Groundwater-Flow System


Conceptual Model of the Flow System


The study area and conceptual model of Lite and Grondin (1988) was extended to likely natural hydrologic boundaries for the purposes of groundwater-flow simulation. The major hydrogeologic processes and concepts used to define the study area are described in this section. Data collection, groundwater recharge, movement, discharge, and water-level changes are summarized in the following sections.


Lite and Grondin (1988) presented a conceptual model of flow in the study area along the transect A–A′ (fig. 5). The principal hydrogeologic features considered were the basalt aquifers and their interactions with the Rocky Prairie thrust fault and incised creeks, primarily Mosier Creek. Because the Mosier Creek gradient is less than the dip of the CRBG units, the creek cuts across several basalt aquifers along its length, with lower aquifers exposed at higher elevations in the watershed (compare B–B′ with C–C′, fig. 5). The Rocky Prairie thrust fault acts as a groundwater-flow barrier, causing groundwater recharge in the uplands to fill the basalt aquifers until springs and seeps form where aquifers intersect the land surface (fig. 7). 


For the current study, the area covered by the Lite and Grondin (1988) conceptual model was extended to include natural hydrogeologic boundaries appropriate for groundwater flow simulation. The southeastern boundary is coincident with the Columbia Hills anticline where a combination of the anticline and the draped Chenoweth thrust fault suggests that water recharged to aquifers will flow away from the anticline on both sides, implying that groundwater and surface water do not flow laterally across the anticline.


The eastern extent of the Lite and Grondin conceptual model was Rowena Creek and an associated mapped fault. Because the hydrogeologic role of Rowena creek is not readily apparent, and because it may have a hydrogeologic effect similar to Mosier Creek, the study area boundary was moved further east to the Columbia River Gorge and the Columbia Hills anticline, encapsulating the entire Rowena Creek drainage. 


The western extent of Lite and Grondin was a north‑south line generally corresponding to the westernmost edge of the Rocky Prairie thrust fault and the extent of the available data. Rock Creek likely has a similar hydrogeologic effect on parts of the groundwater system as Mosier Creek, so the remainder of the study area was defined by adding the combined total drainage area of Mosier and Rock Creeks (fig. 1). This new boundary is coincident with the ridge to the west of Rock Creek, which is paralleled by a high offset normal fault. The combination of ridge and fault is a likely barrier to groundwater flow, making this boundary a barrier to groundwater and surface water.


The Columbia River forms the entire northern boundary of the study area. All groundwater and surface water in the study area naturally drain from the uplands toward the river, and the river crosscuts all of the basalt aquifers of interest along some part of its length.


In this extended study area, two additional faults were identified as potential barriers to flow based on geologic modeling: the Maupin wrench fault and the Chenoweth thrust fault. Mosier and Rock Creeks intersect multiple aquifers and flow across these potential flow barriers, creating a complex geometry between aquifers and creeks. 


The part of the study area to the south of the Chenoweth thrust fault (fig. 2) is completely covered with overburden. The geometry of the aquifer system is unknown and no groundwater-level data exist. It was assumed that the overburden is similar to the low permeability units described in the northern areas of the study area and that the Chenoweth thrust fault is likely a hydrogeologic barrier similar to the Rocky Prairie thrust fault. Given this combination of overburden and thrust fault, much of the groundwater above the Chenoweth thrust fault likely drains into the creeks above the fault.


Rainfall, and consequently recharge, varies gradationally across the watershed, with more rainfall occurring, and presumably more recharge entering, the aquifer system in upland areas. The older sheet flow basalts that underlie the entire study area are only exposed at land surface in uplands near the structural anticline, allowing recharge into these deeper units (fig. 2). Downslope, parts of the younger intracanyon CRBG lava flows are exposed at land surface, and parts are buried beneath the overburden, which allows recharge into the younger basalt aquifers.


