Groundwater Status and Trends


Trend analyses were done on the compiled groundwater-level data for the CRBG aquifers to characterize and document changes in the hydrologic status of the system. Water levels in individual wells vary in response to natural and anthropogenic stresses on daily, seasonal, decadal, and longer time scales. Because the purpose of the USGS Groundwater Resources Program study is to evaluate the long-term availability and sustainability of groundwater resources within the CPRAS, the focus of the data analysis is to examine persistent decadal or longer trends that have occurred in many wells since the onset of widespread irrigation and groundwater pumping.


For the trend analysis, groundwater measurements representing conditions within the CRBG aquifers were used. The geologic model of Burns and others (2011) was used to identify wells with bottom elevations between the simulated top of the CRBG geologic units and the older pre-Miocene basement rocks. Water-level measurements from these identified wells then were used in the analyses. To remove daily and seasonal variation in groundwater levels induced by irrigation pumping, the median groundwater level measured in winter between January and March of each year was used in the analysis. The resulting data reflect the influence of multi-year precipitation patterns and the cumulative effects of pumping and irrigation recharge.


Water-level measurements made prior to 1951 are assumed to represent pre-development conditions because large scale irrigation started after the beginning of 1951. Because summer water levels in wells were not yet affected by large scale groundwater pumping, annual median water levels were used when winter water-level data were not available prior to 1951.


Water-level measurements in reports filed by well drillers at the time of new well installation were included to obtain the broadest spatial and temporal distribution possible in areas or periods of sparse data. Groundwater-level measurements reported as non-static measurements were excluded from analysis. The subsampled dataset includes 7,735 CRBG wells, with data representing pre-development conditions in 1,265 wells (fig. 7).


Although pumping and irrigation effects began to have appreciable effects on groundwater levels during the 1950s in the CPRAS, most persistent regional declines from pumping started after 1970. A map of the linear trends in groundwater levels for 1968–2009 was constructed to show areas of widespread declines in CRBG aquifers (fig. 8). Groundwater‑level trends were computed as the slope of the best-fit line to the winter median water-level data for each well, provided that at least four data points were available that spanned at least 50 percent of the period 1968–2009. Of the 761 wells in the CRBG aquifers with sufficient data to compute water-level trends, overall declines were measured in 72 percent of the wells. The mean of the slopes of the water-level trends for all wells was a decline of 1.9 ft/yr. Rates of declines greater than 1.0 ft/yr were measured in 50 percent of wells, declines greater than 2.0 ft/yr in 38 percent of wells, declines greater than 4.0 ft/yr in 29 percent of wells, declines greater than 6.0 ft/yr in 9 percent of wells, and rates of decline greater than 8.0 ft/yr in 4 percent of wells. These results are similar to the values obtained by Snyder and Haynes (2010, table 1, p. 8) for 273 CRBG aquifer wells in the CPRAS during 1984–2009 where water levels declined in 81 percent of wells and water levels changed at an average rate of a 1.5 ft/yr decline (calculated as a weighted-mean average for all CRBG aquifer wells). Because complex anomalous behavior can occur for any single well hydrograph (for example, step changes in water level associated with nearby well construction activities), the linear water-level trend for any single well may not represent conditions over an area of interest, but the general pattern of water-level declines in multiple wells illustrates the persistent patterns across 
the CPRAS.


The clusters of wells with linear declines generally correspond to areas that are the subject of previous hydrogeologic studies and to continued data collection efforts by local, State, Federal, or non-governmental agencies in Oregon in the Mosier Watershed and Umatilla Basin, and in Washington in the Yakima Basin and the Palouse Slope/eastern Yakima Fold Belt in the Columbia Basin Ground Water Management Area (GWMA). The Columbia Basin GWMA, which encompasses Adams, Franklin, Grant, and Lincoln Counties (fig. 8), was designated by the Washington State Department of Ecology (WADOE), for the protection of groundwater in the area. Detailed analyses of well hydrographs for the Mosier Watershed, Oregon, and Yakima Basin, Washington, are available in Burns and others (2012) and Keys and others (2008), respectively. Well hydrographs are examined for parts of the Umatilla Basin, Oregon, and the Palouse Slope/eastern Yakima Fold Belt (within the GWMA), Washington, in this report. Groundwater-level hydrographs for the Umatilla Basin show that barriers to groundwater flow are readily identifiable. Conversely, groundwater-level hydrographs from the Columbia Basin GWMA exhibit a significantly different behavior, with fewer well-defined barriers to flow and with groundwater-level changes being dominated by the large-scale irrigation projects in the lowland near the Columbia River. An analysis describing the relation between well hydrographs and geologic features in the GWMA is presented by Porcello and others (2009); therefore, the discussion here is restricted to a complementary discussion of regional-scale flow patterns and hydraulic changes resulting from development of water resources in the GWMA since 1950.


