Integrating long-term trends in groundwater depth and low streamflow across the United States using regional classifications

The U.S. Geological Survey (USGS) Water Resources Mission Area (WMA) is working to address a need to understand where the Nation is experiencing water shortages or surpluses relative to the demand for water need by delivering routine assessments of water supply and demand and an understanding of the natural and human factors affecting the balance between supply and demand. A key part of the Integrated Water Availability Assessments (IWAAs) Trends and Drivers project is identifying long-term national trends in water availability, including groundwater and surface water quantity, quality, and use. This data release contains analyses of previously published Mann-Kendall monotonic trends data (see URLs below) and new Regional Kendall trend analyses for streamflow (annual minimum mean daily discharge (cfs) over a consecutive 7-day period) and groundwater (annual mean water-level depth). A total of 4,992 streamflow sites had 7-day low flow trend results in at least one of the three trend periods, whereas there were 19,304 groundwater sites with groundwater depth trend results for these periods. Five regional classifications, that represent unique characteristics of the surface and/or subsurface as contiguous and non-contiguous regions, were used to calculate Regional Kendall trends. These include Principal/Secondary Aquifers, Hydrologic Landscape Regions, Modified Soller Surficial Geology, Van Metre Regions, and Modified HUC04s.

The Regional Kendall test (Helsel and Frans, 2006) is nonparametric and calculates consistent regional trends across an area with multiple sampling points, where individual Mann-Kendall trend tests (Mann, 1945) are performed at each sampling location within the region. Test statistics for each location are summed to form the Regional Kendall test statistic. The characteristics and mathematics of the test are identical to the Seasonal Kendall test of Hirsch et al. (1982). A modification for serial correlation in the Seasonal Kendall test can be performed (Hirsch and Slack, 1984) and is recommended for studies where each location has 10 or more observations. This correction adjusts the variance of the S-statistic by estimating the covariance between blocks (i.e., seasons) and was applied to spatial autocorrelation, where the blocks are now locations, for the Regional Kendall test by Helsel and Frans (2006). For this data release, we used the rkt package (version 1.7) in R to perform Regional Kendall tests, with a minimum of three sites per region. Regional Kendall results should be interpreted to represent the consistent trend among sites within a region. A correction for correlation between sites (i.e., blocks) within a given region was performed as part of the rkt package. We also performed an iterative pre-whitening (IPW) procedure (Zhang and Zwiers, 2004) to the annual low flow and GW depth data prior to calculating Regional Kendall trends. This procedure only pre-whitens data with a lag-1 autocorrelation > 0.05 by first estimating the trend slope, then removing the trend from the original data, estimating a new lag-1 autocorrelation, and repeating until the autocorrelation is ≤ 0.05 or the changes between two iterations are small.Trend magnitudes were computed for 4,992 qualifying streamflow and 19,304 qualifying groundwater monitoring locations. Trend analyses were computed between years 1980-2020 and trend periods are between 21-41 years long. Trends at each site are available for three main periods: (i) 1980-2020, (ii) 1990-2010, and (iii) 2000-2020.

Caution must be exercised when utilizing monotonic trend analyses conducted over periods of up to several decades (and in some places longer ones) due to the potential for confounding deterministic gradual trends with multi-decadal climatic fluctuations. Abrupt trend changes, such as changes to water withdrawals and wastewater return flows, or episodic disturbances with multi-year recovery periods, such as wildfires, are also not evaluated for any site. Sites with pronounced abrupt changes or other non-monotonic trajectories of change may require more sophisticated trend analyses than those presented in this data release.

Streamflow trends source: https://doi.org/10.5066/P9VBR38I
Groundwater trends source: https://doi.org/10.5066/P9ZACZ6H

REFERENCES
Helsel, D.R. and L.M. Frans (2006). Regional Kendall test for trend, Environ. Sci. Technol. 40(13), 4066-4073. https://doi.org/10.1021/es051650b.

Hirsch, R.M., J.R. Slack, and R.A. Smith (1982). Techniques of trend analysis for monthly water quality data, Water Resour. Res. 18(1), 107-121. https://doi.org/10.1029/WR018i001p00107.

Hirsch, R.M. and J.R. Slack (1984). A nonparametric trend test for seasonal data with serial dependence, Water Resour. Res. 20(6) 727-732. https://doi.org/10.1029/WR020i006p00727.

Mann, H.B. (1945). Non-parametric tests against trend, Econometrica 13(3) 245-259. https://doi.org/10.2307/1907187.

Zhang X. and F.W. Zwiers (2004). Comment on "Applicability of prewhitening to eliminate the influence of serial correlation on the Mann-Kendall test" by Sheng Yue and Chun Yuan Wang, Water Resour. Res. 40, W03805. https://doi.org/10.1029/2003WR002073.

First release: 2026-07-06 (ver. 1.0)