Introduction

Precipitation—the process by which water vapor in the atmosphere condenses and falls to the Earth’s surface in the form of rain, snow, sleet, or hail—is a fundamental component of the hydrologic cycle and the primary input to hydrologic models and budgets. Accurate measurements and modeling of precipitation amount and timing are critical for understanding and quantifying various hydrologic processes, including surface runoff, groundwater recharge, and ET, which have direct impact on agricultural planning, water resource management, infrastructure development, and flood forecasting. In addition to precipitation, the total amount of atmospheric moisture, which represents the vertically integrated amount of water vapor in the atmosphere, is a useful indicator of the total amount of atmospheric water vapor available for precipitation at any given location.

Precipitation Results

Results from analysis of the CONUS404 dataset showed that CONUS-wide precipitation averaged 857 millimeters per year (mm/yr) of precipitation, ranging from the driest year (729 millimeters [mm] in 2012) to the wettest year (995 mm in 2019; Foks and others, 2024a). When viewed across the entire CONUS, May was the month with the most precipitation (92 mm) and January was the month with the least precipitation (58 mm).

The national scale patterns are composites of distinct regional signals that drive water availability in unique ways across the country (figs. 4 and 5). The three hydrologic regions receiving the most precipitation are the Pacific Northwest (1,688 mm annually), the Mississippi Embayment (1,470 mm annually), and Florida (1,468 mm annually). In the Pacific Northwest, cool, wet winters and warm, dry summers define precipitation patterns, with topographical effects of the Cascade Mountains strongly influencing the spatial distribution. In the Mississippi Embayment and Florida hydrologic regions, monthly precipitation totals are relatively consistent throughout the year, with high totals in the summer attributable to convective thunderstorms. Mountainous regions in the West are among the driest areas in the United States. The Central Rockies and California–Nevada hydrologic regions receive less than 450 mm/yr of precipitation, and the Columbia–Snake hydrologic region receives 643 mm/yr. These hydrologic regions are heavily dependent on winter snowfall and snowfall accumulation to sustain water resources throughout the drier summer months. This pattern is most clear in the California–Nevada and Columbia–Snake hydrologic regions, where 42 percent and 38 percent of annual precipitation falls in the winter months, respectively. The driest hydrologic region is the Southwest Desert, which receives 306 mm/yr of precipitation. This hydrologic region is heavily dependent on the Southwest monsoon, which typically delivers 37 percent of annual precipitation during the summer months. The central part of the country (including the Souris–Red–Rainy, Northern High Plains, Central High Plains, Southern High Plains, and Texas hydrologic regions) is characterized by low overall precipitation, with spring and summer periods that are wetter than the dry winters.

The strong seasonality that defines precipitation and thus water availability in the Western aggregated hydrologic region is much less pronounced in the Northeast through Midwest and the Southeast aggregated hydrologic regions, where precipitation falls more consistently throughout the year (fig. 5). Seasonal patterns are present in the Mississippi Embayment, Tennessee–Missouri, and Northeast hydrologic regions, driven by more consistent spring and summer rains, but they are much less pronounced than those in the Western aggregated regions. The differences in seasonal patterns of precipitation drive water availability across the Nation. Many areas in the Northeast through Midwest and Southeast aggregated hydrologic regions have a more consistent water supply compared to regions in the Western aggregated hydrologic regions, which are heavily reliant on precipitation from a single season to sustain water supplies for the remainder of the year. This pattern is particularly true in snow-dominated regions (see section, “Snow Water Equivalent”).

Interannual variability in precipitation totals has a considerable effect on water availability, with more variable precipitation necessitating increased storage infrastructure, long-term planning, and diversification of water sources. Precipitation variability was assessed by calculating the coefficient of variation (CV) of the mean annual precipitation across the period of analysis, where the CV is calculated as:

A higher CV value indicates more year-to-year variability in precipitation and lower values indicate more consistent rainfall totals. The California–Nevada and Texas hydrologic regions had the greatest CV values (0.31 in both instances), indicating that the rainfall in these regions can vary considerably from year to year. The Florida (0.08) and Tennessee–Missouri (0.11) hydrologic regions had the lowest CV values, and the average across the CONUS was 0.08.

Introduction

Evapotranspiration (ET) is the flux of water vapor from the surface of the Earth to the atmosphere. ET encompasses two subprocesses: (1) evaporation, which occurs on the surface of waterbodies, vegetation, and bare ground; and (2) transpiration, in which plants and other vegetation transfer water from the soil and aquifer system through plant roots, stems, and eventually leaves to the atmosphere (Senay and others, 2013). ET is generally the second largest water budget component behind precipitation, and accurate estimates of ET are important for water budgeting, drought monitoring, irrigation timing, crop-yield modeling, and many other applications.

Evapotranspiration Results

According to the ensemble estimates from NHM–PRMS and WRF–Hydro, mean annual ET across the CONUS during water years 2010–20 was 539 mm/yr, about 63 percent of the average annual precipitation (Foks and others, 2024a, 2024e; Sampson and others, 2024). ET ranged from a minimum value of 488 mm in 2012 (67 percent of annual precipitation) to a maximum value of 591 mm in 2019 (59 percent of annual precipitation). Annual ET showed a strong seasonal signal with highest values of 77 mm in June and lowest values of 17 mm in December. Summertime high ET values are driven by a combination of increased transpiration from the growth of natural vegetation and agricultural crops and an increase in summer temperature and precipitation in some regions. Additionally, greater solar radiation during summer leads to greater evaporation.