Umatilla Area, Oregon


The Oregon Water Resources Department (OWRD) established several administrative areas (Critical Ground Water Areas and Ground Water Limited Areas) in the area of the Umatilla Basin that cover most locations with large long-term groundwater declines. Administrative areas include Butter Creek, Ella Butte, Ordnance Basalt, and Stage Gulch (fig. 9; Oregon Water Resources Department, 2012). A total of 286 wells within and immediately adjacent to the OWRD administrative areas in the CRBG aquifers were divided into clustered groups of wells with similar water levels and trends (fig. 10). Figure 11 presents hydrographs showing the winter median water levels for individual wells within each well group shown in figure 10 and a trend line representing the overall water-level trend of all wells in the group constructed using the LOESS algorithm in the statistical software program TIBCO Spotfire S+ (TIBCO Software, Inc., 1988–2008). Zones of low permeability may separate the groups of wells with similar groundwater levels and trends. These zones represent leaky barriers to groundwater flow and compartmentalize the CRBG aquifer system. The degree of compartmentalization is variable, but it occurs in both the vertical and horizontal directions in the Umatilla area.


Horizontal flow barriers (barriers prohibiting or reducing horizontal groundwater flow) were identified between adjacent well groups when water levels and trends were different for wells open to the same aquifer. Because aquifers have not been mapped extensively within the study area, drilling of each well was presumed stopped when the target aquifer was found. The elevation where drilling stopped was used as an estimate of the elevation of an aquifer at that location. However, the originally flat-lying CRBG lava flows have been deformed over geologic time, so the CRBG interflow zones hosting most productive aquifers in the CPRAS are not horizontal. To correct for the departure from horizontality, the geologic model of Burns and others (2011) was used to compute estimates of the stratigraphic positions for water producing zones in each well. Correction was accomplished by subtracting the estimated top elevation of the Grande Ronde Basalt geologic model unit from each well bottom elevation. Positive values indicate aquifers are in the lavas above the top of the Grande Ronde Basalt, and negative values indicate aquifers are in the Grande Ronde Basalt. Plotting the groundwater level against this stratigraphic position for each well allows a rapid assessment of whether or not hydraulic head values are representative of the same aquifers (fig. 12). However, the flow margin of several younger (post Grande Ronde Basalt) lava flows are within the area of interest with thicker deposits to the north. The resulting wedge shape of the overlying lavas complicates the interpretation of positive elevations (fig. 12) corresponding to aquifer horizons, especially for north-south transects. Additional details regarding this method are provided in appendix A. 


An example of strong horizontal compartmentalization is provided by comparing groups 2 (dark blue triangles) and 12 (gray triangles) (fig. 10). Measurements for these two groups likely are from the same aquifer within the Grande Ronde Basalt although water levels differ on average by about 500 ft (figs. 11B and 12D-F). Within each group, water levels for nearby wells commonly are within a few tens of feet of each other. Group 2 wells have a wider range of values than group 12 wells, which corresponds to a relatively smooth hydraulic gradient from Pendleton, Oregon, to the center of the Stage Gulch administrative area.