Water enters the CRBG aquifers, flowing from the uplands towards the Columbia River. The Rocky Prairie thrust fault interrupts the lateral continuity of several shallow CRBG aquifers, forming a barrier to flow. Hydraulic heads in the upper CRBG aquifers north of the fault are similar to the stage of the Columbia River, although heads in the same aquifers to the south are hundreds of feet higher. Although the fault is a barrier to flow, it is likely imperfect, so some groundwater may flow through or past the fault while other groundwater drains into creeks from springs and seeps. No groundwater-level data exists for the deeper CRBG aquifers to the north of the Rocky Prairie thrust fault, so it is not known to what extent the thrust fault acts as a barrier to groundwater flow in these deeper aquifers.


At locations where aquifers intersect the creek, groundwater and surface water are in direct connection with each other (fig. 7). If the groundwater level is below the creek, water leaks from the creek into the aquifer, and vice versa. If the aquifer is exposed above the creek level, then water may also flow out of the aquifer through springs and seeps, draining into the creek. All these conditions occur along the length of Mosier Creek between the Chenoweth and Rocky Prairie thrust faults, depending on location and time of year.


Recharge


Three sources contribute recharge to the basalt aquifers. The primary source of recharge is precipitation that infiltrates past the plant root zone to the groundwater system. Second, part of the water pumped for irrigation and domestic usage may return to the groundwater system by infiltration. Third, leakage from streams to the groundwater system can occur in locations where streambeds are permeable and stream water levels are higher than the hydraulic head in the connected aquifer. Recharge infiltration past the root zone from precipitation, irrigation, and domestic usage was estimated, and recharge from streams was estimated during the groundwater simulation process. 


Recharge from precipitation was estimated by two independent methods. The primary method, and the method that provided an estimate of recharge over the entire study area, was based on Precipitation Runoff Modeling System (PRMS) (Leavesley and others, 1996). PRMS is a watershed model that balances the input of precipitation with numerous outputs, including evaporation, runoff, and of primary interest for groundwater flow simulations, water that recharges the groundwater system. Calibration of PRMS was accomplished for the part of the Mosier Creek basin upstream of the stream‑gaging station (streamflow measurement site 4 on fig. 1) by adjusting model parameters to minimize the difference between simulated and observed daily streamflows. The PRMS model was then expanded in space and time to include the entire study area and the period of groundwater‑flow simulation. The simulated values of streamflow were compared to streamflow measurements and seepage estimates made at 14 additional streamflow measurement sites (fig. 1) to ensure the expanded model is a reasonable representation of the entire study area.


The second method to estimate recharge from precipitation uses a computer program, RORA (Rutledge, 1998), to estimate the part of each peak in the streamflow record contributed by flow through the groundwater system. Applied over a long period, the program estimated the mean rate of groundwater recharge that returns to streamflow upstream of the Mosier Creek gaging station (streamflow measurement site 4 on fig. 1). Details of PRMS and RORA are provided in appendix B.


The spatial distribution of annual average groundwater recharge was estimated using PRMS for the period 1955 to 2007. The average value over the entire study area was 9.6 in. (41,100 acre-ft/yr) with recharge varying from about 4 in. in the eastern part of the study area to about 19 in. in the southern upland area. The annual average groundwater recharge to the drainage area upstream of the Mosier Creek gaging station (fig. 1) was estimated at 9.7 in. using PRMS, compared to 13.6 in. estimated using RORA (appendix B.3). This result suggests a range of possible recharge values for the study area, with PRMS providing a relatively conservative lower estimate of recharge compared to RORA. 