In the eastern parts of the CONUS, ET is primarily limited by the available energy, resulting in a lower percentage of total precipitation recycled back to the atmosphere. By contrast, in the western parts of the CONUS, ET is often limited instead by available water. Although there are exceptions to this broad generalization, such as the Pacific Northwest hydrologic region, it explains a large amount of the regional patterns in ET across the CONUS (figs. 3B, 4E–H, 5). Model simulations show that the highest ET was in the Mississippi Embayment, Gulf Coast, and Florida hydrologic regions, where ET averaged 896, 899, and 935 mm/yr, respectively, or 61, 62, and 64 percent of annual precipitation, respectively. Areas with the lowest ET were the Southwest Desert, Alaska, and California–Nevada hydrologic regions, where ET averaged 248, 277, and 278 mm/yr, respectively or 81, 69, and 76 percent of total annual precipitation, respectively.

Seasonal patterns in ET across regions are driven by unique combinations of water supply, including soil moisture, precipitation, and surface water, and demand for ET from vegetation differing by type and coverage, solar radiation, and air temperature, and specific humidity. For example, in the Atlantic Coast hydrologic region, where precipitation is relatively constant throughout the year, ET follows a pattern similar to the CONUS-wide pattern with highs in summer and lows in winter (when most vegetation ceases transpiration after frost). In the Florida and Texas hydrologic regions, where solar radiation is consistently high, ET closely follows precipitation totals, accounting for a relatively constant fraction of precipitation (fig. 5). In contrast, areas with snowpack and summer lows in precipitation (such as the Columbia–Snake or the Pacific Northwest hydrologic regions) show highest ET earlier in the year because of melting snowpack followed by low summertime ET attributable to limited water availability even though evaporative demand is highest. In the Southwest Desert hydrologic region, late summer increases in ET are largely attributable to increased precipitation from the Southwest monsoon, a consistent, regionally important phenomena (Woodhouse and Udall, 2022).

Introduction

Streamflow is the flow of water in a natural channel on the land surface. Streamflow can be conceptualized as having two primary components—base flow and quickflow. Base flow is subsurface water that enters the stream channel along any reach that intersects an aquifer; it maintains streamflow between precipitation events. Quickflow is short-term drainage off the landscape following precipitation or snowmelt events. Although streamflow dynamics are often more complex, this conceptualization is useful and consistent with how fluxes are represented in the hydrologic models used in this assessment. To illustrate these definitions, in figure 6, two diagrams show stream-groundwater interactions, and an example hydrograph shows base-flow partitioning. Although streams can either gain or lose flow through connections to aquifers through bidirectional exchange (Jasechko and others, 2021), in the hydrologic models’ conceptualization of base-flow processes, base flow only adds flow to the stream.

Streamflow Results

Analysis of ensemble results from NHM–PRMS and WRF–Hydro simulations showed that annual average streamflow across the CONUS for water years 2010–20 was 249 mm/yr (Foks and others, 2024e; Sampson and others, 2024). The lowest streamflow occurred in water year 2012 (214 mm/yr) and the highest streamflow occurred in water year 2019 (303 mm/yr). Averaged across the CONUS, the month with highest average streamflow was May (30 mm) and the month with the lowest average streamflow was August (15 mm).

Hydrographs showing total streamflow for each water year aggregated at the scale of the hydrologic region indicate distinct patterns in amount and seasonality (fig. 7). Percentile bands, calculated across the 11-year period (water years 2010–20), provide context for extreme high or low flows for the period, with the highest and lowest cumulative flow years highlighted. Consistent, distinct seasonal patterns across the period of analysis are evident in the regions most dominated by snowpack including California–Nevada, the Central Rockies, and the Columbia–Snake hydrologic regions. Seasonal patterns are also evident in the Mississippi Embayment and Tennessee–Missouri hydrologic regions; however, because the total streamflow in those regions was greater than in many regions in the Western aggregated region, a similar variability in total streamflow represents a lower percentage of flow. In the Souris–Red–Rainy, Northern High Plains, and Central High Plains hydrologic regions, streamflow was consistently lowest in winter and higher in spring and summer. In the Texas, Southern High Plains, and Southwest Desert hydrologic regions, the seasonal patterns are less obvious and emphasize the importance of event-driven streamflow there.

Annual hydrographs show that water years 2011 and 2012 had the lowest streamflow in 12 of the 18 regions, located primarily in the southern parts of the Western, High Plains, and Southeast aggregated hydrologic regions, coincident with the 2011–12 drought. This drought affected more than one-half the conterminous United States and was the most extensive drought in the United States since the 1930s, resulting in an estimated $30.0 billion ($40.5 billion in 2023 dollars) in damages and 123 attributed deaths (Smith, 2023). In the Columbia–Snake, Pacific Northwest, and California–Nevada hydrologic regions, 2017 was the water year with the highest streamflow. The elevated streamflow led to flooding in California–Nevada and was largely attributed to a deep snowpack and heavy precipitation from atmospheric river systems (Henn and others, 2020).

