1.1. Purpose and Scope

Sustainability of groundwater resources in Virginia Coastal Plain aquifers is a concern for water-resource managers and planners. The U.S. Geological Survey (USGS) and the VDEQ have cooperated on numerous studies to better understand the hydrogeologic framework, groundwater flow, and water use of the Virginia Coastal Plain aquifers (including McFarland and Bruce, 2006; Pope and others, 2008; Heywood and Pope, 2009; and Pope and others, 2016). This study builds upon previous USGS investigations, namely Pope and others (2008), to update estimation of domestic water use in the Virginia Coastal Plain across hydrogeologic, spatial, and temporal scales. These revisions will support higher resolution groundwater modeling and better inform water-resource managers and decision makers. Because each component of the revised analysis—distribution of wells among the aquifers, spatial distribution of users, and temporal distribution and amount of withdrawals—was completed with independent data sources and methods, this report covers each topic separately and in turn.

1.2. Description of Study Area

The Virginia Coastal Plain occupies the entire eastern part of Virginia between approximately longitude 77°30′ W and the Atlantic Ocean, including the Virginia Eastern Shore on the peninsula between the Chesapeake Bay and the Atlantic Ocean (fig. 1). The area is bordered by Maryland and North Carolina to the north and south, respectively. Geologically, the Virginia Coastal Plain is defined by a wedge of unconsolidated sediments of fluvial-deltaic and marine origin that rests on consolidated bedrock, hereafter referred to as the basement. This sediment wedge thins to the west and pinches out in the Fall Zone, where crystalline rock of the Piedmont Physiographic Province outcrops. Along the Atlantic Coast of Virginia, the total thickness of Virginia Coastal Plain sediments is greater than 6,000 feet (ft; McFarland and Bruce, 2006).

Fifty Virginia counties and independent cities, henceforth referred to as localities, lie entirely or partially within the Virginia Coastal Plain. Population is concentrated in large urban areas, including the northern Virginia metropolitan area near Washington, D.C.; the Cities of Fredericksburg, Richmond, and Petersburg along the western boundary of the Coastal Plain; and the Cities of Chesapeake, Hampton, Norfolk, Portsmouth, and Virginia Beach in the southeast near the mouth of the Chesapeake Bay (fig. 1; U.S. Census Bureau, 2021). The remainder of the Virginia Coastal Plain is more sparsely populated and includes small towns and outlying rural areas of forest, agriculture, and wetland. The climate of the Virginia Coastal Plain is humid and temperate, with annual mean precipitation of approximately 45 inches (National Oceanic and Atmospheric Administration, 2024). The topography of the Virginia Coastal Plain is characterized by rolling terrain with deeply incised stream valleys in the northwest and gently rolling to level terrain with broad stream valleys in the east and south. Land-surface elevations generally decline seaward, from over 300 ft in the western Virginia Coastal Plain to 0 ft (sea-level) along the Atlantic coast. Several major rivers drain eastward into the Chesapeake Bay, creating three prominent peninsulas. From north to south these peninsulas are referred to as the Northern Neck (between the Potomac and Rappahannock Rivers), the Middle Peninsula (between the Rappahannock and York Rivers), and the York-James Peninsula (between the York and James Rivers; McFarland and Bruce, 2006). These major rivers become brackish as they enter estuarine areas east of the Fall Zone. Consequently, the Virginia Coastal Plain is heavily reliant upon groundwater resources (Pope and others, 2008).

The USGS previously estimated groundwater use at 5-year intervals for the 50 localities all or partially within the Virginia Coastal Plain as part of national water-use data compilation efforts (table 1; fig. 2). Note that this estimated groundwater use is for the entire locality and not limited to the Virginia Coastal Plain. From 2000 to 2015 (the latest year for which USGS data were available), overall groundwater use has slightly decreased, predominantly because of a decline in groundwater use for industrial activity since 2005 (U.S. Geological Survey, 2024b). By 2015, self-supplied domestic groundwater withdrawals collectively accounted for 46 million gallons per day (Mgal/d) and made up the largest portion of groundwater use (39 percent). Industrial use of groundwater accounted for 39 Mgal/d (33 percent), and public supply from groundwater accounted for 24 Mgal/d (21 percent).

Most of the Virginia Coastal Plain is included within the two State-designated Groundwater Management Areas: Eastern Virginia and Eastern Shore. In these areas, large groundwater withdrawals are regulated by the VDEQ, which generally requires well owners to obtain a permit to withdraw 300,000 or more gallons per month. Under the conditions of these permits, well owners collect various information, including well construction details, withdrawal amounts, and water-level and (or) water quality data. This information is reported to the VDEQ to support greater understanding of the effect large withdrawals have on the aquifers of the Virginia Coastal Plain (B. Campbell, Virginia Department of Environmental Quality, written commun., 2025).

