Unconfined Groundwater Storage Condition
The filtering process resulted in 356,785 monitoring wells being identified as completed in an unconfined aquifer (Extended Data Fig. 1a), for a total of 4,117,920 water table measurements (Extended Data Fig. 1b). Most wells (95%) in unconfined aquifers are out of service, although we still use their data to quantify historical temporal trends in groundwater availability and storage. The state of Nebraska has the most measurements (619,538; Extended Data Fig. 1b and 1c), followed by New York, California, Colorado, New Mexico, Wisconsin, South Dakota, and Louisiana. States with relatively few wells and measurements include Alabama, West Virginia, and Illinois. Measurements in the early decades (e.g., 1940) were taken generally from the High Plains Aquifer, Utah, and the Mississippi Alluvial Plain, with measurements extending to the northeast states, western states, and north plain states in 1960–1980. The highest measurement activity occurred during the 1980s (peak of approximately 80,000 measurements per year, from 22,463 wells; Extended Data Fig. 1d), with an average of 11,516 measured wells per year between 1920 and 2020. As of 2020, measurements were taken from only 8,183 wells.
Spatial patterns of groundwater head h (Extended Data Fig. 2) align with elevation patterns. However, the temporally averaged saturated thickness (m) for each monitoring well location, filtered for being in an unconfined aquifer (Fig. 1a) shows similar dramatic differences spatially throughout the 48 states. The average saturated thickness is 38.9 m, ranging from 0 m to 1,116 m (Nevada). 10% of locations have an average saturated thickness < 10 m; 46% of locations < 25 m; 69% of locations < 40 m; and 9% of locations > 100 m. These latter locations (black dots in Fig. 1a) are in the northern HPA (Nebraska), New Mexico, Arizona, Utah, Nevada, and California. Similar conditions occur for specific years. For the CV (Extended Data Fig. 3a), approximately one-fourth of locations in 2010 had a saturated thickness of 10 m or less, and 40% with 50 m or more. For the HPA, these fractions are 8% and 70%; and for the MRVAA, they are 7% and 70%, demonstrating, in terms of relative number of locations, the similarities between the HPA and the MRVAA regarding available saturated thickness in the aquifers.
Depth to water table dwt varies dramatically spatially (Extended Data Fig. 4), with much of the country (northwest, northern plains, northeast) experiencing depths between 1 and 10 m, whereas the HPA, CV, southern Arizona, and southern Nevada, have depths over 75 m. Regions with shallow groundwater typically are locations of groundwater-stream exchange. By comparing the map of 1960 with 2010 (Extended Data Fig. 4), the water table has become much deeper in portions of eastern Colorado, western Kansas, the panhandles of Oklahoma and Texas, and the portions of Arkansas, Mississippi, and Louisiana that reside within the MRVAA (Fig. 1c, 1d). These results coincide with locations of reported groundwater depletion due to pumping for irrigation12–15,18. For the CV, HPA, and MRVAA, the fraction of locations with dwt > 50 m is 32%, 49%, and 10% (Extended Data Fig. 3b). Besides a sign of groundwater depletion, deeper water tables also result in higher pumping costs19–21, as 1) pumps must lift water over a greater vertical distance to the ground surface, leading to higher energy costs; and 2) wells must be drilled deeper in the aquifer, leading to higher capital costs.
Rates of Change in Groundwater Head
We used the temporal data for each monitoring well location to estimate rates of change (m/yr) in groundwater head h and saturated thickness s. Specifically, we determined, for each monitoring well in the 48 states that had more than one measurement, whether h increases or decreases through the measurement history (Extended Data Fig. 5a), the number of years between the earliest and latest measurement events (Extended Data Fig. 5b), and the cumulative change in h between these events (Extended Data Fig. 5c). 51.7% of wells (44,301) had a decrease in h, and hence a decrease in groundwater volume, whereas 48.3% (41,319) had an increase in h. Decreases in h likely are due to groundwater pumping for irrigation and drinking; increases are due to increases in recharge (rainfall percolation, irrigation percolation, canal seepage, reservoir seepage). Using the values of time and head change, we calculated the change rate in h (Extended Data Fig. 5d). Most wells have a head change of less than 1 m (either increase or decrease), although many of these wells have a short sampling period of less than 6 years. The change rate in h (m/yr) for the CV, HPA, and MRVAA systems show areas of both decrease and increase, with many locations either decreasing by 5 m/yr or increasing by 5 m/yr (Extended Data Fig. 6). Areas of highest decline rates are in southwest Kansas and the panhandles of Oklahoma and Texas.
