
Sourcing & Methodology
To report this story, The Times built a dataset of groundwater levels from the U.S. Geological Survey’s National Water Information System, the U.S.G.S. National Groundwater Monitoring Network and 28 states and regional authorities: Arizona Department of Water Resources, California Department of Water Resources, Colorado Division of Water Resources, Delaware Geological Survey, Southwest Florida Water Management District, St. Johns River Water Management District, South Florida Water Management District, Suwannee River Water Management District, Hawaii Department of Land and Natural Resources, Iowa Geological Survey, Idaho Department of Water Resources, Illinois State Water Survey, Kansas Geological Survey, Kentucky Geological Survey, Minnesota Department of Natural Resources, Mississippi Department of Environmental Quality, Montana Bureau of Mines and Geology, North Dakota Department of Water Resources, University of Nebraska Conservation and Survey Division, New Mexico Bureau of Geology and Mineral Resources, Nevada Division of Water Resources, Oklahoma Water Resources Board, Oregon Water Resources Department, South Dakota Department of Agriculture and Natural Resources, Texas Water Development Board, Utah Geological Survey, Washington State Department of Ecology and Wyoming State Engineer’s Office.
Data from each agency may include water level measurements made by staff members as well as additional data submitted by contributing agencies, researchers or private firms. State agencies say they perform quality checks, but those checks don’t guarantee complete accuracy. Data were limited to reflect stable groundwater measurement conditions as closely as possible. Duplicate sites that appeared in the U.S.G.S. and regional data sets were removed based on matching site numbers where available, and by matching latitude, longitude and well-depth combinations.
Based on a survey of scientific papers and consultations with researchers, The Times used Theil-Sen median regressions and the Mann-Kendall test for significant trends to estimate increasing and decreasing patterns at each site. The Mann-Kendall test can overestimate significance in cases with fewer observations, and underestimate in cases with more observations. Trend analysis used annually averaged data. Sites were limited to those with a minimum of 10 years of observations, and at least one observation within the first and last 5 years. Theil-sen trends represented in the animated map were calculated on 20-year ranges, every year, from the end years of 1940 to 2022. Each site had a minimum of 10 years with observations, and at least one observation within the first and last three years.
For the analysis of record lows, annual averages based on January-to-May measurements were used to calculate record highs and lows in order to mitigate the effects of large swings during pumping seasons. Only sites that had a minimum of five years with observations within the past decade, and 15 years with observations before the past decade, were included.
United States aquifer map data from GebreEgziabher, Jasechko and Perrone, Nature Communications (2022)