Total streamflow, base-flow depth, and base-flow fraction are shown for the hydrologic regions on a seasonal basis in figure 8. Base-flow depth is the area-normalized amount of base flow while base-flow fraction is the fraction of total streamflow derived from base flow and serves as an indicator of groundwater influence in the regional water supply. Within the context of the results from the models used in this assessment, which only simulate a shallow groundwater system, high base-flow fraction can indicate that inputs of subsurface water buffer a stream against short-term decreases in water supply, for example during the dry season. However, in prolonged drought, the subsurface water can become depleted making streamflow vulnerable to longer-term shifts. If shallow groundwater is also withdrawn for water use, streamflow can be substantially reduced during prolonged drought. Conversely, regions with a lower base-flow fraction have precipitation that replenishes streamflow and shallow groundwater storage on a regular basis. Highest base-flow fractions in summer, autumn and winter are simulated across the Western aggregated hydrologic regions and Northern High Plains hydrologic region where total streamflow is among the lowest of all hydrologic regions. The Pacific Northwest hydrologic region is the outlier in the Western aggregated hydrologic regions, with a distinctly lower base-flow fraction than other western hydrologic regions in all seasons except summer, emphasizing its high total streamflow and snowmelt-dominated hydrology.

In the Western hydrologic regions during the spring, the base-flow fraction is generally lower than during other seasons, reflecting the snowmelt-dominated streamflow inputs during spring and summer in those hydrologic regions. In the eastern and central hydrologic regions of the CONUS, where total streamflow is generally higher than elsewhere in the CONUS, the base-flow fraction is lower and less variable season-to-season. The highest base-flow fraction in the Southeast aggregated hydrologic regions is present in the Florida and the Atlantic Coast hydrologic regions, which have abundant moisture year-round and where high vegetation density and ET rates may account for a higher base-flow fraction there.

Streamflow Capture and Depletion

Streamflow depletion occurs when groundwater pumping “captures” water that would have contributed to streamflow, reducing or even reversing the flow of water from a shallow aquifer to a stream (Zipper and others, 2022). Streamflow depletion may be a consequence of pumping near-stream groundwater for industrial or municipal supplies or for irrigation. It is difficult to quantify groundwater withdrawals that lead to streamflow depletion at regional scales; datasets on pumping wells are scarce and withdrawals often need to be inferred from other evidence. Several approaches using statistical or analytical methods and models have been developed recently (Zipper and others, 2022; Brookfield and others, 2024).

On a regional scale, streamflow depletion is most likely to occur where potential evapotranspiration (atmospheric demand) is greater than precipitation, although individual streams can have streamflow depletion anywhere groundwater pumping is sufficiently high. An investigation of base-flow trends and climatic drivers over a 30-year period (Ayers and others, 2022) established that precipitation deficits and rising temperatures were associated with declining base flow in the southern part of the United States whereas annual cycle shifts and increased base flow in the Northern United States were associated with increasing precipitation and temperatures. Ayers and others (2022) noted that, in addition to climate drivers, irrigation and groundwater pumping were major influences on decreasing base flow across arid and semiarid regions.

The models presented in this report treat streams and shallow, subsurface water as a linked system; streamflow at any given time comprises shallow, subsurface storage-derived base flow, with or without a quickflow component from recent precipitation or snowmelt. Future assessments can further determine the magnitude and spatial extent of net losses or gains of streamflow attributable to subsurface pumping from this linked system; for example, by addition of deep groundwater from outside the linked system through irrigation or public-supply return flows, or by capture of the base-flow volume that would occur in a setting without groundwater withdrawals. Assessment tools (for example, monitoring of changes in streams from gaining to losing, or from perennial to intermittent streams) would aid in the development of this capability. Likewise, better integration of surface water and groundwater components and, in particular, groundwater pumping for withdrawals, would also aid in the development of this capability.

Introduction

Soil moisture is a general term for water that occurs below the surface of the earth, but above the saturated zone in an area often referred to as the unsaturated zone or the vadose zone. Soil moisture plays a critical role in many hydrologic processes such as infiltration, groundwater recharge, runoff generation, and evapotranspiration (Vereecken and others, 2022). It serves as a key linkage between hydrologic, biological, and climatic processes by providing water for plants in the root zone and recycling water back to the atmosphere through plant transpiration and evaporation from bare soils (Legates and others, 2011). Soil moisture acts to partition incoming solar radiation between latent and sensible heat, which subsequently drives the climate system (Entekhabi and others, 1996; Seneviratne and others, 2010).

Although the quantity of water in the unsaturated zone can be large (Long and others, 2013), volumetric estimates of soil moisture are difficult to measure directly at applicable scales. Therefore, to compare the soil-moisture simulations from both models (WRF–Hydro and NHM–PRMS), we present relative soil-moisture volumes ranging from 0 (representing the permanent wilting point) to 1 (representing the field capacity) to emphasize the accessible part of soil water. These values are independently calculated for each model, and then averaged across the two models and multiplied by 100 to calculate percent saturation. Although the two models represent soil processes differently, this approach serves to unify the different conceptualizations of soil-water processes and facilitates the comparison of spatiotemporal patterns of available soil moisture to those seen in the benchmarking dataset.