In comparison, information from the approximately 200,000 private-domestic wells in the Virginia Coastal Plain is far less comprehensive (Pope and others, 2008). Well-completion records are required by Virginia state regulation and maintained in various forms by the Virginia Department of Health (VDH) or the VDEQ, depending on the well’s age and location within the Groundwater Management Areas. The availability, quality, and variability of well-completion records represents change in reporting requirements, data entry, and data management over time, as well as the different, and sometimes overlapping, authorities of the VDH and the VDEQ over private-domestic wells in the Virginia Coastal Plain (B. Campbell, Virginia Department of Environmental Quality, written commun., 2025). Efforts to improve data collection and move from paper records to electronic well databases are ongoing among both agencies, but legacy records still contain missing or imprecise information (Pope and others, 2008). Given the magnitude of self-supplied domestic water use and the lack of complete data for private-domestic wells, accurate estimation of the source aquifer, location, and rates of these groundwater withdrawals is important to better understand the water resources of the Virginia Coastal Plain.

1.3. Aquifers of the Virginia Coastal Plain

The Virginia Coastal Plain aquifer system is a part of the more extensive Northern Atlantic Coastal Plain aquifer system. The geology of this region consists of a seaward-thickening wedge of eastward-dipping strata of unconsolidated to partly consolidated Cretaceous, Tertiary, and Quaternary sediments that overlie a basement of consolidated bedrock (fig. 3; McFarland and Bruce, 2006). The groundwater within the Virginia Coastal Plain primarily originates from precipitation that infiltrates the surficial, water-table aquifer and then flows laterally to discharge in nearby streams or downward to recharge deeper confined aquifers (Pope and others, 2008). The hydrogeology of the Virginia Coastal Plain is covered extensively in previous USGS publications, including McFarland and Bruce (2006), Pope and others (2008), Masterson and others (2016), and McFarland and Beach (2019). A brief overview of the aquifers of the Virginia Coastal Plain is provided in this report to provide context for the analysis herein.

2.1. Study Methods

Electronic private-domestic well-completion records from Virginia Coastal Plain localities were obtained from the Virginia Department of Health (VDH) Environmental Health Database, the successor to the Virginia Environmental Information System (M. Senseman, Virginia Department of Health, written commun., 2021). Coordinate locations provided by the VDH for these wells were derived from the Virginia Geographic Information Network (VGIN) Composite Geocoding Service using the North American Datum of 1983 (NAD 83; Virginia Department of Emergency Management, 2013). VDH well records without VGIN geocoding were processed with the Esri ArcGIS World Geocoding Service (Esri Inc., 2023) and projected using NAD 83. Many of the VDH well records were “location only” and therefore lacked data on well or screen depth. Of the 100,930 well records provided by the VDH, 4,884 (4.8 percent) had usable location and well or screen depth information.

Additional electronic well-completion records from Virginia Coastal Plain localities were obtained from the VDEQ Groundwater Characterization Program (T.S. Bruce, Virginia Department of Environmental Quality, written commun., 2019). VDEQ well records were filtered by “domestic” or “household” use categories to identify private-domestic wells. Well coordinates were provided using NAD 83. The source of coordinate data was an included field of the well construction record, indicating use of GPS or map interpolation, but most of the records were reported without this information. From 3,555 VDEQ private-domestic well construction records, 3,401 (97 percent) had usable location and well or screen depth information. Twenty-one duplicate well records shared by the VDH and the VDEQ were removed before combining the agencies’ data into a common dataset with location and well or screen depth information for 8,264 private-domestic wells in the Virginia Coastal Plain.

Elevations for the land surface and subsurface hydrogeologic layers at each well coordinate were obtained by extracting raster values from a Digital Elevation Model (DEM; 100 ft grid cells) and the digital representation of the individual aquifers within the Northern Atlantic Coastal Plain aquifer system (1/2-mile grid cells; Pope and others, 2016) using ArcGIS Pro (Esri Inc., 2023). Ladd and others’ (2024) hydrogeologic framework (125-meter grid cells) was used for wells in the Virginia Eastern Shore. A script was written in the statistical programming language “R” to compute the well-screen elevation using the DEM land-surface elevation and screen depth (R Core Team, 2023). For wells without screen depth information, the well depth or hole depth was assumed equivalent to the screen depth. Where well construction records indicated multiple well screens (44 records), the bottom of the bottommost screen was used (greatest depth). The well-screen elevation was compared to the aquifer elevations in a scripted series of logical expressions to estimate the aquifer intersected by each well (fig. 4).

