Groundwater recharge estimates for Maine using a Soil-Water-Balance model—25-year average, range, and uncertainty, 1991 to 2015

Martha G. Nielsen; Stephen M. Westenbroek

doi:10.3133/sir20195125

Home/Archive/USGS

USGS · sir20195125

Groundwater recharge estimates for Maine using a Soil-Water-Balance model—25-year average, range, and uncertainty, 1991 to 2015

Martha G. Nielsen·Stephen M. Westenbroek

View original source ↗Read source document ↗Export metadata ↓

Abstract

To address the lack of information on the spatial and temporal variability of recharge to groundwater systems in Maine, a study was initiated in cooperation with the Maine Geological Survey to use the U.S. Geological Survey Soil-Water-Balance model to evaluate annual average potential recharge across the State over a 25-year period from 1991 to 2015. The Maine Soil-Water-Balance model was calibrated using annual observations of recharge, runoff, and evapotranspiration for 32 calibration watersheds in the State during 2001–12 (902 total observations). Observations of recharge, runoff, and evapotranspiration were developed for each watershed to reduce the possibility of nonunique combinations of model parameters during the calibration. The Maine Soil-Water-Balance model was run using an optional evapotranspiration calculation method that provides more control for calibration than the standard method. The model was calibrated using the Parameter ESTimation software suite. The overall mean model error (average of all annual residuals for recharge, runoff, and precipitation) was 0.39 inch. The mean of the absolute value of the residuals, or the mean absolute error, was 2.32 inches. The root mean squared error for the calibrated model overall was 3.14 inches. Statistical tests indicated that the model residuals are normally distributed. To determine the potential uncertainty in the median annual potential recharge that results from uncertainty in the parameters as they relate to information contained in the observations, 300 alternate model realizations were run, and the standard deviation of the median potential recharge value at every pixel was calculated. Simulated 25-year median potential recharge across the State is widely variable; this variability closely follows patterns of precipitation, with additional variability contributed by the patchwork nature of the combinations of land-use class and hydrologic soil group inputs, and distribution of available water capacity in the soil across the State. Overall, the 25-year median annual potential recharge across the State is 7.5 inches, ranging from a low of about 5 inches to over 30 inches. The statewide range in the 25-year minimum values is from just over 2 inches to just over 20 inches. The statewide range in the 25-year maximum potential recharge is between 15 and 48 inches per year. The model areas with the highest simulated median potential recharge include areas underlain by type A soils (sandy and well drained), particularly those that also have land uses with low or little vegetation (blueberry barrens, developed, open space, scrub/shrub, and cropland, for example). The potential recharge values for these areas are similar to previously published values for comparable soil types. The 25-year average potential recharge grids were compared to recharge evaluated through groundwater-flow models or other methods in four hydrogeologic settings at six study areas in the State. A key factor in the ability of the Soil-Water-Balance model to reproduce the earlier study results was whether the available water-capacity data were an appropriate match for the hydrologic soil groups. The Maine Soil-Water-Balance model does a good job in representing an accurate potential recharge under circumstances where the surficial mapped soils extend below the surface to the water-table aquifer and where the available water-capacity data are in an appropriate range for the hydrologic soil group. One hydrogeologic setting that was challenging for the model was where a silt and clay layer was below a shallow soil unit that did not have available water-capacity data that were appropriate for the hydrologic soil group. In these cases, typically the available water-capacity data were very low, not accounting for the impedance of water flow provided by the underlying soil. The model also does not simulate well areas where bedrock surfaces are above the water table but below the plant rooting zone. The data products accompanying this report are intended to be used to provide first-cut estimates of recharge for geographic areas no smaller than the smallest watersheds used in the calibration of the model—or about 1.5 square miles. It is recommended that the grids are used to calculate an area-wide average potential recharge for any given area of study, and an uncertainty around the mean should be calculated from the standard deviation grid at the same time.

Explore related subjects

Scientific Investigations Report Report

Source-reported bounding extent: 43.059825° to 47.459686° latitude; -71.084334° to -66.950569° longitude. This indicates report coverage, not an exact sampling location. View area on OpenStreetMap.

Keep this discovery

Explore connections, maps & timelines

BibTeX RIS

Martha G. Nielsen, Stephen M. Westenbroek. 2019-12-30. Groundwater recharge estimates for Maine using a Soil-Water-Balance model—25-year average, range, and uncertainty, 1991 to 2015. https://doi.org/10.3133/sir20195125

Cite the original work for its findings. Save a collection to share your selection of sources.

