Study Methods

The method of analysis used to evaluate the configuration (shape or form) of the current (2009) groundwater surface in the CPRAS units relied on water-level data from wells measured in spring 2009. These data were used with statistical methods to determine the presence of spatial correlation of groundwater levels with distance and direction from each other. This information was used to interpolate the groundwater elevations between observations for each hydrogeologic unit. The change in groundwater levels between 1984 and 2009 was determined by computing the change in water-level elevations from wells with observations in 1984 and 2009, and analyzing these changes for spatial patterns to aid in interpolation.

Sources of Data

The spring 2009 groundwater-levels for the CPRAS were synoptic (collected simultaneously) sets of measurements of the depth to water in wells representative of a single hydrogeologic unit as part of routine monitoring. In addition to the water levels measured by the USGS, data were provided by the U.S. Department of Energy, Oregon Water Resources Department, Confederated Tribes of the Umatilla Indian Reservation, Washington Department of Ecology, South Columbia Basin Irrigation District, Walla Walla Basin Watershed Council, Palouse Basin Aquifer Committee, East Columbia Basin Irrigation District, and Bureau of Reclamation. These agencies provided well information and water-level data collected during late winter and early spring 2009, prior to the typical start of significant pumping for agriculture. The spatial distribution of the wells measured in the existing monitoring networks was analyzed and determined to be insufficient for understanding groundwater levels in large areas of each of the hydrogeologic units. Therefore, the monitoring network data were supplemented with measurements from wells specifically selected for this study. Wells were selected from those measured during the 1984 USGS synoptic water-level investigation (Vaccaro, 1999) on the basis of the primary hydrogeologic unit to which the wells were open (a screened or open part of the well through which water enters—referred to as the open interval) and their proximity to areas of sparse data for the corresponding hydrogeologic unit. Priority was given to wells with water-level measurements from spring 1984 to facilitate the comparison of water levels between 1984 and 2009. Sixty‑six of these supplemental wells that fit these selection criteria were identified and located, and water levels were measured in early spring 2009 by the USGS (pls. 2, 3, 4, and 5; appendix A). The number of wells measured during spring 2009 from all sources totaled 1,752 and were apportioned between the hydrogeologic units as follows: Overburden, 984 wells; Saddle Mountains, 118 wells; Wanapum, 288 wells; and Grande Ronde, 362 wells.

The water-level change analysis evaluated water-level data from spring 1984 obtained from the USGS and eight other agencies. Of the 1,752 wells measured in spring 2009, corresponding water levels had been measured in 470 wells in spring 1984. The 1984 water-level measurements were apportioned between the hydrogeologic units as follows: Overburden, 197 wells; Saddle Mountains, 50 wells; Wanapum, 87 wells; and Grande Ronde, 136 wells (pls. 6, 7, 8, and 9; appendix A).

Information on land-surface elevation was needed to determine groundwater elevation from depth-to-water information. Land-surface elevations for many wells were provided by the agencies that measured the well. For the remaining wells, land-surface elevations were estimated as the median land-surface elevation using a 10-m lateral resolution Digital Elevation Model (DEM) (U.S. Geological Survey, 1999) within a 100-ft radius of the reported location of the well to account for uncertainty with regard to the exact location of wells used in the study (appendix A).

The hydrogeologic units to which the wells were open were determined on the basis of assignments made by the agencies that provided the well data. For wells with no assigned hydrogeologic unit, open-interval information was compared with estimated unit-top elevations from a three‑dimensional hydrogeologic framework model (E.R. Burns, U.S. Geological Survey, written commun., 2009) and was assigned to a specific unit if that unit comprised 75 percent or greater of the open interval (appendix A). If no units comprised 75 percent or greater of the open interval (multi-unit) or if information was insufficient to determine the hydrogeologic unit (unit unknown) then the water level from the well was excluded from all analyses.

Generalized Groundwater-Elevation Maps

Generalized groundwater-elevation maps for the three basalt hydrogeologic units were developed from the 2009 water-level data using a three-stage geostatistical process (Bossong and others, 1999, p. 25–29; Davis, 2002, p. 258–259). A similar map for the Overburden unit was not developed because the 2009 data density was insufficient to develop reliable surfaces of groundwater elevation for the isolated and discontinuous areas over which thick overburden deposits generally occur. The first stage of the process involved estimating the regional groundwater-elevation trend for each unit for the entire CPRAS extent using a regression‑smoothing technique called locally weighted scatterplot smoothing (LOESS). The groundwater elevation estimated using the regional trend surface at the location of each well was subtracted from the groundwater elevation measured at the well. The surface represented by the residuals (remainders) is the configuration of the groundwater surface without the regional influence. The second stage consisted of variography (analysis of the correlation of measurements with distance and direction) using the groundwater-elevation residuals within defined structural regions for each hydrogeologic unit to determine trends in the groundwater surface due to the hydrogeologic conditions within the regions. The third stage was the interpolation of the groundwater-elevation residuals for each hydrogeologic unit using kriging (a statistical method of data analysis and interpolation), with the required kriging parameters determined by the variography analysis.