For any well initially determined to exist in a confining unit (21 percent), the script reassigned the well to the nearest available aquifer above or below that confining layer using the absolute value of the elevation differences. Complex hydrogeology (for example, two confining units adjacent to each other without an aquifer in between) confounded the scripted logic and 261 well records were flagged for manual review and corrected to the nearest aquifer. A well located in the confining unit could be an error related to the 1/2-mile resolution of the hydrogeologic information; an error in either coordinate, screen, or well-depth information; or poor well-screen placement. However, local-scale geologic variability, including potential heterogeneity in the confining unit itself, could allow a well drilled in the confining unit to yield adequate water (Pope and others, 2008). For the purposes of this study, all wells were assumed to utilize the known aquifers.

The method used to estimate the distribution of private-domestic wells among Virginia Coastal Plain aquifers is summarized in figure 5.

2.2. Discussion of Results

The basic distribution of domestic well aquifer utilization in the Virginia Coastal Plain matches current understanding of aquifer extent, characterization, and yield (fig. 6). Shallow wells in either the surficial aquifer or the basement predominate in the northern and western extremes of the Virginia Coastal Plain near the Fall Zone. Deeper aquifers, particularly the Potomac aquifer, are used throughout the central portion of the Virginia Coastal Plain. In the south and east near the mouth of the Chesapeake Bay and along the Atlantic coast, use of shallower aquifers like the Yorktown-Eastover and surficial aquifers is more prevalent because the deeper aquifers become brackish.

Overall, as shown in table 2, available well records suggest that the greatest domestic use occurs in the Yorktown-Eastover aquifer (35 percent), followed closely by the Potomac aquifer (31 percent). The surficial aquifer accounts for 21 percent of the aquifer distribution, with the remaining 6 aquifers and the basement combined for the final 13 percent of the domestic use. Despite the use of more generalized well and hydrogeologic information in Pope and others (2008), the private-domestic well aquifer distribution using the revised method in this report corroborates many of the findings of the previous effort. The most substantial differences exist with the smallest aquifers, especially the Aquia aquifer, because the generalization techniques of 2008 likely assumed a larger aquifer extent or thickness than those digitally modeled by Pope and others (2016). The difference in basement wells in the two analyses is due primarily to the exclusion of certain counties in the northern and western extremes of the Virginia Coastal Plain by Pope and others (2008).

With 8,264 location-specific data points, including the estimated aquifer assignment of each well, aggregation of aquifer utilization can occur on various spatial scales and with different boundary datasets, including Virginia Coastal Plain localities (table 3). The USGS Watershed Boundary Dataset (U.S. Geological Survey, 2023a), includes watersheds of various nested sizes defined by nested hydrologic units that may provide another useful spatial context for groundwater withdrawals, particularly for hydrologic modeling or water budgets. Appendix 1 contains aquifer distribution by watershed at the 8-digit hydrologic unit code (HUC8) scale and appendix 2 contains aquifer distribution by watershed at the 10-digit hydrologic unit code (HUC10) scale.

Caution is advised when using these aggregation results. As spatial units become smaller, fewer wells exist in each unit, approaching the limits of representative aggregation and increasing uncertainty. Furthermore, because geopolitical or watershed boundaries do not necessarily align with hydrogeologic ones, comparisons from one scale to another are difficult to make. For example, while a certain percentage of a locality’s total number of wells may be in the basement, the use of that hydrogeologic unit declines rapidly eastward of the Fall Zone.

Tabular data of the aquifer distribution of private-domestic wells for the Virginia Coastal Plain at both the county- and watershed-scale are available as a USGS data release in Kearns and Pope (2025).

2.3. Assumptions and Limitations

The primary limitation of the aquifer distribution analysis is the quality and availability of the electronic private-domestic well data from the VDH and the VDEQ. These data are derived from Water Well Completion Reports submitted to the VDH and the VDEQ. The availability, quality, and variability of electronic well-completion records represents change in reporting requirements, data entry, and data management over time, as well as the different, and sometimes overlapping, authorities of the VDH and the VDEQ over private-domestic wells in the Virginia Coastal Plain (B. Campbell, Virginia Department of Environmental Quality, written commun., 2024).

A further limitation is whether these electronic well-completion records are representative of the actual distribution of aquifer use. The electronic well-completion records used in this analysis were not systematically sampled but compiled from available sources. This methodology introduces uncertainty while also increasing the sample size nearly three times over the number of paper well records reviewed by Pope and others (2008). The 8,264 electronic well records in this analysis represent approximately 4 percent of the estimated 200,000 private-domestic wells in the Virginia Coastal Plain (Pope and others, 2008). A sample of this size is adequate to statistically represent the population with a margin of error of about 1 percent with a standard 95-percent confidence interval. Although the computed numerical sampling error is small, the other potential sources of uncertainty discussed here are difficult to quantify and likely vary spatially across the study area.