WRF–Hydro simulates soil moisture to a depth of 2 meters (m) in a four-layer soil zone (0–10, 10–40, 40–100, and 100–200 centimeters [cm]) that can acquire a range of saturations to contribute to lateral flow, exfiltration, drainage to deeper layers, or contributions to streamflow based on an empirical storage-discharge formulation. For this report, to facilitate comparison between the models and benchmark dataset, the soil-moisture fluctuations in the top 10 cm of the soil zone were considered.

In NHM–PRMS, soil moisture is not simulated at a specific depth; rather, it is represented in two conceptual reservoirs: (1) the capillary reservoir, which represents the soil moisture between field capacity and the permanent wilting point; and (2) the gravity reservoir, in which water over the field capacity contributes to slow lateral interflow and drainage to groundwater.

In addition to soil saturation, we also present estimates of soil-moisture volumes, as simulated by NHM–PRMS, to facilitate volumetric comparisons to other fluxes and storage components. We use the combined volume in the capillary reservoir and the gravity reservoir to represent all moisture below the land surface and above the water table. To summarize soil-moisture volumes, we use the average monthly soil-moisture volume and normalize by the area to generate a depth of soil moisture.

Soil Moisture Results

Analysis of soil-moisture volumes from NHM–PRMS showed that across the CONUS, soil moisture was relatively higher in the mountainous regions of the Western United States and the northern parts of the Northeast and Great Lakes hydrologic regions, and relatively lower in the High Plains and Southeast aggregated hydrologic regions (Foks and others, 2024e; fig. 9A). Soil-moisture saturation was highest in winter, when ET was low, and lowest in summer, when demands from ET were high. Soil-moisture saturation showed similar spatial patterns to ET, with relatively higher values in the more mesic eastern part of the country and relatively lower values in the more arid West, where a relative lack of precipitation and high evaporative demand reduced soil-moisture storage, particularly in the upper layers of the soil (fig. 10).

Several hydrologic regions—such as Florida, Texas, Central Rockies, Southwest Desert, the Northern High Plains, Central High Plains, and Southern High Plains—showed minimal seasonal variation in soil moisture. In these hydrologic regions, ET and streamflow balanced precipitation and there was little accumulation of soil moisture throughout the year (fig. 5). In other hydrologic regions, seasonal highs in soil moisture occurred in winter and spring, when precipitation was high and was not matched by losses attributable to ET, as was the case in the Pacific Northwest, Northeast, Great Lakes and Mississippi Embayment hydrologic regions.

Soil moisture provides a critical storage component, which is closely tied to precipitation-streamflow dynamics in watersheds. Soil moisture can influence flood generation through rainfall-runoff response where saturation-excess mechanisms lead to flooding (Berghuijs and others, 2016; Crow and others, 2018), and knowledge of soil moisture can improve streamflow forecasts (Koster and others, 2010; Harpold and others, 2017). Soil moisture is driven by topography, vegetation, soil depth, and antecedent precipitation (Vereecken and others, 2014) and can be affected by soil properties such as bulk density and porosity at the local scale (Gaur and Mohanty, 2013). Seasonal, cumulative soil-moisture volumes were compared across hydrologic regions and time periods through normalizing by cumulative-precipitation totals, highlighting areas with strong imbalances between precipitation and storage (fig. 11). For example, soil-moisture stores in the Pacific Northwest hydrologic region in summer show the highest values (1.53 decimal fraction of 1). Summer precipitation is low in the Pacific Northwest hydrologic region, so the excess moisture helps to sustain base flow, which constitutes a large fraction of streamflow (fig. 8). In contrast, during summer in the Sours-Red-Rainy, Texas, Southern High Plains, and Southwest Desert hydrologic regions, soil-moisture volumes are low (all less than 0.07 decimal fraction of 1), indicating that the precipitation that falls during these times is directly lost to streamflow and (or) ET and relatively little is partitioned to storage. The lowest values are all present in summer, except in autumn in the Southwest Desert and Texas hydrologic regions (both 0.05 decimal fraction of 1), and in spring in the Texas hydrologic region (0.04 decimal fraction of 1). These results highlight the potential vulnerability of these hydrologic regions, as there is little storage to buffer supplies during periods of low rainfall.

To assess the temporal patterns in soil moisture across the period of analysis, the minimum and maximum soil saturation for each month was calculated. The percentage of each of the aggregated hydrologic regions that had extreme soil saturation (defined as the minimum or maximum soil saturation for that month across the period of analysis) was calculated and is shown in figure 12. Extremely dry conditions were simulated in >25 percent of the Southeast aggregated region in water year 2011. In water year 2012, >75 percent of the High Plains and >60 percent of the Northeast through Midwest aggregated region had extremely dry conditions because of anomalously low rainfall totals in spring and summer. These extremely dry conditions were associated with the most severe and widespread drought in the period of analysis and one of the most extreme droughts since the 1930s dust bowl, when similar conditions were present throughout the High Plains aggregated region and much of the western part of the Northeast through Midwest aggregated region (Hoerling and others, 2014; Bell and others, 2015). Large parts of the Southeast and Northeast through Midwest aggregated regions had the wettest soil-moisture conditions in water years 2019 and 2020, whereas the Western aggregated region showed extremes in soil-moisture saturation in response to the heavy snowpack and rainfall in water year 2017.