The advantage of kriging, variography, and regional trend removal is that smaller scale structure of the data is revealed and preserved in areas where data are sufficiently dense, whereas the regional trend provides an indication of the general tendency of groundwater elevation and flow direction in areas where data are sparse or nonexistent. Because the 2009 water-level data were relatively sparse in many parts of the CPRAS, only the generalized configurations of the groundwater surfaces for the basalt units could be reliably determined for the entire study area. These generalized water‑level configurations are shown on plates 3–5 using color shading rather than elevation contours to indicate that the maps are simplifications of the large and complex actual flow system. Generalized directions of groundwater flow derived from this analysis and depicted on maps were modified on the basis of detailed maps of regional and subregional groundwater levels that better illustrate groundwater flow directions that are available for many parts of the CPRAS (Whiteman, 1986; Whiteman and others, 1994; Vaccaro, 1999; and Vaccaro and others, 2009).

Because the CPRAS includes several distinctive structural regions (pl. 1) with markedly different densities of folds and faults that may be impediments to groundwater flow, an analysis was done to determine whether the spatial correlation between groundwater levels differed within these structural units for each hydrogeologic unit. This process included variography, which provides statistical measures of the degree of correlation between a groundwater measurement and other groundwater measurements based on the distance and direction that separates them. Typically, wells that are closer together in the same aquifer have similar water levels and respond similarly to stresses. As distance between wells increases, however, the similarity of measured values generally decreases. On the other hand, if a flow barrier exists between two closely spaced wells due to folds, faults, or changes in lithology that affect the hydraulic properties of the aquifers, then the hydraulic response to a given stress may be different in each well. The presence of a large number of flow barriers will shorten the correlation distance, and the absence of barriers will result in a larger average distance of correlation. Variography provides an objective way of evaluating the average distance over which groundwater-elevation measurements may be used to make an estimate of groundwater elevation at another, non-measured location, although the results are heavily dependent on having data available throughout the defined structural region.

The CPRAS was divided by geographic or structural regions on the basis of significantly different variography to improve the estimates of groundwater elevation developed using the kriging analysis. The structural regions used for analysis during this study were modified from Reidel and others (2002) and initially included the Yakima Fold Belt, Palouse Slope, Blue Mountains, and Clearwater Embayment. The extent of these structural regions (pl. 1) was modified on the basis of recent modeling of the extent and thickness of the hydrogeologic units and the geologic structures that form geologic compartments that may influence groundwater levels (E.R. Burns, U.S. Geological Survey, written commun., 2009). The Blue Mountains and Clearwater Embayment structural regions were subsequently eliminated from the analysis because few of the spring 2009 groundwater-level measurements were within these regions. The arc-shaped Yakima Fold Belt is generally composed of long east-west trending valleys separated by fold-fault pairs defining the valley ridgetops (E.R. Burns, U.S. Geological Survey, written commun., 2009). Compared to the Yakima Fold Belt, the Palouse Slope is less deformed, with relatively few faults and less intense folding. On the basis of the significant difference in variography between the Palouse Slope and Yakima Fold Belt, whenever data from a specific hydrogeologic unit were sufficiently dense in each of these structural regions, water‑level data within the regions were kriged separately.

Kriging was used for interpolation of the groundwater‑level residuals using a lateral grid size of 1,000 ft. Kriging is based on a geostatistical theory predicated on the observation that values of spatially distributed data commonly are correlated—values at nearby locations are more highly correlated than values at distant locations (Alley, 1993, p. 87). Parameters for the kriging analysis were selected on the basis of the variography analysis and used to interpolate groundwater levels for each hydrogeologic unit by structural region. The regional trend, estimated using LOESS, and the kriged residuals, representing the configuration of the groundwater surface without the regional influence, were subsequently added together to produce the generalized groundwater-elevation maps for each hydrogeologic unit.

The estimated groundwater-elevation surfaces for each basalt hydrogeologic unit were compared to land‑surface elevation to identify areas where artesian conditions (potential groundwater elevations greater than land-surface elevation) were indicated. Artesian conditions are common in selected areas within the CPRAS; however, the interpolated groundwater surface indicated that artesian conditions were in some areas where no data are available to support their presence, or that the magnitudes of the predicted artesian levels were deemed unreasonable. A small number of such areas occur along the edges of each hydrogeologic unit, where few data exist to constrain the interpolation or in topographical depressions or incised canyons where land-surface elevation decreases rapidly and no data are available for areas within the depression or along the canyon to guide the interpolation. In these situations, the interpolated groundwater elevations were modified such that the mapped groundwater elevations would not exceed the maximum artesian levels observed within each unit. The maximum artesian levels are 50, 225, and 300 ft above land-surface for the Saddle Mountains, Wanapum, and Grande Ronde, respectively.

The estimated groundwater-elevation surfaces for each basalt hydrogeologic unit also were compared to surfaces representing the estimated unit-top elevations from a three-dimensional hydrogeologic framework model (E.R. Burns, U.S. Geological Survey, written commun., 2009) to identify potentially unsaturated areas. On the basis of these comparisons, several areas within the Saddle Mountains and Wanapum Units may be unsaturated. These areas generally are associated with higher land-surface elevations where depths to water are greater and unit thicknesses are thinner. Because of the uncertainty associated with the estimates of the groundwater and unit-top elevations, these areas are retained on the groundwater-elevation maps.

Development of Groundwater-Level Change Maps

Groundwater levels measured during spring 1984 and spring 2009 were compared for 470 wells to evaluate water‑level changes for each of the hydrogeologic units and to construct groundwater-level change maps. The annualized rates of water-level change for the 25-year period between measurements were calculated for each well. This average rate of change based on only two measurements represents a long-term net change in water levels; the sign and rate of change in any particular year could be different and may not be representative of recent trends in water-level changes. The groundwater-level changes presented are for wells at specific locations and may not be representative of conditions throughout the CPRAS.