The methods in this report inherit any error or bias present in the electronic well-completion records. Small coordinate errors, in either measurement or data entry, that incorrectly place the well elsewhere in the Virginia Coastal Plain are difficult to detect, and records using geocoding for coordinates are less precise than coordinate measurements made at the well. Horizonal datum and method of coordinate determination were seldom recorded; NAD 83 was assumed in these instances. One potentially significant source of bias could be the age of the well records; for example, a bias towards modern wells having electronic records. Unfortunately, the inconsistent format used for data entry of the “date” field in the well records prevented machine-readable import and manual entry¹ of that variable and any analysis of a temporal trend or bias in the data. Additionally, some well records could be for out-of-service private-domestic wells, particularly in the Cities of Norfolk and Chesapeake or in other areas with rapidly expanding public water system (PWS) service areas.

General private-domestic well data availability was also a limiting factor, with unequal sample size across Virginia Coastal Plain localities. Sparse coverage of well records in the southwest corner of the Virginia Coastal Plain and areas with a larger portion of public supply water use results in some localities or watersheds with few records and thus, higher uncertainty in the aquifer distribution. Nine counties had fewer wells included in this revised aquifer analysis than the one completed by Pope and others (2008) using paper records. Conversely, an abundance of electronic well records from the Virginia Eastern Shore, which lacks aquifers deeper than the Yorktown-Eastover aquifer, potentially skews overall aquifer source distribution of the Virginia Coastal Plain. Aquifer distributions might also vary significantly within a locality, depending on the scale of analysis and complexity of local hydrogeology. For example, although localities in the west of the Virginia Coastal Plain may have an overall percentage of wells in the basement, the likelihood of basement wells decreases rapidly with distance east of the Fall Zone.

Finally, there is the limitation of the current hydrogeologic framework of the Virginia Coastal Plain. As a subset of the digital model for the greater Northern Atlantic Coastal Plain (Pope and others, 2016), Virginia Coastal Plain aquifer characterization is available in 1/2-mile grid cells. This resolution supports the analysis of general aquifer-utilization patterns across a region; however, variability in hydrogeology or other unique, localized conditions could lead to discrepancies at finer scales. Contemporary understanding of the aquifers of the Virginia Coastal Plain suggests small heterogeneities and discontinuities are present in many of the aquifers. However, the overall similarities in the well distribution between this report and the previous efforts of Pope and others (2008) using different methods and data suggests a relative insensitivity of general private-domestic well patterns to minor variations. Such similarities may also retroactively confirm the assumptions and generalizations made by Pope and others (2008) in adapting the hydrogeologic framework at the locality scale for the low-relief topography and hydrogeology of the Virginia Coastal Plain.

3.1. Study Methods

The 2020 census-block population, housing, and geospatial data for Virginia were downloaded from the TIGER line dataset published by the U.S. Census Bureau (2021). The ratio of households using private-domestic wells in each census block was determined by dividing the 2020 estimated well count from Murray and others (2023) by the 2020 census housing-unit count. This well-use ratio was then multiplied by the 2020 census-block population and rounded to the nearest whole person to estimate the self-supplied domestic population in each census block. Estimates for the 2010 self-supplied domestic population and number of wells, used for comparison to previously published USGS estimates, were obtained by weighting the 2010 census block population and housing-unit counts to 2020 census block geographies using National Historic Geographic Information System (NHGIS) crosswalk tables (Manson and others, 2023), multiplying each by the 2020 well-use ratio, and rounding to the nearest whole number.

The primary dataset used for the spatially weighted estimated distribution of self-supplied domestic population was a 2018 building footprint sum analysis at 30-meter resolution, available for each state in the conterminous United States, by Heris and others (2020). For example, a 30-meter cell that was 50-percent covered by building footprints would have a cell value of 450 square meters. The building footprint dataset for Virginia (Heris and others, 2020) was re-projected in ArcGIS Pro to match the grid of the 2019 National Land Cover Database (Dewitz, 2021). A “Zonal Statistics” ArcGIS Pro tool was used to calculate the total building footprint area from all 30-meter cells within each 2020 census block (Esri Inc., 2023). The previously estimated self-supplied domestic population for each census block was divided by this total building footprint area, to derive a density value of the estimated self-supplied domestic population per square meter of building footprint. Using the “Polygon to Raster” and “Raster Calculator” ArcGIS Pro tools (Esri Inc., 2023), the density values unique to each 2020 census block were converted to a raster format and multiplied by the sum value of each cell in the building footprint dataset. This operation converts the density value into a number representing the estimated self-supplied domestic population for each 30-meter building footprint cell within any given census block.