Introduction

Water storage as snow is an important part of the hydrologic cycle and snowpack dynamics have a substantial influence on water availability (Dettinger, 2005). The amount of water stored as snow, which is a function of snow depth as well as snowpack density and structure, is called snow water equivalent (SWE). SWE is comparable to other hydrologic storage terms such as soil moisture and surface-water storage. SWE is a temporary storage of water that feeds soil moisture, streamflow, and downstream reservoirs, which in turn are vital to human water uses. Snowpack accumulation is limited to high-latitude and high-elevation areas, and in most places in the CONUS it melts completely during spring and summer. The cycle of snowpack accumulation and melting is important to the maintenance of base flow during summer, especially in semi-arid environments (Godsey and others, 2014).

Recent changes toward earlier snowpack melting imperil water supplies in the Western United States. Earlier melt typically occurs more slowly than later melt, allowing more of the stored water to be lost to evaporation rather than replenishing the subsurface storage that sustains dry-season streamflows and reservoir levels (Milly and Dunne, 2016; Sexstone and others, 2020). The maximum monthly amount of SWE is one measure of snow accumulation that captures the peak extent of annual accumulation without double-counting snow that persists from one month to the next, hereinafter referred to as maximum SWE. We use maximum SWE to assess spatial patterns and temporal variability in snowpack across the CONUS and OCONUS.

Snow Water Equivalent Results

Results from our analysis showed that average maximum SWE during water years 2010–20 across the CONUS was highest in the mountainous Western aggregated hydrologic regions, such as the Columbia–Snake (169 mm/yr) and Pacific Northwest (117 mm/yr) hydrologic regions (fig. 13A; table 3; Foks and others, 2024e; Sampson and others, 2024). Two hydrologic regions in the Western aggregated hydrologic region in which snowpack is an important water source showed somewhat lower maximum SWE: the Central Rockies (69 mm/yr), and the California–Nevada (48 mm/yr) hydrologic regions. The highest maximum SWE in all these hydrologic regions occurred in water year 2017. The lowest maximum accumulation in all these regions except the Central Rockies hydrologic region occurred in water year 2015; the Central Rockies hydrologic region had a lower maximum SWE in water year 2018.

Among these Western aggregated hydrologic regions, the California–Nevada hydrologic region had the highest interannual variability, with a CV of 0.60, compared to a CV of less than or equal to 0.36 for the Columbia–Snake, Pacific Northwest, and Central Rockies hydrologic regions. In the California–Nevada hydrologic region, the water year with the lowest maximum SWE (2015) showed less than 30 percent of the average maximum SWE and was among the driest on record (Margulis and others, 2016). However, just 2 years later in water year 2017, the same hydrologic region had a maximum SWE accumulation >200 percent of average. This highlights the intense year-to-year variability in western snowpacks, making interannual water storage and planning both difficult and critical (Siirila-Woodburn and others, 2021). The California–Nevada and Central Rockies hydrologic regions had regionwide average maximum SWE of 40–60 mm/yr during water years 2010–20, which was similar to other northern regions across the CONUS. More southern hydrologic regions had much lower snowpack and SWE overall.

Although SWE is an important source of water supply, especially in the mountainous Western aggregated hydrologic region of the CONUS, its accumulation is subject to extreme spatial heterogeneity, meaning that large areas of these regions rely on water supplies that originate from a limited number of HUC12s. In the mountainous hydrologic regions within the western part of the CONUS (California–Nevada, Central Rockies, Columbia–Snake, Pacific Northwest, and Southwest Desert hydrologic regions), the HUC12s with the highest snowpack accumulate several meters of SWE during the cold months, which is much larger than the regionwide average. The Central High Plains and Northern High Plains hydrologic regions also contained HUC12s that could accumulate >1 m of maximum SWE. In other parts of the northern CONUS, the snowiest HUC12s accumulate several hundred millimeters of SWE.

The relative importance of SWE to water supply can be expressed as its ratio to annual precipitation (SWE/P). Maximum SWE in a HUC12 is the total amount of water stored as snowpack that can potentially become available for water supplies during snowmelt. This amount, in comparison to annual precipitation, provides perspective on the potential importance of SWE to overall water supplies. The Columbia–Snake, Central Rockies, and California–Nevada hydrologic regions had the highest average SWE/P; all were >0.1 (fig. 13B). Other hydrologic regions in the northern CONUS had much lower SWE/P despite having appreciable maximum SWE. The Souris–Red–Rainy, Great Lakes, Northeast, and Central High Plains hydrologic regions all had SWE/P values less than 0.05 (fig. 13B). Other parts of the CONUS had still lower values. However, SWE/P also had high spatial heterogeneity, as was also the case with maximum SWE.

