate understanding of the upstream sources, fate, and transport of contaminants in contributing catchments. For nutrients, such understanding is often confounded by the multiple nitrogen and phosphorus sources and interacting hydrogeologic and geochemical conditions typical of large catchments. SPARROW models employ a hybrid empirical approach with deterministic constraints (such as flow routing and mass balance) to resolve such confounding factors affecting in-stream nutrient fluxes in large watersheds. SPARROW has been applied in many areas to a variety of water resources challenges, including predicting water quality in unsampled streams and into the future, targeting management investments, and improving the understanding of water-quality trends, contaminant fate and transport, and the effectiveness of management practices. Such models have traditionally been calibrated to mean-annual flow-normalized contaminant fluxes and therefore estimate mean fluxes over multiple years and during average hydrologic conditions. Recent advances in SPARROW, however, incorporate dynamic storage in terrestrial uplands and represent seasonal or annual fluxes over 21 years in catchments of the Illinois and Upper Colorado Rivers (respectively) in the central United States (US). Ongoing development of similar seasonal models for nitrogen and phosphorus over the conterminous US is intended to provide an improved representation and prediction of nutrient conditions affecting water availability as well as an improved understanding of observed changes in water quality over time. In a parallel effort, understanding of nutrient sources, fate, and transport quantified in calibrated SPARROW models is being used along with simulated annealing to demonstrate options for optimizing water-quality monitoring in support of regional flux estimates. Although monitoring is often conducted to address multiple objectives, such optimizations suggest opportunities for substantially improved efficiency in monitoring for the specific purpose of estimating regional fluxes to the Delaware River and to Chesapeake Bay along the mid-Atlantic US coast.

The lack of comprehensive water supply prediction capacity in most areas of the U.S. poses challenges in evaluating water availability. In order to improve water availability prediction and assessment, in 2019, the U.S. Geological Survey initiated planning efforts to intensively study five medium-sized basins throughout the U.S. over the next decade, including the Upper Colorado River Basin (UCOL). Research in the UCOL aims to provide insight into how past, present, and future snow conditions – including amount, timing, melt, and transitions from snow- to rain-dominated systems – impact water supply (quantity and quality) and the ability to meet demand. A specific emphasis is placed on how these processes affect water budget components and dissolved solids concentration and loading in the UCOL. A fully integrated groundwater-surface water hydrologic model (GSFLOW), and temporally dynamic dissolved solids models (SPARROW) that explicitly represent the groundwater contribution to dissolved solids loading to streams are being applied to meet the study objective. Comprehensive information on water diversion and groundwater pumping has been compiled and is being explicitly represented in the models to better represent human demand. Temporal and spatial patterns in predicted water quantity and quality conditions are being evaluated in the context of past and projected future changes, summarized over 30-year time periods, in snow metrics. Historical trends in multiple snow metrics, water budget components, groundwater levels, and dissolved solids provide context for evaluations of current conditions and motivation for further investigation and modeling. The tools and concepts developed in the UCOL will contribute to ongoing work in the other regional study basins as well as a forthcoming national water availability assessment.

Stream water-quality and its drivers vary across time and space, but we only monitor a small fraction of streams consistently over long periods of time. Such limited monitoring necessitates the development and application of spatially explicit and dynamic models to predict water quality at unmonitored locations. Historically, data and computational limitations have hindered temporally variable prediction efforts across large spatial scales. However, hybrid statistical and process models, such as Spatially Referenced Regression on Watershed attributes (SPARROW), can provide spatially explicit, accurate predictions of water quality constituents with substantially lower computational cost than process-only models while retaining process-level information that can be obscured within machine learning models. An emerging next generation of such hybrid models moves beyond temporally static predictions into dynamic predictions. Here, we present regional- and continental-scale dynamic SPARROW models developed across the United States to simulate annual salinity and seasonal nutrient loads and concentrations over decades. Dynamic SPARROW models account for temporal variability of constituent sources and processes that deliver constituents from the landscape to streams. In addition, dynamic SPARROW models quantify lagged delivery of contaminants to streams that may have accumulated in soils, groundwater, and vegetation. Results quantify that legacy sources can vary by constituent, location, and time, and provide inference into river responses and lags to management activities. For example, groundwater storage contributes between 66 and 82% of the dissolved solids load to streams in the Upper Colorado River Basin, while lagged storage contributes on average between 20% to nearly 50% of the total nutrient load to Illinois River Basin streams. Ongoing work to expand dynamic representation of loading up to the continental United States will provide further insight into the continually evolving impacts of legacy and other sources on riverine water quality. Dynamic representation of key processes across spatial scales provides new opportunities for more informed management that can improve water quality for human and ecosystem uses.