Because of errors in either census or building footprint data, 2 percent of 2020 census blocks had a self-supplied domestic population greater than zero where no building footprints were identified. For these census blocks, the 2019 National Land Cover Database (NLCD; Dewitz, 2021) was used as an alternate data source. The NLCD includes a developed cell classification generally defined by the percentage of impervious surface cover (Dewitz, 2021). Although impervious features other than residential housing (including roads, parking lots, or commercial and industrial structures) make the NLCD less desirable for spatially distributing the self-supplied domestic population, the wider categorization ensures fewer census blocks are omitted from the analysis. The 30-meter cells classified as high-, medium-, and low-intensity developed were extracted from the NLCD and converted to a raster value of 900, 600, and 300, respectively. This conversion approximated square-meter values from the Heris and others (2020) building footprint summary dataset based on the percent of impervious cover from the NLCD cell classification. These NLCD-derived proxy values then were totaled for each census block and used in the same density and population conversion operations and calculations as the building footprints.

The final raster analysis combined the self-supplied domestic population estimated from the building footprints with the self-supplied domestic population estimated from the NLCD. By combining the estimates, the number of census blocks with a self-supplied domestic population greater than zero but with no identified building footprints or NLCD developed cells decreased to 0.1 percent of 2020 census blocks, representing 0.08 percent of the 2020 self-supplied domestic population. Without a corresponding building footprint or NLCD developed cell within the census block, this small part of the self-supplied domestic population is lost from the analysis. Temporal differences in the estimated population (2020), building footprints (2018), and NLCD (2019) data sources may provide some explanation of this variability, as does the potential of intentional noise or unintentional miscounts in the census data itself (Hill and others, 2022; U.S. Census Bureau, 2022).

Note that the mathematical operations of this disaggregation method will produce non-integer population estimates in each raster cell. The authors recommend that, if any rounding is necessary, it is delayed until raster cells are aggregated to the desired spatial zone or scale. For example, as a quality control check on the disaggregation technique, ArcGIS Pro’s “Zonal Statistics” (Esri Inc., 2023) was used to sum the 30-meter raster of estimated 2020 self-supplied domestic population for each census block in the Virginia Coastal Plain and compare it to the original values calculated from the 2020 census-block population and ratio of well use from Murray and others (2023). Aside from the 0.08 percent population discrepancy for census blocks without qualified building or development land-cover cells previously noted, the results were equivalent.

Self-supplied domestic population estimates were constrained to the Virginia Coastal Plain by extracting only those raster cells within the spatial extent of the Northern Atlantic Coastal Plain (Pope and others, 2016) with the “Extract by Mask” ArcGIS Pro tool (Esri Inc., 2023). The method used to estimate the spatial distribution of the self-supplied domestic population across the Virginia Coastal Plain is summarized in figure 8. Thirty-meter rasters of the estimated self-supplied domestic population of the Virginia Coastal Plain for 2020 and 2010 are available as a USGS data release in Kearns and Pope (2025).

3.2. Discussion of Results

In 2020, there were an estimated 475,332 people, approximately 12.5 percent of the total population, that depended upon private-domestic wells across the Virginia Coastal Plain.² Because the 30-meter population raster produced in this analysis has cells too small for practical use as a page-sized figure of the Virginia Coastal Plain in this report, the USGS 1-kilometer National Groundwater Model Grid (Clark and others, 2018) was used to aggregate the data release and is presented in figure 9. This figure indicates that, although the proportion of the population served by private-domestic wells is highest in the most rural areas of the Virginia Coastal Plain, a higher total population shifts the areas of greatest overall self-supplied domestic population to the suburban fringes of the larger cities. Other regions with significant self-supplied domestic population include the Virginia Eastern Shore and the Middle Peninsula.

The self-supplied domestic population of the Virginia Coastal Plain increased 5.7 percent during the period 2010–20; this increase is slightly less than the total population change of approximately 8 percent (fig. 10; table 4) estimated from census blocks within the Virginia Coastal Plain. The suburban fringe experienced the most self-supplied domestic population growth, and the more rural areas held constant or experienced slight decreases in population. As mentioned in section 3.1, it is critical to note that in the context of this analysis, any change in the self-supplied domestic population from 2010 to 2020 is due solely to changing population dynamics in census blocks dependent upon private-domestic wells. The changes presented here are not due to expansion of PWS service areas over time, and the rate of well use was held constant from 2010 to 2020 using Murray and others (2023).