An example of a vertical flow barrier is provided by groups 3 (dark green squares) and 14 (light blue squares) (fig. 10). In map view, these groups apparently occupy much of the same area; therefore, horizontal compartmentalization does not explain the hydraulic head contrast between the groups that began in the 1960s (fig. 11A). However, based on the shallow and deep well pairs shown in figure 10, group 3 well bottoms are consistently above group 14 well bottoms (fig. 13), indicating that the aquifers are separated vertically, with the shallow group 3 wells exhibiting no persistent water-level declines, and the deeper group 14 wells exhibiting substantial declines since the 1960s. Whereas small groups of wells show ideal vertical separation (fig. 13), the pattern is obscured when considering all wells in groups 3 and 14 at the same time (fig. 12A-F). Well bottoms for group 3 wells trend to lower elevations of stratigraphic position to the south (not shown), indicating that the group 3 aquifer(s) are sloping relative to the estimated top of Grande Ronde Basalt consistent with the thickening of the younger lavas to the north. If aquifers were mapped based on similarity in hydraulic head, the apparent overlap of groups 3 and 14 (fig. 12) would disappear and the ideal vertical separation (compare fig. 13 with appendix A, fig. A1A) would be more apparent.


Barriers to groundwater flow result from primary characteristics of the basalt, the depositional environment, and post-depositional folding and faulting. Vertical flow barriers typically result when aquifers are separated by dense impermeable CRBG lava flow interiors that are laterally extensive. Near flow margins, however, the dense interiors are discontinuous, and aquifers may be well-connected hydraulically. In the OWRD administrative areas, vertical separation is apparent only in the groundwater-level data near groups 6 (light blue triangles), 11 (pink triangles), and 14 (light blue squares), which have higher hydraulic heads than nearby wells completed in lower stratigraphic units. For example, the head in group 3 wells stayed nearly constant at about 500 ft, whereas the deeper group 14 wells have shown declines in groundwater elevation from about 450 to about 300 ft since the 1960s (fig. 11B).


For all other well groups, there is a conspicuous lack of vertical hydraulic gradients across thick sequences of lava that should contain many individual lava flows. Within most groups a uniform hydraulic head exists across thousands of feet of basalt thickness despite the fact that many (approximately 100-ft thick) lavas are intersected (fig. 12A-F; ideal behavior is shown in fig. A1D). These hydraulic heads are uniform even though the older lava flows were more extensive than the younger flows, which implies that dense flow interiors should separate the aquifers, creating the conditions necessary for vertical hydraulic gradients. The uniformity of hydraulic heads in these aquifers may be the result of hydraulic equilibration through commingling wells. For confined or other low storage aquifers, time for equilibration can be short. In the Mosier Watershed, Oregon, commingled wells equilibrated within 2 years (Burns and others, 2012). Because groundwater-level monitoring typically begins after multi-year declines have been documented, data representing pre-commingling conditions are limited.


Examination of the mapped geologic structure shows horizontal compartmentalization is frequently correlated to the structure, although not always. The correlation along the Willow Creek monocline in groups 4 (orange circles), 5 (dark red circles), 10 (purple circles), and 15 (dark blue circles) show consistent head patterns along the sinuous fold structure and high head differences between the groups across the fold indicating that the fold functions as a longitudinal conduit and a perpendicular barrier to groundwater flow (figs. 10 and 11A). The Service anticline also is an apparent horizontal flow barrier, although the deeper aquifers have complex, yet similar patterns particularly later in time, which indicates that some aquifers may be better hydraulically connected across the structure than others, indicating that commingling may be rendering this barrier less effective over time. To the contrary, the Reith anticline apparently is not a horizontal barrier to flow between the Pendleton area and the eastern side of the Stage Gulch administrative area as evidenced by group 2 wells (dark blue triangles) in figure 11B. From north to south, the hydraulic gradients increase proportionally to the amount of geologic structure (compare water levels from figure 11A to the well locations in figure 10), indicating that geologic structure may be impeding lateral recharge from the uplands. Curiously, the exceptionally high hydraulic head contrast between groups 2 and 12 (gray triangles) (figs. 10 and 12D-E) does not correspond to a mapped geologic structure, indicating the presence of a previously unmapped geologic feature.


Near the OWRD administrative areas, a few shallow CRBG aquifers (groups 3 [dark green squares], 6 [light blue triangles], 7 [yellow squares], and 11 [pink triangles]) are receiving recharge from irrigation projects and have stable or slightly rising hydraulic heads (fig. 11B). Water levels in many of the deeper CRBG aquifers, in contrast, have declined 100–300 ft since 1970 as shown in groups 4 (orange circles), 8 (red squares), 10 (purple circles), and 14 (light blue squares) (fig. 11A). Hydraulic heads in groups 15 [dark blue circles], 16 [green circles], 17 [white circles], and 18 [magenta circles] (figs. 10 and 11A) to the south have lower total decline because they are protected from the high pumping rates in the north by horizontal flow barriers immediately north of 
these groups.


