NOTE: The authors believe that this dataset is likely to be LESS accurate than using a clip of the larger San Francisco Bay groundwater dataset, here: https://dash.berkeley.edu/stash/dataset/doi:10.6078/D1W01Q. The reason is that additional datapoints and editing were used to improve the San Francisco Bay groundwater map after this Alameda County map was produced. We expect that the most recent San Francisco Bay map is the most accurate, even for smaller geographic scales.

This dataset contains a comparison of four interpolation methods used to estimate a minimum depth to groundwater surface for Alameda County, within one kilometer/0.6 mi of San Francisco Bay. The interpolation is based on well data from the CA State Water Board GAMA GeoTracker database, and the depth to water was calculated using a 2m USGS Digital Elevation Model.

Well data - California State Water Control Board: GAMA GeoTracker http://geotracker.waterboards.ca.gov/data_download_by_county. Ground elevation data - U.S. Geological Survey https://topotools.cr.usgs.gov/coned/sanfrancisco.php. 2 meter DEM. SF Bay extent (includes open water and tidal wetlands) San Francisco Estuary Institute (SFEI): Bay Area Aquatic Resource Inventory (BAARI) http://www.sfei.org/baari. GIS FILE PROPERTIES: File format: ESRI Layer Package. Cellsize: 6.56. Linear unit: Feet. Z unit: Feet. Projected Coordinate System: NAD_1983_2011_StatePlane_California_III_FIPS_0403_Ft_US. Geographic Coordinate System: GCS_NAD_1983_2011.

METHODS: We subtracted the minimum depth to water at each well point from the ground elevation (extracted from the 2m DEM) to determine groundwater elevation at each well point. These elevations represent the maximum measured groundwater table height in the past 20 years. We then performed the interpolation on this groundwater elevation dataset, a total of 3,183 individual well points. Wells within one mile of the coast were included in the interpolation; results are shown within one kilometer (0.6 miles) of the coast, a distance used in previous studies of sea level rise-induced groundwater inundation. Wells within one-half mile north and south of the county borders were included in the interpolation to ensure continuity, but results are shown only for area within Alameda County.

We tested a variety of methods available in ArcGIS and used cross-validation to determine which method minimized prediction error most. We compared root mean square error (RMSE) to see how accurately each model predicted values at non-sampled locations, and examined mean error (ME), or the averaged difference between actual and predicted values, to see if each model was skewed in one direction or another. For each interpolation technique, we chose the input parameters (e.g. power, number of neighbors included) that minimized RMSE most.

After performing the groundwater table interpolation, we subtracted the output from the original elevation surface to display estimated minimum depth to water values. The interpolation and subtraction method we used produced some negative values for depth to water, indicating water above the ground surface, especially in areas where there were no well sample points at the base of a slope or in a valley. In the provided data files, we have changed these negative values to zero for clarity.

This data package contains the minimum depth to water results obtained by using each of the four interpolation methods, as well as files showing the minimum and maximum of the four methods for comparison. Also included are files showing the bay edge file used and no data areas (greater than 1km/0.6mi from the nearest well point).