The final component of recharge is return flow from pumping for irrigation and domestic usage. Water pumped for use by rural residents on small acreages is either applied to lawns and gardens, or used for household needs. Unless over-watering occurs, most of the water applied to lawns and small gardens is consumptively used by evapotranspiration by plants, whereas household drinking and wash waters are non-consumptively used, returning to the uppermost aquifer through septic drain fields. Because typical rural residential‑exempt water wells are shallow, tapping the uppermost aquifer, most septic drain water is assumed to return to the aquifer being pumped, indicating a negligible net change in water in the uppermost aquifer resulting from pumping for non-consumptive uses. For this reason, only consumptive use pumping was estimated for rural residential small acreages for representation in the groundwater simulation model (see Pumping of Groundwater), and recharge and non‑consumptive pumping from rural residential wells was not simulated.


Irrigation water applied in excess of plant requirements returns to shallow aquifers by percolating through the root zone past the point where plants access the water. Groundwater recharge from irrigation into the principal CRBG aquifers is assumed to be negligible for two reasons. In the relatively small area supplied by large irrigation wells (generally coincident with the OWRD groundwater administrative area, fig. 1), hydraulic heads in the confined basalt aquifers were significantly higher than the overburden water table aquifers, indicating that the amount of recharge to basalt aquifers from irrigation is negligible. This recharge likely enters the overburden aquifers and returns to nearby streams. Additionally, the amount of recharge from irrigation is less than 1 percent of recharge from precipitation because the estimated 740 acre-ft of irrigated water applied to all crops in 2006 (see section Pumping of Groundwater) was 1.8 percent of the estimated average annual recharge due to precipitation (41,100 acre-ft). The high-efficiency sprinkler systems used in the area resulted in only a small fraction of this water infiltrating into the groundwater system. 


Groundwater Flow Direction


Groundwater-Level Monitoring Network


Groundwater levels provide a measure of hydraulic head and water stored in the aquifer system. Hydraulic head is a measure (in units of feet above a datum) of the potential to cause flow due to gravity and water pressure. Groundwater flows from high to low hydraulic head.


A network of wells (table 1, fig. 8) was established by USGS and OWRD to monitor changes in groundwater levels over a 2-year intensive period and for comparison with historical levels. The monitoring network is limited to the eastern side of the study area, in part because this is the area where significant groundwater-level declines have been observed, and because few wells exist in the study area west of the Maupin wrench fault and south of the Chenowith thrust fault (fig. 2). Water levels were collected quarterly (4 wells), bimonthly (26 wells), and continuously (7 wells, measurements recorded bihourly) in a network of 37 wells representative of the aquifers in the study area. These wells were privately owned domestic, irrigation, and unused wells where owner permission was granted and were selected to represent each aquifer over the maximum lateral extent possible.


Many of the groundwater-level measurements in the study area are from wells that are potentially open to multiple aquifers. The groundwater level in each of these wells is a composite hydraulic head, representing a flow-weighted combination of hydraulic heads that occur separately in each of the aquifers. 


Hydraulic Gradients and Groundwater Movement 


The hydraulic gradients and groundwater movement in the study area are controlled by flow barriers associated with the geology. High offset faults interrupt the lateral continuity of the thin basalt aquifers, forming effective barriers to flow, resulting in high gradients across the faults. Laterally extensive, thick confining units separate the basalt aquifers resulting in high vertical hydraulic gradients. 


Pre-1970 water levels in shallow basalt wells south of the Rocky Prairie thrust fault were at a water-level elevation of about 475 ft (figs. 9, and 10), and shallow basalt wells north of the fault had water levels between 70 and 90 ft, similar to the Columbia River stage (about 70 ft). Since 1970, groundwater levels have steadily declined to the south of the Rocky Prairie thrust fault, reducing the gradient across the fault by about 175 ft.