Palouse Slope/Eastern Yakima Fold Belt, Washington


The Palouse Slope and the eastern Yakima Fold Belt (fig. 1), which forms a transition area between the two physiographic provinces, encompass the entire area of the Columbia Basin GWMA (hereafter referred to as the Palouse Slope/eYFB). Within the GWMA, WADOE has established several administrative areas (Ground Water Management Subareas) that cover most locations with large long-term groundwater-level declines (Washington State Department of Ecology, 1988a, 1988b, and 1988c). These include the Odessa, Quincy, and 508-14 subareas (fig. 14). The largest groundwater-level declines measured in the central GWMA (1968–2009) are along a north-south swath near the center of the area (fig. 14). Hydrographs for 1,195 wells blanketing the area of largest declines in the CRBG aquifers were examined and divided into groups of wells exhibiting similar changes in hydraulic head over time and a subset of these groups was selected to illustrate these trends (fig. 15). Horizontal barriers to flow are not as evident in this area as in the Umatilla area, but groups are still identifiable based on similar response to hydraulic stresses. This is consistent with the geologic interpretation of the Palouse Slope as being a gently folded structure created during subsidence where CRBG lavas were deposited in voluminous sheet flows. For this area, hydrographs of well groups are most easily viewed along general flow paths that have developed as the result of irrigation stresses on the aquifer system (fig. 15). These flow paths generally trend toward the area with large declines (fig. 14).


Figures 16–19 present hydrographs showing the winter median water levels for individual wells within each well group shown in figure 15 and the trend line for each group constructed using the statistical software program, TIBCO Spotfire S+ (TIBCO Software, Inc., 1988–2008). Water levels in wells in groups 1 (pink circles), 2 (light blue circles), 3 (dark green circles), and 4 (black circles and white circles for shallow and deep wells, respectively) associated with the western flow path (shown as circles in figure 15) start to rise during the 1950s (fig. 16). Prior to 1950, the limited data suggest lower groundwater levels and flatter hydraulic gradients. After 1950, groundwater levels rise and the hydraulic gradient steepens from west to east indicating increased groundwater flow toward Moses Lake and farther east where groundwater declines are associated with widespread irrigation from groundwater.


The wells in groups 8 (brown squares), 9 (orange squares), 10 (dark green squares), 11 (black squares and white squares, shallow and deep), and 12 (gray squares) associated with the eastern flow path (shown as squares in figure 15) show a hydraulic gradient sloping from northeast to southwest, down the Palouse Slope (fig. 17). Water‑level declines for most of these wells begin after 1970. Group 11 shallow wells (black squares) are less affected, indicating hydraulic separation between most of these wells and the wells downslope (to the west). The wells showing significant declines exhibit various decline patterns, although total drawdown is similar in most wells.


Between the eastern and western flow paths and well groups, there is a middle set of well groups, groups 5 (dark blue triangles), 6 (orange triangles), and 7 (pink triangles), (shown as triangles in figure 15), with transitional behavior of groundwater levels. Groundwater levels in most of these wells begin to increase about 1950, as observed with the western well groups. Some of these wells also show groundwater-level declines starting in the 1970s as observed with the eastern well groups (fig. 18). Since 1970, a groundwater mound has formed between Moses Lake, Potholes Reservoir, and the eastern flow path wells. Groundwater-levels in groups 6 (orange triangles) and 7 (pink triangles) generally are higher than in groups to the east and west, which consist of group 4 shallow (black circles), group 5 (dark blue triangles), group 8 (brown squares), and group 12 (gray squares) (figs. 15–18).


The southern well groups (groups 13 [red diamonds], 14 [black diamonds and white diamonds], and 15 [yellow diamonds]) associated with the southern flow path (shown as diamonds in figure 15) show a hydraulic gradient from north to south toward the Snake River (fig. 19), which locally has an elevation of about 350 ft. The hydraulic behavior is complex, although the narrow spacing between adjacent well groups and dissimilar temporal changes indicates there may be some hydraulic barriers between these groups.