Snow persistence (SP) is the amount of time that snow remains on the ground over a given time period, which can be useful for determining the geographic extent of snow coverage regimes; the timing of melt generation; and the boundaries between zones of low, intermittent, and persistent snow coverage (Bales and others, 2006). Following previous work, we defined ranges of SP for January–July of each year to compare coverage across hydrologic regions (Hammond and others, 2018; fig. 14). Any HUC12 was considered to have snow for a given month if SWE was >10 mm, and each HUC was classified as having (1) low snow if SP was less than or equal to (≤) 3 months; (2) intermittent snow if SP was >3 months and ≤6 months; or (3) persistent snow if SP >6 months. This analysis revealed that, in the snowiest hydrologic region of the CONUS (the Columbia–Snake hydrologic region), 53 percent of the region had at least intermittent snow, indicating relatively widespread snow coverage during the winter, whereas only 6 percent of the hydrologic region had persistent snow (fig. 14D). In contrast, in the Central Rockies and California–Nevada hydrologic regions—two regions that are heavily dependent on snowmelt for water availability—only 27 and 14 percent of the hydrologic regions had at least intermittent snow, respectively, and 4 percent and less than 1 percent had persistent snow, respectively (fig. 14D). The mountains of the Western United States are commonly referred to as the “water towers” of those areas because of their importance to water supply, especially during snowmelt in summer and early autumn (Viviroli and others, 2007; Scholl and others, 2025). The snow persistence results of the Western aggregated hydrologic regions (fig. 14A) align with the “water tower” perspective and emphasize that SWE storage and persistence have a skewed spatial distribution such that a small proportion of these regions is responsible for a large amount of the SWE storage and subsequent water supply in the regions.

Introduction

The storage of fresh water in lakes and reservoirs is a critical component of the hydrologic cycle, providing for water supply, as well as biogeochemical and ecological processing functions (Cole and others, 2007). Lakes and reservoirs are crucially important for water storage and can buffer water supplies during periods of low precipitation. The construction of reservoirs for water supply, flood control, recreation, hydropower, or other purposes has helped facilitate rapid growth and population expansion, particularly in the Western United States, but has also had a large impact on hydrologic flows (Magilligan and Nislow, 2005; Harvey and Schmadel, 2021). Although many of the largest lakes and reservoirs have well-characterized size, capacity, and fluctuations in storage volume, many of the smaller waterbodies, although still important, are less well-characterized. Neither of the hydrologic models used in this assessment simulate the amount of storage in lakes and reservoirs, so external sources were used to assess this storage component.

To estimate the water storage in lakes and reservoirs, we used the HydroLAKES global spatial dataset (Messager and others, 2016), the result of a geostatistical model using land-surface topography, which estimates the volume, surface area, and residence time of lakes with a surface area of 0.10 km² and greater. This dataset provides a fixed estimate of lake characteristics, and therefore, represents a static upper limit of storage capacity. The time-varying component to water storage in lakes and reservoirs is a first-order consideration, attributable to the drawdown of lake and reservoir volumes that occurs as a result of drought and (or) increased use. Such temporal dynamics are not captured in our assessment; nevertheless, the static estimate can provide important information examining regional differences and overall hydrologic budgets. The accuracy of the HydroLAKES dataset was estimated by comparing the modeled lake volume to independent datasets, which revealed a global coefficient of determination (R²) =0.92, with higher uncertainty associated with smaller lakes. The HydroLAKES dataset includes natural lakes and human-made reservoirs by incorporating data from the Global Reservoir and Dam database (Lehner and others, 2011). The HydroLAKES dataset includes coverage of the CONUS, OCONUS, and U.S. Virgin Islands. Lake- and reservoir-storage volumes are presented for all of these areas, but comparison to other storage components does not include the U.S. Virgin Islands and is limited to the CONUS and OCONUS.

Lake and Reservoir Results

Our analysis of the HydroLAKES dataset shows that the CONUS, OCONUS, and U.S. Virgin Islands have a total of 102,528 natural lakes and 1,891 reservoirs, with volumes of 23,567 and 715 km³, respectively (Messager and others, 2016). This is equivalent to 2,416 and 73 mm, respectively, when converted to a depth by dividing by the combined area of the CONUS, OCONUS, and the U.S. Virgin Islands, or a total of 2,489 mm. The Great Lakes (Lakes Superior, Michigan, Huron, Ontario, and Erie) store >95 percent (22,553 km³) of the total volume of lakes in the CONUS, OCONUS, and U.S. Virgin Islands. For this reason, the Great Lakes are broken out separately for the regional analysis, and their storage volumes are not attributed to any hydrologic region (fig. 15A, note the different scale). These large lakes and reservoirs, including the Great Lakes, contribute considerable storage during dry seasons and dry years, and are of substantial regional and national importance to water supply. Additionally, complex networks of water transfers have been developed that allow for water to be moved across hydrologic boundaries to buffer water supplies in other basins and regions (Siddik and others, 2023). Aside from the Great Lakes, the volume of water stored in reservoirs constitutes about 41 percent of total storage in lakes and reservoirs across the CONUS, OCONUS, and U.S. Virgin Islands. Although there are relatively few reservoirs compared to natural lakes, they are of great importance from a water-supply perspective. The spatial distribution of lakes and reservoirs is reflective of several natural and anthropogenic factors including natural depressions and more humid conditions in the Northeast through Midwest aggregated hydrologic regions, with recent glaciation and large constructed reservoirs in the Western aggregated regions (Brosius and others, 2021). When each of the Great Lakes is assessed individually and all other lakes and reservoirs are grouped by hydrologic regions, volume and surface area are highly correlated, with some exceptions (including the Great Lakes hydrologic region), indicating potentially shallower average lakes and reservoirs than in the rest of the hydrologic regions (fig. 15B). The California–Nevada hydrologic region stands out, with a large volume relative to surface area and number of lakes and reservoirs, largely because of the considerable volume of water stored in Lake Tahoe, which accounts for about 46 percent of the total lake and reservoir storage in that region. Hydrologic regions with lower total volume of storage generally show a higher fraction of that storage in reservoirs compared to lakes (fig. 15C). For example, Alaska is the hydrologic region with the highest volume and surface area and with approximately 100 percent in lakes, whereas The Mississippi Embayment hydrologic region has the lowest water storage in the CONUS and with >95 percent in reservoirs. Mountainous hydrologic regions (such as the Central Rockies, Columbia–Snake, and California–Nevada hydrologic regions) have more lake storage than reservoirs, whereas >90 percent of storage volume in the hydrologic regions of Texas, Southern High Plains, Tennessee–Missouri, and Central High Plains is within reservoirs.