We equated a change in h with a change in saturated thickness s, then used the latter to calculate the fraction change in s for each monitored location. This fraction change provides a measure of relative change in groundwater storage at each location. Fraction changes range from 99% decrease (i.e., complete depletion of the aquifer) to 100% increase (doubling of saturated thickness) (Fig. 2a), although most areas (92%) have a fraction change of less than 25%, either as a decrease (Fig. 2a) or an increase (Fig. 2b). Decreases or increases of more than 75% occur throughout the regions of the 48 states. Of these locations (474; 0.6% of locations), 67% had an initial s of less than 20 m, 27% had an initial s of between 20 m and 50 m, and only 6% had an initial s of more than 50 m. Fraction changes for the CV, HPA, and MRVAA systems (Extended Data Fig. 7) exhibit the same pattern as the 48 states, with areas of both fraction increases and decreases. In general, areas of large fraction decreases coincide with areas of groundwater pumping (Extended Data Fig. 8) and low rainfall, as groundwater is extracted at a rate greater than replenishment from recharge.
Time to Groundwater Depletion in Unconfined Aquifers
We used the datasets of saturated thickness s (m) established in this study to estimate the number of years to depletion for any areas exhibiting historical decline in groundwater levels. We used only monitoring well locations with more than 10 years between the earliest and latest measurement time. Using historical trends in rate changes in s, we estimated years to depletion for each qualifying location by dividing the most recent value of s (m) by the calculated rate of decline in s (m/yr), assuming that the same historical rate of change will continue into the future until groundwater is depleted14,15. This may not occur, however, as irrigators and municipalities adapt to a dwindling yield from pumping wells in the vicinity. Regardless, the estimates we provide here likely coincide with the expected lifetime of appreciable groundwater use for that location.
The resulting “time to depletion” map of the 48 states (Fig. 3a) shows 3% of monitored locations with declining levels, principally in the western United States (western and southern HPA, southern Arizona, CV, Snake River Valley of Idaho), that have a depletion time of less than 20 years (Fig. 3b). 8% of locations have a depletion time < 50 years; 16% of locations < 100 years, and 51% > 500 years (Fig. 3b). These depletion terms can be categorized as “Severe” (< 20 years), indicating a short-term shortage of groundwater supply; “Intense”, indicating that current irrigators and water operators likely will have sufficient water for demand, but not the next generation (20–50 years); “Manage”, areas where groundwater must be managed wisely to sustain irrigation and drinking water practices; and “Observe”, where groundwater levels should continue to be monitored but storage is adequate for centuries. Of the CV, HPA, and MRVAA, the CV has the highest proportion of locations with severe conditions, followed by the HPA and MRVAA (Extended Fig. 3c).
Estimates of time to depletion (Fig. 3a,b) provide information regarding the serviceable lifetime of the unconfined aquifer, but not the relative volume of groundwater available. A location may have a long lifetime as indicated by historical groundwater levels, but not sufficient groundwater storage for irrigation and/or drinking water. A plot of s (m) vs. years to depletion (Fig. 3c) demonstrates that, although there most locations have a relatively thin (< 40 m) saturated thickness, there are many locations (“Long-term Production”) that have both sufficient groundwater and a long (> 150 years) lifetime for appreciable groundwater use, particularly with management of irrigation technology and crop choice.
Estimates of time to depletion (yr) were added to the most recent measurement event to calculate a depletion year for the CV, HPA, and MRVAA (Fig. 4). The depletion year of locations in Arizona is also included (Fig. 4c), due to the severe decline in s and time to depletion in the southern portion of the state (Fig. 3a). Many locations in these regions, particularly in western Nebraska, eastern Colorado, southwest Kansas, the panhandles of Oklahoma and Texas, have a depletion year before 2050. Locations of rapid depletion within these four regions generally coincide with areas of intensive groundwater pumping for irrigation, as estimated by the USGS for the year 201522 (Extended Data Fig. 8). Results for these stressed aquifers agree with region-specific studies. For example, estimated aquifer service life of 20–30 years for portions of the central and southern regions of the HPA14,15.
These estimates of time to depletion, of course, assume a future climate that is similar to historical climate in terms of rainfall and temperature. If future climate is drier in certain regions (e.g., CV, HPA), then pumping for irrigation may intensify, leading to shorter depletion times than what we present here (Fig. 3). Some areas, due to a projected decrease in annual rainfall, may require new or additional irrigation23, again leading to shorter depletion times, or even a switch from historical storage increases to future storage decreases. However, an expected increase in frequency of rainfall extremes24 may lead to more rapid and higher volumetric recharge events, leading to replenishment of unconfined aquifers in areas of intensive irrigation. These competing factors require further investigation.