KEEP EXPLORING

Related USGS reports

USGS2026

Evaluating the number of realizations (train-test splits) necessary to ensure robust estimates of conventional hydrothermal resource favorability using Monte Carlo cross-validation, the Great Basin, Western USA

Supervised machine learning approaches used to produce favorability maps in resource assessments commonly partition an input dataset into training and testing subsets (that is, through train-test splits). However, if a single split is not representative of underlying resource conditions, the resulting favorability map may be biased, particularly with sparse or heterogeneous datasets. To mitigate the potential for unrepresentative favorability maps, recent works by U.S. Geological Survey researchers have used Monte Carlo cross-validation to implement a large number ( n) of randomized train-test splits to produce n realizations. The realizations are then aggregated using cell-wise median predictions to produce the resulting final favorability map, recognizing that the median will remove extreme values that might have been estimated from a comparatively few unrepresentative realizations. Theoretically, the median prediction at every cell converges to a single estimate as n approaches infinity, and the question addressed herein is: How many realizations are necessary to get a robust estimate (that is, an estimate unaffected by additional realizations)? Therefore, convergence metrics are described to identify the number of train-test splits and realizations needed to make a reliable map using the data and supervised machine learning workflow for conventional hydrothermal resource assessment of the Great Basin of the Western United States. For this study, robust prediction occurs with n approximately equal to 600 realizations, and smaller ensembles (for example, n is approximately equal to 200 or even 100) may be suitable during preliminary or exploratory analyses. Although the resulting threshold numbers will vary depending on the data and modeling approach used, the convergence metrics proposed and applied herein are applicable to other supervised machine learning studies that use Monte Carlo cross-validation.

Scientific Investigations Report↗

USGS2026

Groundwater budget for the Portneuf River Basin, southeastern Idaho

The U.S. Geological Survey, in cooperation with the Idaho Department of Water Resources, developed the first groundwater budget for the upper Portneuf River Basin in southeastern Idaho, a tributary of the Snake River. A groundwater budget for the Portneuf River Basin supports water resource and administrative management decisions within the basin and for downstream users of the eastern Snake River Plain aquifer. The analysis quantifies inflows, outflows, and residuals for an aquifer system defined by the Idaho Department of Water Resources under three recent precipitation conditions: an average year (2016), a dry year (2021), and a wet year (2023). The budget was calculated for the basin and for two subbasins delineated by the U.S. Geological Survey streamgage Portneuf River at Topaz, Idaho (13073000): the above Topaz subbasin, which includes the Portneuf-Gem valley, and the below Topaz subbasin, which includes Marsh Valley. Aquifer inflows include mountain-front recharge; areal recharge; recharge from applied water, including recharge from irrigation, canal seepage, and effluent returns; groundwater flow into the basin; and reservoir seepage. Outflows include groundwater pumping for agricultural, municipal, and domestic uses; subsurface outflow through the Portneuf gap; and discharge to streams as base flow. Estimated annual inflows for the full basin ranged from 116,880 acre-feet in 2021 to 154,880 acre-feet in 2023, and outflows ranged from 132,190 acre-feet in 2016 to 158,490 acre-feet in 2023. Basinwide residuals, the change in storage plus uncertainty, were −2,410 acre-feet in 2016, −18,410 acre-feet in 2021, and −3,610 acre-feet in 2023, and each subbasin had a mixture of negative and positive residuals. These basinwide results, interpreted as change in storage plus water-budget uncertainty, indicate minor net losses that increase with dry conditions, consistent with well depth-to-water measurements indicating gradual multi-decadal declines. Patterns in subbasin residuals are driven by irrigation practices and surface-water management; for example, the above Topaz subbasin relies more on groundwater for irrigation, resulting in more negative residuals during dry years. In contrast, the below Topaz subbasin depends more on surface water, producing positive residuals during dry years because of irrigation losses and canal seepage. The analyses for both subbasins indicate decades-long groundwater declines, which agree with the overall negative residuals calculated. The largest sources of uncertainty include climate inputs (precipitation and evapotranspiration) and poorly constrained components such as groundwater inflow and tributary stream infiltration. These findings provide insight into how climatic variables and irrigation practices affect groundwater inflows and outflows within the Portneuf River Basin.

Idaho↗

USGS2026

World minerals outlook—Antimony, copper, graphite, niobium, potash, silicon, silver, and tungsten through 2030

Demand for mineral commodities continues to increase owing to rapid technological advancements and global economic growth. To support this increase in demand, information about current and near-term supplies of these mineral commodities is essential. Such data help evaluate the health and stability of global supply chains and protect against potential supply risks. Current production and production capacity data coupled with information regarding near-term expected changes (such as facility closures, planned expansions, or the completion of new mineral production projects) can provide insight into expected future market developments and expected available supply of mineral commodities. This report within the “World Minerals Outlook” series analyzes data for antimony, copper, natural graphite, niobium, potash, silicon metal, silver, and tungsten—eight mineral commodities listed in the 2025 List of Critical Minerals. World production data from 2019 through 2025, as well as estimated world capacity data from 2023 through 2025 were analyzed. Using these data, near-term (2026–2030) expected world capacities were estimated. For antimony, copper, niobium, silicon metal, silver, and tungsten, modest increases in production capacity of 4 to 10 percent were expected between 2025 and 2030, whereas capacities for natural graphite and potash were expected to increase by about 18 to 20 percent.

Scientific Investigations Report↗