The density of lakes and reservoir storage was calculated by dividing the total volume by the land surface area of the hydrologic region (fig. 15D). This showed that the Alaska, California–Nevada, the Gulf Coast, and the Northern High Plains hydrologic regions had the highest storage density. The Alaska hydrologic region has 59,481 lakes, more than six times more than the next highest hydrologic region, the Great Lakes hydrologic region (with 9,574 lakes). The Southwest Desert hydrologic region and the combined area of Hawaii, Puerto Rico, and the U.S. Virgin Islands constitute the other end of the distribution, with 222 and 70 total lakes and reservoirs, respectively.

Like the distribution of lakes worldwide, the number of lakes in the United States is skewed towards small lakes (that is, volume less than or equal to 0.1 km³; fig. 16). These small lakes are important for aquatic terrestrial interface (Pi and others, 2022) yet contribute less to storage of water. For example, our analysis showed that all but 1,212 lakes and reservoirs have a volume of ≤0.1 km³, accounting for 99 percent of the total number of lakes and reservoirs; however, these waterbodies account for only 1 percent of the total volume stored in lakes and reservoirs throughout the CONUS, OCONUS, and the U.S. Virgin Islands. Although the exact distribution of surface areas and volumes is uncertain, especially with regard to the abundance of small lakes, the analysis using modeled volumes referenced here addresses the regional trends and volume of water in lakes and reservoirs nationwide, and likely still holds true despite this uncertainty (McDonald and others, 2012; Verpoorter and others, 2014).

Methods

Changes in the depth of groundwater at representative monitoring locations provide an easily identifiable metric from which to visualize changes in groundwater storage. The USGS developed summaries of the status of groundwater levels and trends over the prior 21-year period (water years 2000–20) in 38 of the 67 principal and regional aquifers across the CONUS. The 38 aquifers analyzed accounted for about 82 percent of the aquifer withdrawals in the CONUS, OCONUS, and U.S. Virgin Islands in calendar year 2015 (Lovelace and others, 2020). The High Plains aquifer was divided into three regions for this analysis (north, central, and south) as done in previous efforts to estimate water-budget components in the High Plains aquifer (Stanton and others, 2011). Wells from aquifer systems with confined and unconfined aquifers were separated by confinement status and analyzed separately. Some aquifer systems have confined and unconfined parts such as the northern region of the High Plains aquifer, or the Basin and Range basins-fill aquifers. In these cases, the aquifers were analyzed separately for confined and unconfined conditions. We analyzed data from a total of 29 unconfined and 23 confined aquifers. Compared to confined aquifers, unconfined aquifers generally are nearer to the land surface and have the water table as the upper extent of the saturated zone. Confined aquifers are typically deeper than unconfined aquifers and are bounded by confining layers (aquitards) above and below (Freeze and Cherry, 1979).

Water-level data for each aquifer were downloaded from the National Groundwater Monitoring Network (NGWMN; U.S. Geological Survey, 2023a) using the dataRetrieval package in R (DeCicco and others, 2023). Data providers to the NGWMN provide the information on whether a well is monitoring a confined or unconfined aquifer. Data were retrieved from the NGWMN in April 2023, resulting in about 281,000 individual water-level measurements at 6,750 wells that had data in water years 2000–20. Water-level measurements were collected at varying temporal frequencies across the sites, and for those sites with multiple measurements in a single day, a daily measurement was randomly selected. In order to analyze data for an entire aquifer, the annual median water levels were determined for each well, and subsequently the annual water level of the aquifer was determined from a median of all wells in the aquifer. To be included in the analysis, each well had to have at least one water-level measurement during each water year 2000–20. This approach has been used for developing datasets to create aquifer composite hydrographs as described in Myers (2022).

An annual median water level was computed for each principal and regional aquifer and aquifer type (confined or unconfined) with at least 10 wells, and the annual aquifer-wide groundwater-level percentiles for these water levels were determined for water years 2000–20. Other aquifers with data in the NGWMN had an insufficient number of wells with long-term data for analysis. The medians of the calculated percentiles for the 11-year period (water years 2010–20) are presented in figure 18 and tables 4 and 5.