The hydraulic behavior of group 12 (gray squares) is at the nexus of the eastern, middle, and southern well groups and apparently is influenced by each of the adjacent groups (figs. 17 and 19), although its component wells are not easily separated into the adjacent groups based on location. Hydraulic head values for group 12 wells commonly are intermediate between the other groups, with some wells exhibiting declines similar to group 8 (brown squares) wells from the eastern well groups (fig. 17), and other group 12 water levels rising similarly to water levels in group 7 wells (pink triangles) from the middle well groups (fig. 18). No clear pattern of depth or location allows group 12 wells to be subdivided into the adjacent groups.


Water levels on the Palouse Slope/eYFB were plotted against stratigraphic position relative to the simulated top of the Grande Ronde Basalt (fig. 20A-I). Analogous to the head distribution in the Umatilla area, heads are similar over large vertical intervals of basalt in the Palouse Slope/eYFB area indicating possible commingling of aquifers. Unlike the Umatilla area, even within a well group and for closely spaced wells, those wells constructed at the same elevation may have hydraulic heads hundreds of feet different from each other. This noisy data and the lack of apparent flow barriers between well groups, results in overlap of the hydrographs between groups. Even within well groups, division of hydrographs by approximate stratigraphic layer does not yield smoothly-varying groundwater-level data that generates well-behaved potentiometric surface maps. This may be due to the complex connectivity between wells and commingled aquifers resulting in a range of composite head elevations (fig. 6).


Generalized potentiometric surfaces were developed for pre-development conditions (prior to 1951; fig. 21) and post‑2000 conditions (fig. 22). Relatively high-error generalized surfaces were created using all CRBG wells without distinction between CRBG hydrogeologic unit because hydrographs within well groups show consistent general trends (figs. 16–19); however, few measurements within each group support the division of wells into hydraulically distinct zones vertically (fig. 20). Figures 21 and 22 were constructed by using median water levels for the defined periods for each CRBG well, and smoothing the data with LOESS local linear trend models (Cleveland and others, 1992). At any location, potential error of the simulated hydraulic head is high; the measured value could be hundreds of feet higher or lower than these generalized surfaces. Because trend models were used, fit is biased toward shallow data because few deep wells with significantly different water‑levels were measured (fig. 20). However, when evaluating model fit, there is little spatial bias in the residuals between the measurement points and predicted surface; therefore, the resulting surfaces are good representations of the patterns in the hydrograph groups. Because of the large amount of smoothing of these potentiometric surfaces, smaller scale features associated with streams and structural barriers to flow, for example, the Frenchman Hills (fig. 1) are not well represented.


Groundwater flow during the pre-development period was from the uplands in the northeastern part of the GWMA toward the Columbia and Snake Rivers to the west and south, respectively (fig. 21). Although the locations where groundwater has the potential for local drainage to surface water features are shaded (fig. 21), the generalized potentiometric surface was not corrected for interaction with surface drainages because flow to local streams depends on the hydraulic head and local connection between the aquifers and the streams. Following the onset of large surface-water irrigation projects near Moses Lake (fig. 22) in the 1950s, water levels in the upper CRBG aquifers increased and formed a groundwater mound, resulting in a reversal of the hydraulic gradient toward the area of declines (compare figs. 14, 21, and 22). For wells examined, groundwater levels under surface-water irrigation areas typically rise about 50 ft, with larger rises occurring locally (for example, in the fault bounded valleys to the south of the Frenchman Hills and the Saddle Mountains). To the east, in the south central area of the GWMA where groundwater pumping is the primary source of irrigation, a trough-shaped depression sloping toward the south has formed, and groundwater flows toward it from east and west (fig. 22). The axis of the trough generally is coincident with the easternmost area of dense irrigation pumping of groundwater shown in the center of figure 22, much of which is occupied by well group 8 (brown squares) in figure 15.