The spatial disaggregation technique used in this study permits analysis of self-supplied domestic population at scales other than census blocks or county boundaries. Appendix 1 contains an estimate of the 2010 and 2020 self-supplied domestic populations by HUC8 watersheds and appendix 2 contains the 2010 and 2020 self-supplied domestic populations by HUC10 watersheds. Note that because these two different spatial units were rounded to the nearest whole person, totalized self-supplied domestic population values may vary slightly.

Comparisons to previous self-supplied domestic water-use studies can help ascertain the uncertainty of the estimates in this report. Pope and others (2008) had previously estimated the domestic self-supplied population of the Virginia Coastal Plain to be 514,390 in 2000. However, their 2000 estimate is higher than the 475,332 estimated by this study for 2020, despite the general population growth in the Virginia Coastal Plain in the intervening 20 years, and possibly indicative of even greater uncertainty in the methods or datasets used by Pope and others (2008). Other comparisons are possible using county-level estimates of water use for the Nation, including self-supplied domestic population, published by the USGS every 5 years since 1985; however, a transition in the methodology used at the national level has delayed the release of the 2020 estimates (U.S. Geological Survey, 2024a). Therefore, at the time of publication, estimates made for the 2010 self-supplied domestic population by Maupin and others (2014) and Johnson and others (2019) provide the best opportunity to assess previous results against the analysis of this report (table 5). To compare equivalent population estimates, both Johnson and others (2019) and the results of this analysis were aggregated to the locality level for the 50 localities with all or part of the locality in the Virginia Coastal Plain. In other words, although the population numbers presented in table 4 only include the portion of each locality within the Virginia Coastal Plain, the population numbers in table 5 also include the portions of those localities outside the Virginia Coastal Plain for the sake of comparison.

The 2010 estimate of self-supplied domestic population from this analysis closely aligns with that from Maupin and others (2014). It is 27 percent higher than the estimate made by Johnson and others (2019), which was done at a national scale with coarser census-block groups. Direct comparison beyond 2010 is difficult because of the temporal differences in the datasets. However, a relative comparison suggests a ratio of self-supplied domestic population to total population of 12.6 percent in 2010 (Maupin and others, 2014), 10.7 percent in 2015 (Dieter and others, 2018), and 12.4 percent in 2020 from this analysis. Given the reliance on private-domestic well-use data from 1990, the difficulty of obtaining datasets that represent the continued expansions to PWS service areas, and the lower estimates from Dieter and others (2018) and Johnson and others (2019), it is likely that the self-supplied domestic population estimates from this report are toward the upper end of the range of actual private-domestic well use.

3.3. Assumptions and Limitations

One of the primary limitations of self-supplied domestic population estimation is the reliance on the 1990 long-form census. For current high-resolution water-use modeling, this original census-block group dataset is coarse in scale and dated. Thirty years of subsequent development and population growth have undoubtedly altered the private well and public-supply conditions extant during the 1990 census. Housing density thresholds have been explored by Johnson and others (2019) and Murray and others (2021) to account for new development and accompanying expansion of public supply; however, these modeled relations will not precisely hold in every census block. Furthermore, the retroactive application of 2020 census block private-well-use rates to 2010 in this report would capture increases and decreases in the self-supplied domestic population solely from census-block demographic change and not from PWS service area expansions.

Good data on the current extent of PWS service areas could supplement the 1990 long-form census data and other estimates of PWS or private well distribution, including Murray and others (2023). The authors of this report found PWS data within the Virginia Coastal Plain to be inconsistent, available in a patchwork from local governments or PWS, and with substantial variation in data quality, publication date, and update frequency. With a lack of consistent PWS service area data for the entire Virginia Coastal Plain, the decision was made to use nationally available and (or) peer-reviewed datasets for our analysis.

For further analysis at local scales or in the fastest growing areas, the 30-meter raster dataset of 2020 self-supplied domestic population estimates produced by the methodology described in this report could potentially be improved by using geographic information system (GIS) clipping or masking tools to remove the raster cells of self-supplied domestic population within a given spatial boundary dataset of a PWS service area. A PWS service area may contain private-domestic wells, but use of such wells in those areas would likely be greatly reduced. The USGS offers a national-scale estimate of PWS service area boundaries (Buchwald and others, 2022); however, service area proxies, such as municipal political limits, were used in areas of limited data availability, including for PWS in the Virginia Coastal Plain. Therefore, given the spatial and temporal variability of PWS service areas over time, obtaining the most recent GIS files from local authorities likely would prove more accurate. For example, the use of a 2009 GIS file of PWS service areas within the Hampton Roads Planning District Commission (HRPDC; I. Ozmon, Hampton Roads Planning District Commission, written commun., 2024) to extract the 2020 self-supplied domestic population estimated from this analysis described in this report suggested 62,174 people use private-domestic wells within the HRPDC PWS service areas, against a total 2020 population of 1,415,651 (estimated from census blocks within the HRPDC PWS service areas; U.S. Census Bureau, 2021). Within the greater HRPDC PWS service area, the Virginia Beach PWS accounts for the largest share of this potential discrepancy, with an estimated 22,162 self-supplied domestic population within the PWS service area of 434,245 people (approximately 5 percent). It is known that some private-domestic well use is occurring in the Virginia Beach PWS service area, particularly for outdoor water use (Eggleston, 2010); however, it is likely that the water-source data from the 1990 census and subsequent adjustments by Murray and others (2023) are not keeping pace with actual PWS expansion in this area. By applying this example to other fast-growing localities in Virginia Coastal Plain, the authors’ previous conclusion that this analysis likely represents the upper end of the actual self-supplied domestic population is further supported.