Within each aquifer, the hydraulic head is higher in the uplands than near the Rocky Prairie thrust fault. Horizontal gradients are smaller near the thrust than in the uplands. The reasons for this are not clear, but three potential contributing factors have been identified. First, the transmissivity of younger basalts is known to be high near the OWRD management area in the watershed with possible lower transmissivity as these aquifers extend toward the Columbia Hills anticline (Lite and Grondin, 1988). Second, most water‑level measurements near the anticline are from Frenchman Springs aquifers, and lower in the watershed, measurements are more frequently from the younger aquifers, indicating that the higher gradient may be the result of the Frenchman Springs aquifers being less transmissive. Third, recharge is higher in the uplands where the Frenchman Springs geologic unit is exposed, which would also result in a steeper hydraulic gradient. Lite and Grondin (1988) provide hydraulic head maps for the Pomona and Priest Rapids aquifers. The maps are complicated, with attempts to account for composite heads and seasonal changes near streams. However, the general patterns summarized here and implied by the conceptual model hold true.


The highest vertical gradient measured between any two adjacent aquifers south of the Rocky Prairie thrust fault is a head difference of about 70 ft across the Selah interbed (Lite and Grondin, 1988). Anecdotal evidence and well logs indicate that in the OWRD administrative area (fig. 1) groundwater levels are higher in deeper aquifers when they are encountered during drilling. This is evidence of a persistent upward gradient above, and including at least part of, the Frenchman Springs aquifers. 


Because deepening of wells is often in response to declining water levels and because most of the deeper wells were installed after groundwater declines began, no reliable estimates of pre-development vertical gradients are available. Moreover, interpretation of the water-level measurements made during drilling is further complicated because water‑level data collected during drilling are composite head measurements of units open to the borehole. 


The vertical hydraulic gradient near the Columbia Hills anticline (fig. 2) is downward. This is the likely source of recharge to the Grande Ronde aquifers. No groundwater‑level data are available for the Grande Ronde aquifers in the OWRD groundwater administrative area, and based on geologic modeling results, these aquifers may be connected to the Columbia River to the east, south of the Rocky Prairie thrust fault. The absence of this flow barrier could result in lower groundwater levels in the Grande Ronde aquifers than in the upper CRBG aquifers of the OWRD administrative area, and a resulting downward gradient starting within or below the deeper Frenchman Springs aquifers.


Discharge


Discharge includes all pathways through which water leaves the groundwater system. Groundwater is discharged to surface water features (streams, rivers, springs, and wetlands) as it leaks out of the system or is discharged by pumping from wells. Leaky wells that allow water to flow from one aquifer to another (commingling wells) are internal flow paths, and can affect the rate of discharge into surface water features, but are not considered discharge points from the aquifer system (see section Commingling Wells for a description).


Discharge to Surface Water 


Most groundwater in the study area discharges to streams and the Columbia River with the pattern of gaining and losing stream reaches generally controlled by hydraulic compartmentalization of aquifers by geologic faults. When an aquifer is intersected by a stream, groundwater flows into the stream when the hydraulic head in the aquifer is higher than the water level in the stream or river. Groundwater can also flow into streams from springs and wetlands where water is seeping out of the ground above the stream. The amount of streamflow contributed by groundwater is referred to as base flow. Hydraulic head in the aquifers varies over time, providing variable amounts of flow to streams, springs, and wetlands. Flow rates can vary following storms, seasonally, or on longer timescales in response to decadal precipitation patterns or long-term aquifer declines. Because there is low precipitation during summer months, streamflow during this period consists almost entirely of base flow. Average annual base flow is estimated to be approximately 70 percent of total streamflow at the Mosier Creek gaging station for water years (October 1 to September 30) 1964–81 and 2006–07 (appendix C.2).


The spatial distribution of groundwater exchange with study area streams was estimated by measuring streamflow at many points along the Mosier Creek (compare figs. 1 and 11). The amount of seepage to, or from, a stream reach from the aquifer system is calculated as the difference between the upstream and downstream steamflow after accounting for tributary inflows to and diversions from the reach. Seepage studies of Mosier Creek were conducted in 1962 in a regional groundwater study (Newcomb, 1969), in 1986 as part of a water-availability study by the Oregon Water Resources Department (Lite and Grondin, 1988), and for the current study in 2005 and 2006 (appendix C.2). These latter seepage studies were conducted at various times throughout the year to account for seasonality of water exchange. These data were used during calibration of the PRMS hydrologic model (see section Recharge and appendix B).