Groundwater Resources Results

Analysis of groundwater-level data from wells in unconfined aquifers showed that six aquifer systems had median groundwater-level percentiles less than (<) 30: (1) California Coastal Basin aquifers, (2) Colorado Plateaus aquifers, (3) central and (4) southern regions of the High Plains aquifer, (5) Rio Grande aquifer system, and (6) Snake River Plain basaltic-rock aquifers, indicating that the water levels were considerably lower than normal during water years 2010–20, relative to the 21-year period analyzed (fig. 18A; table 4; U.S. Geological Survey, 2023a). A threshold of the 30th percentile was chosen to highlight those aquifers with the lowest median groundwater levels. Several other aquifer systems had median groundwater-level values less than the 50th percentile, including the Basin and Range and Pacific Northwest basin-fill aquifers, as well as the Edwards-Trinity aquifer system and Pennsylvanian and Rush Springs aquifers. The median water-level percentiles in 14 of the unconfined aquifers, or about one-half of them, were >50, but only 2 aquifer groupings were >70 (the lower Tertiary aquifers and the Ozark Plateaus aquifer system; fig. 18A; table 4).

Our analysis of groundwater-level data from wells in confined aquifers showed results similar to those of the analysis for unconfined aquifers. Six aquifers had median groundwater-level percentiles <30 percent: (1) Basin and Range basin-fill aquifers, (2) California Coastal Basin aquifers, (3) Colorado Plateaus aquifers, (4) Colombia Plateau basaltic-rock aquifers, (5) Northern Atlantic Coastal Plain aquifer system, and (6) Rio Grande aquifer system (fig. 18B; table 5; U.S. Geological Survey, 2023a). Several other aquifers had median groundwater-levels <50 percent, including the Castle Hayne aquifer, northern region of the High Plains aquifer, Northern Rocky Mountains Intermontane Basins aquifer systems, Willamette Lowland basin-fill aquifers, and Coastal lowlands aquifer system. No confined aquifers had median water-level percentiles >70 percent (fig. 18B; table 5).

The analysis of groundwater-level data has some important limitations. Data availability is not evenly distributed among aquifers and subjectivity surrounds decisions about what constitutes a representative and sufficient record for inclusion in this analysis. The limited data availability resulted in some aquifers having a minimal number of wells (tables 4 and 5). Some of the aquifers analyzed had as few as 10 wells, which may not be representative of the entire aquifer. Additionally, the 21-year period of analysis was selected to include as many sites and aquifers as possible in the analysis. Many of these aquifers have documented depletion of groundwater that began during the 20th century (Konikow, 2013), so the starting point (water year 2000) was a time when some aquifers were already depleted, resulting in water levels that had already begun to decline well before the start of our analysis.

One of the major aquifers absent from this analysis is the Central Valley aquifer system in California, which is estimated to account for about 14 percent of the annual freshwater use in the CONUS, OCONUS, and the U.S. Virgin Islands, or >11,000 Mgal/d in 2015 (Lovelace and others, 2020). Other aquifers with considerable water use, missing from this analysis, include the Columbia Plateau basin-fill aquifers (391 Mgal/d) and the Texas coastal uplands aquifer system (375 Mgal/d; Lovelace and others, 2020). These aquifers were not included in the analysis because of insufficient data in the NGWMN.

Aquifers with low groundwater levels during water years 2010–20 likely represent a general decline in water levels in this 11-year period relative to the 21-year period; for example, the confined and unconfined Colorado Plateaus aquifers where the maximum percentile was 50 for both aquifer types and median percentiles were less than 30 (tables 4 and 5). However, some aquifers showed a wider range of variability throughout the period of analysis; for example, the confined Southeastern Coastal Plain aquifer system had a median groundwater-level percentile of 50 and minimum and maximum percentiles of 4.5 and 90.9, respectively, indicating water-level response to shorter-term events (table 5).

The central and southern regions of the High Plains unconfined aquifer system offer examples of clear declines over the period of analysis relative to water years 2000–20. These declines are part of the history of groundwater-level declines that have dated back to the middle of the 20th century, when irrigated agriculture became widespread in the area (McGuire, 2013). The High Plains aquifer supplies 12,300 Mgal/d, of which nearly 95 percent is used for irrigation (Lovelace and others, 2020). Although irrigation withdrawals are high in the northern region of the High Plains aquifer, water levels have not shown the same degree of consistent decline, owing largely to the differences in natural recharge, which is higher in the northern region and lower in the central and southern region of that aquifer (Scanlon and others, 2012). This pattern has led to the conclusion that much of the withdrawals in the central and southern region are from “fossil groundwater” that was recharged over the last 13,000 years (McMahon and others, 2004). The declines in water level in this analysis have occurred even as withdrawals from the High Plains aquifer decreased by 32 percent (in 2015 relative to 2000; Lovelace and others, 2020), suggesting that despite efforts to reduce levels of groundwater withdrawal, these withdrawals are far from balanced with natural recharge.

The analysis of the status of groundwater storage for this study aligns with previous analyses of groundwater storage for principal aquifers (Konikow, 2013). Aquifers with groundwater-level percentiles <50 in this analysis that were also identified as depleted in the Konikow analyses include the High Plains aquifer, Coastal lowlands aquifer system, and North Atlantic Coastal Plain aquifer system. The depletion could be a result of a combination of factors including increases in use and (or) decreases in natural recharge from shifts in climate or land use. The results from this analysis differ from previous results for the Cambrian-Ordovician and Mississippi embayment aquifer systems, where groundwater levels during water years 2010–20 were in the normal range (>30th percentile) relative to the 21-year period. Differences can likely be attributed to different methods and time periods of analysis. Additionally, some of the aquifers were split into smaller local aquifers in previous studies, making direct comparisons difficult.