Other assumptions and limitations apply to the distribution of the self-supplied domestic population within a census block. The primary dataset used for internal census block disaggregation was the building footprints estimate at 30-meter resolution (Heris and others, 2020). Because this dataset makes no distinction between residential, commercial, industrial, institutional, or auxiliary structures, the method described herein equally distributed any self-supplied domestic population evenly across any building structure footprint within the census block. This does lead to error in “mixed use” census blocks where residential buildings are intertwined with other structures and population is distributed across nonresidential building footprints. However, because each census block has a unique self-supplied domestic population density value, any error within a census block is confined to that block, minimizing the overall impact of any erroneous distribution.

The building footprint dataset from Heris and others (2020) did not identify buildings in approximately 2 percent of census blocks with a population greater than zero. This is due either to an error with the census reporting a population in a block that has none, or an error within the building footprints dataset. The authors of this report noted possible remote sensing errors where isolated linear gaps in the building footprint dataset exist; these errors were most apparent in the Richmond suburbs. Furthermore, the mismatch between the 2018 building footprint dataset and the 2020 census suggests the possibility of new residential construction in certain census blocks that would not show up in the slightly older building footprints. To fill in these data gaps, the 2019 NLCD was chosen to supplement the building footprint dataset because of its national extent and proximate temporal period. However, the NLCD categorization is broader than the building footprint dataset, encompassing all impervious surfaces—including buildings, parking lots, and roads—within the “developed” land-use classification. This broader definition ensures fewer census blocks with a population greater than zero are left without qualifying raster cells for the population density metrics performed in this analysis but would suggest that the self-supplied domestic population within those census blocks are distributed across cells that might not contain only residential structures. However, this tradeoff in using a broader dataset ensures that error of omission is reduced from 2 percent to 0.1 percent of census blocks when the NLCD density analysis is combined with the building footprint density analysis in the final estimate of self-supplied domestic population. The remaining 0.1 percent of census blocks, representing 0.08 percent of the self-supplied domestic population, have no “qualifying” building footprint or NLCD developed raster cells to place the estimated population and are lost in the final raster analysis. Some of this discrepancy could be explained by errors in the 2020 census (such as estimated counts of people in blocks that have zero actual population) or by the statistical noise added by the U.S. Census Bureau to protect individual privacy (Hill and others, 2022; U.S. Census Bureau, 2022).

Although this method for spatial disaggregation is intended to improve the resolution of self-supplied domestic population estimates, as the desired spatial coverage decreases in size (smaller area) the uncertainty of the estimate within proportionally increases. The intent of this data analysis was to support regional groundwater modeling and management efforts with reasonably accurate population and distribution estimates. Groundwater models or other use cases at extremely localized scales would benefit from additional data collection or verification of the estimates herein.

4.1. Study Methods

This study refines the seasonal variability in domestic water use in the Virginia Coastal Plain with a localized, data-driven approach. Using a fundamental assumption that households on public or private water supply would have similar water-use patterns, monthly water-withdrawal records from 293 PWSs from localities within the Virginia Coastal Plain were obtained from the VDEQ (G. Mahadwar, Virginia Department of Environmental Quality, written commun., 2022) and assessed for temporal variation between 2010 and 2019.

Forty-five of these PWSs were excluded from further analysis because of seasonality-exhibiting outlier behavior, defined in this report as any monthly water use above or below three standard deviations from the mean. PWSs with outlier monthly water use likely serve transient or non-community populations, such as a seasonal campground, that do not represent the water-use patterns of primary residents included within the self-supplied domestic population estimated from census data.