Summer estimates of seepage (fig. 11) show persistent groundwater discharge patterns. Flow measurement patterns are complex, consistent with the observation that several aquifers are intersected by Mosier Creek upstream and downstream of the stream gage. The Rocky Prairie thrust fault groundwater-flow barrier is evidenced by increasing streamflow and specific conductance associated with the fault (river mile [RM] 0.8) where groundwater is forced to discharge to the stream. Although there is a pronounced reduction in base flow when comparing the September 1962 streamflow measurements to later measurements, precipitation at the proximal Hood River rain gage (fig. 1) was significantly higher during August and September of 1962 than for the periods preceding all other measurements. For this reason, clear linkages between the declining groundwater levels and base flow cannot be made. 


Data from the five seepage studies conducted during 2005–06 (fig. 11 and table C2) provide evidence that the Chenoweth thrust fault (RM 7.1) is also a groundwater-flow barrier. The percentage of streamflow measured at the Mosier Creek gaging station (site 4 on fig. 1) that is in the creek immediately south of the Chenoweth thrust fault (site 1 on fig. 1) ranges between 74.3 and 106.2 percent (table C2), with a median value of 80.0 percent, indicating that groundwater may be forced into Mosier Creek above the Chenoweth thrust fault rather than flowing across the fault through the aquifer system. The single measured value greater than 100 percent (August 2006) indicates that water was lost to the aquifer system below the fault and upstream of the gaging station. The PRMS estimate of groundwater recharge upstream of streamflow measurement site 1 (near the thrust fault, compare figs. 1 and 2), is approximately 16.5 ft³/s on average for 1955–2007, and the average annual base flow was estimated using the PART hydrograph separation computer program (Rutledge, 1998) to be 20.7 ft³/s at the Mosier Creek gaging station (streamflow measurement site 4) for 1964–81 (see appendix C.2 for details regarding the use of PART). 
If all the PRMS estimated groundwater recharge south of the Chenoweth thrust fault were forced into the stream as base flow above the thrust fault, then it would be 79.8 percent of the estimated annual average base flow at the gaging station, matching the measured streamflow ratios well, providing evidence that the thrust fault may be an effective barrier to groundwater flow. 


The control of the thrust faults on Mosier Creek base flow suggests a relation between geologic faults and streamflow that explains the observed flow patterns of Rock, Rowena, and West Fork Mosier Creeks (fig. 2). West Fork Mosier Creek also is a perennial stream with headwaters above the Chenoweth thrust fault and the Maupin wrench fault, indicating that these faults may also promote groundwater discharge to the creek. Even though Rock Creek flows through gravels with no surface expression low in the watershed during the summer, flow was documented above the gravels during all periods of measurement at site 14 (fig. 1). Rock Creek is crossed by several high-angle faults, creating the potential to force groundwater flow into the Creek. If the Chenoweth thrust fault continues to the west beneath the volcanic deposits (fig. 2), the headwaters of Rock Creek are above this thrust fault, which may result in water from the upper watershed being forced into Rock Creek, similar to Mosier Creek. The ephemeral Rowena Creek is on the eastern side of the study area, receiving less recharge and crossing only one inferred fault, although the creek runs along a mapped fault for some distance (fig. 2). The ephemeral character of Rowena Creek can be explained by the lack of an extensive source area and lack of compartmentalizing faults crossing the creek. 


Pumping of Groundwater


In the study area, groundwater is used for irrigation, public supply, and self-supplied domestic uses. Groundwater use began in the first half of the twentieth century, however, most wells were constructed starting in the 1970s (fig. 12). Even though far more self-supplied domestic wells have been drilled, the consumptive use of water in the study area has been primarily irrigation (fig. 13). Estimates of water usage for each category are summarized below and details are discussed in appendix D.