Because most well groups do not exhibit pronounced persistent vertical gradients, the potential for commingling to contribute to hydraulic head declines in this area is not clear from the data. The limited data show that a significant downward hydraulic gradient is present in the eastern and western well groups (group 4 shallow [black circles] and deep [white circles] wells in figures 16 and 20B-C, group 11 shallow [black squares] and deep [white squares] wells in figures 17 and 20E). Large vertical head differences apparently are abrupt in groups 4 and 11, while other well groups (for example group 10 [dark green squares] in figures 20E-F) show vertical gradients that are more continuous and exhibit considerable variability. The downward vertical gradients are consistent with the geologic model (Burns and others, 2011) that shows the deeper aquifers are exposed through erosion along the Snake and Columbia Rivers at elevations consistent with the lower hydraulic heads in these deeper units. In addition, hydraulic heads in the southern well groups are lower than in the trough-like cone of depression immediately to the north (fig. 22), indicating that natural or commingled well leakage might be allowing flow to pass through the possible horizontal flow barrier between group 12 (gray squares) and group 13 (red diamonds) (figs. 15 and 19). The complexity in flow near group 13 is further illustrated by the hydraulic head patterns of groups 13 and 14 (figs. 19 and 20G-I). Despite the fact that group 14 may be divided into two groups with apparently distinct behavior, groups 14-1 (white diamonds) and 14-2 (black diamonds) cannot be separated laterally or into shallow and deep groups, although each of these groups show similar post-1980 characteristics with various group 13 wells (figs. 19 and 20H-I).


The central Palouse Slope shows a general lack of geologic structure or other evidence for strong horizontal flow barriers to the east, but the east-west trending faults associated with the ridges between the western valleys (the Frenchman Hills and the Saddle Mountains) in the eastern Yakima Fold Belt are likely barriers to flow. Luzier and Burt (1974) identified a “groundwater dam” on the Palouse Slope extending from the northwest to the southeast through the junction of Adams, Grant, and Lincoln Counties that is associated with an apparent change in hydraulic gradient (this feature is only partially represented because of the general nature of the smoothed potentiometric maps shown in figures 21 and 22). This steeper hydraulic gradient also was identified in this analysis at the boundary between group 9 (orange squares) and group 8 (brown squares) of the eastern well groups (figs. 15 and 17) and was present at least as early as 1940 prior to substantial water deliveries from the Columbia Basin Project, which began in the 1950s. One of the few locations with persistent bias in the residuals for the pre-development potentiometric surface was in the area where group 9 hydraulic heads were simulated as too high, indicating a local steepening of the hydraulic gradient to the east, and a flattening to the west. The flattening of water levels is evident in group 8 in 1968 (fig. 20D). As an alternative to the groundwater dam proposed by Luzier and Burt (1974), this steepened hydraulic gradient is hypothesized to result from a complex discharge boundary where the upper aquifers are intersected at land surface by incised canyons, which allows water to flow into the sediment-filled coulees. The potential for this outflow is indicated by the pre-development generalized potentiometric surface above or near the land surface in the coulees (fig. 21). Incision of the coulees into the Wanapum Basalt geologic unit (fig. 2) is sufficient that several lava flows and potentially several aquifers in the Wanapum Basalt may be intersected (Burns and others, 2011), possibly forming pathways for groundwater flow that would result in complex head patterns (not shown in figure 21) and apparent alterations in the regional hydraulic gradient. Most late-time (2000–2010) hydraulic heads are now below the elevation of the coulees, indicating that groundwater discharge to the coulees has declined over time. Two additional pieces of evidence support the hypothesis that groundwater historically discharged in this location. First, because shallow aquifers tend to drain into the coulees possibly dewatering parts of these aquifers, fewer wells are completed in the aquifers intersecting the coulees along a north-south swath in this area. Second, the apparently random highly variable heads about 300–400 ft above the Grande Ronde Basalt geologic model unit (approximate elevation of Wanapum Basalt intersected by the coulees) and flattening of the water levels below this stratigraphic elevation are characteristic of a highly variable flow field where groundwater is flowing to local drainage features (compare group 8 [brown squares] wells in figure 20D to hypothetical well groups in figure A2B and the associated discussion). The lowest land-surface elevation in the coulees where the potentiometric surface is above land surface is the controlling aquifer drainage elevation. This elevation is about 1,180 ft, which is the approximate inflection point of group 8 wells in 1968 (fig. 20D).


First posted February 5, 2013