Depending on the community, a large portion of PWS withdrawals could be used for nondomestic (commercial, industrial, and so on) use, lost to leaks in the distribution system, or other factors generally uncharacteristic of private-domestic water use (Kenny and Juracek, 2012). Initial analysis suggested using the aggregate withdrawal for all PWSs would skew any temporal signal in favor of the biggest systems; the four largest PWSs account for more than half of all public-supply withdrawals in the Virginia Coastal Plain. An additional 28 PWSs with an average withdrawal over 1 Mgal/d from 2010 to 2019 were therefore excluded from further analysis to isolate the smaller PWSs (assumed to be the closest match for domestic water use patterns) within the Virginia Coastal Plain. Using 80 gal/d per capita, this 1 Mgal/d threshold would approximate a community of 12,500 or fewer people. The remaining 220 systems, which withdrawal less than 1 Mgal/d, comprise 70 percent of the total number of PWSs in the Virginia Coastal Plain, but only account for 2.5 percent of the total volume of PWS water withdrawal.

For these 220 PWSs within the Virginia Coastal Plain, the cumulative average withdrawal for each month from 2010 to 2019 was divided by the days in each month. The monthly average gallons-per-day withdrawal was compared to the annual average gallons-per-day withdrawal, creating 12 monthly-to-annual water-use ratios. These monthly-to-annual ratios were multiplied by the annual average 80 gal/d per capita water-use established by Dieter and others (2018) to estimate the variation of monthly domestic water use (table 6).

The method used to estimate the temporal distribution of self-supplied domestic water withdrawals in the Virginia Coastal Plain is summarized in the flow chart of figure 11.

4.2. Discussion of Results

Applying Dieter and others’ (2018) annual estimate of 80 gal/d per capita to the monthly variability of PWS withdrawals in the Virginia Coastal Plain suggests domestic water use ranges from 72 to 93 gal/d per capita over the course of a year. The general temporal variability suggests withdrawals from June to August are 12 percent higher than the annual average, and withdrawals from December to March are 8 percent lower than the annual average (fig. 12). This range is within the seasonal variability of self-supplied domestic water withdrawals from other studies in regions climatically similar to the Virginia Coastal Plain (Shaffer and Runkle, 2007; Shaffer, 2009; Kenny and Juracek, 2012).

Although the confined aquifers of the Virginia Coastal Plain are not affected by seasonal changes in groundwater recharge and availability, approximately 20 percent of private-domestic wells are screened in the surficial aquifer, as discussed in section 2.2 of this report. The higher rate of self-supplied domestic water withdrawal in summer coincides with elevated groundwater withdrawal from other uses (like irrigation) and lower groundwater recharge, driven by increased evapotranspiration. This seasonal mismatch in groundwater demand and supply from the surficial aquifer may strain resources during dry seasons or droughts, as the elevated summer withdrawal comes at a time when demand upon groundwater resources is highest.

It is critical to note that the VDEQ PWS water-withdrawal data were not used to directly revise quantitative estimates of annual or monthly per-capita water-use metrics. Only the monthly temporal variation of the PWS withdrawal above or below the annual average, a dimensionless ratio, was applied to previously published annual water-use estimates to disaggregate the annual value for use in monthly estimates for water-use modeling or other purposes. Therefore, these monthly ratios could be similarly applied to disaggregate previous or future annual water-use estimates. Tabular data of the monthly ratios and estimated monthly self-supplied domestic water withdrawal for the Virginia Coastal Plain are available as a USGS data release in Kearns and Pope (2025).

4.3. Assumptions and Limitations

Because self-supplied domestic water withdrawals are rarely metered or reported, USGS estimates have typically used a coefficient-based approach, multiplying an estimate of the self-supplied domestic population by an estimate of daily per-capita water use. These coefficients are averages often derived from population estimates in areas served by public supply and a general representation of the numerous ways water is used indoors and outdoors at private residences (Dieter and others, 2018). By applying a single, average value for all self-supplied domestic water use across the Virginia Coastal Plain, the greatest uncertainty will exist at the smallest scales, at which localized conditions or factors may cause deviation from the mean. As one example, the extent of outdoor water use in Virginia Beach previously documented by Eggleston (2010) may not be sufficiently represented by Virginia’s average of 80 gal/d per capita for domestic water use estimated by Dieter and others (2018) and used in this report.

Previous research by Kenny and Juracek (2012) suggests there are many demographic and socioeconomic variables that correlate with domestic water use, including population age, housing density, housing stock age, and income level. These variables may be proxies for conditions with a direct impact on water use, including the use of water-saving appliances in the home and the presence and extent of outdoor irrigation. The complexity, variability, and availability of these demographic or socioeconomic variables makes refining self-supplied domestic water-use estimates with this information beyond the scope of this report. However, the USGS has a national effort underway to use large machine-learning models to further refine self-supplied domestic water-use estimates (U.S. Geological Survey, 2024a).