During the intensively measured 2006 irrigation season, irrigation was the largest use of groundwater, accounting for about 80 percent of total volume pumped, with public-supply and self-supplied domestic accounting for about 10 percent each. A total volume of about 740 acre-ft of groundwater was applied to almost 860 acres from 19 wells in or near the OWRD administrative area (fig. 14). This water was used in the production of fruit tree crops, including cherry and to a lesser extent pear and apple. Wine grapes also are becoming a significant crop in the study area.


Three basic types of irrigation methods are currently used in the study area. In 2006, an estimated 534 acres (62 percent) were equipped with micro spray irrigation, 169 estimated acres (20 percent) were using low efficiency impact sprinklers, and an estimated 155 acres (18 percent) were using drip irrigation. Low-efficiency impact sprinklers were once the standard means of irrigation, but this method is being replaced systematically with methods that are more efficient. The proportion of land using drip irrigation has recently increased.


Public supply for the city of Mosier is another major use of groundwater. The city relies on one primary well to supply water to approximately 430 residents with another well serving as backup water supply (fig. 14). In 2006, the primary well pumped approximately 87 acre-ft of groundwater, and the backup well pumped nearly 3 acre-ft for a combined pumpage of 90 acre-ft. This is about 10 percent of total pumping in the study area. Public-supply water usage from 1989 to 2006 was reported by the city of Mosier. Pre-1989 public-supply water use was estimated using the average 1989 per capita water use rate and estimates of historical population. Details of the city’s pumping estimation process are included in appendix D.


In 2006, about 1,200 rural residents pumped an estimated 490 wells, totaling about 114 acre-ft of consumptively used (water used for lawn irrigation etc.) groundwater (about 10 percent of the total pumping in the study area). Non-consumptively used (water used in households) water was assumed to recharge the uppermost aquifer, which is typically the aquifer being pumped by rural residents. Because this indicates no net change in aquifer storage from non‑consumptive pumping and recharge (estimated as 60 percent of total annual rural residential pumping), both the non-consumptive pumping and recharge from rural residential wells are neglected in water budgets and model input. Time‑varying pumping for rural residential use was estimated based on assumptions about historical population, typical water use per capita, and the percentage of water typically used consumptively (estimated as 40 percent of total annual pumping) (details provided in appendix D).


Commingling Wells


Well boreholes drilled through multiple aquifers can allow water to flow between aquifers unless seals are installed to prevent this. Vertical flow through the borehole occurs when there are differences in the hydraulic heads of the aquifers penetrated by the well. Water flows from high hydraulic head to low hydraulic head through the well bore. This mixing (or mingling) of waters from different aquifers provides the name commingling well, which also is sometimes called a cross connecting well. 


A number of wells in the study area are documented as being drilled through multiple basalt layers but having a minimum length seal (approximately 20 ft of sanitary seal immediately below land surface) between well casing and the geologic formation. Frequently, wells also only have casing that extends from land surface to the top of the uppermost CRBG unit. Because wells are commonly uncased and open below the top of CRBG, commingling can occur freely between basalt aquifers intersected by the well. Even when casing is installed, if there is no well seal between the casing and formation that prevents flow, commingling occurs in the annular space between the casing and geologic formation (including flow into or out of the overburden aquifer). In the case where aquifers receiving water have a significant ability to retain the water, groundwater levels can increase. However, in the OWRD administrative area, the glaciofluvial overburden and CRBG aquifers have low water storage capacity and are highly transmissive, so most of the water passes through the commingled aquifers into local springs and streams with only a small increase in storage within the receiving aquifers. 


Commingling wells allow leakage from the aquifer system that can result in groundwater-level declines. Prior to installation of wells, water levels were higher in the highly permeable CRBG aquifers (fig. 15A) (Lite and Grondin, 1988). Installation of a well with an ineffective seal allows water to flow out of the basalt aquifers into the overburden (fig. 15B). If the overburden is sealed off, then water flows from the deep basalt aquifers into shallow basalt aquifers. In either case, hydraulic heads decline in the deeper aquifers, with the amount of head reduction depending on how easily water flows out. 


During geophysical testing of a known commingling well to the south of the Rocky Prairie thrust fault (well 454033121230101, appendix F), the measured upward flow rate through the well ranged from 70 to 135 gal/min (11–22 percent of total annual pumping in the study area). Historically, when aquifer water levels in the deeper basalt aquifers were 150–200 ft higher, and the head contrasts between the deeper and shallower aquifers were higher, this flow rate would have been correspondingly higher.


Possibly commingling wells were identified using a rule‑based algorithm for representation in the groundwater‑flow simulation model. The probable deepest aquifer was selected by using well depth data and the digital geologic model. Because common practice is to have no casing installed in the length of borehole open to competent basalt, boreholes passing through more than one aquifer were identified as possibly commingling, unless well construction data indicated an effective seal was in place. Rural residential wells with no well depth data were assumed to pump from only the shallowest aquifer and as a result, not commingling any aquifers, but this assumption possibly underestimates the number of commingling wells. To the contrary, the number of commingling wells may have been overestimated because, even though a geologic contact is present in a borehole, productive aquifers do not occur at all locations due to depositional variability of the basalt interflow zones. 


Regardless of the possible complicating factors, applying the aforementioned assumptions allowed creation of a reasonable distribution and chronology of well construction (fig. 16) that allows testing of the net effects of commingling wells with a groundwater-flow simulation model. Since approximately 1995, the number of possibly commingling wells has stabilized at about 150. This is presumed to be the result of improved well construction practices in the OWRD administrative area (where most deep irrigation wells exist) and the fact that most new wells are rural residential wells that are typically constructed in shallow aquifers.


Temporal Variation in Groundwater Levels and Changes in Groundwater Storage


Changes in groundwater levels correspond to changes in water storage within the aquifer system. The amount of water in the groundwater system varies in time as a result of hydraulic stresses. Annually, water storage increases during wet winter months as precipitation recharges the system, and decreases during the drier summer months as water continues to discharge from the system into creeks and the Columbia River. Groundwater storage may also vary on longer timescales, such as decadal, resulting from multi-year wet or dry periods. If long-term average groundwater recharge remains the same and no additional water is removed from the aquifer system, groundwater levels will oscillate over time, but the average levels remain constant. This condition is called dynamic equilibrium. 


The addition of pumping and commingling wells to the aquifer system has resulted in declines in groundwater storage. These declines are present in many study area wells (fig. 9), and until the aquifer system reaches a new dynamic equilibrium, groundwater levels will continue to decline. The persistent groundwater-level declines are superimposed with seasonal and decadal oscillations, representing the effects of seasonal recharge, pumping, and decadal wet-dry periods.


Persistent Groundwater Level Declines


Steadily declining water levels in CRBG wells to the south of the Rocky Prairie thrust fult are generally coincident with the OWRD groundwater administrative area and the majority of groundwater pumping in the study area (figs. 9, 10B, and 14). Long-term water level measurements were examined in wells to identify groups of wells with similar hydrologic response (fig. 10A). The largest declines in the study area were measured in Group 1 wells, which have declined at a persistent 4 ft/yr, beginning during the early to mid-1970s. Group 2 wells have a similar response, although the starting water levels are lower, and the rate of decline is smaller (fig. 9). 


By 2006, water levels in most Group 1 wells had declined about 150 ft over 35 years with water levels in these wells typically within 25 ft of each other for most of this period. Water levels in several Group 1 wells seemed to be distinctly different when originally drilled (well 454037121205601, for example); however, within 1–2 years, water levels in these wells became similar in magnitude and rate of decline to water levels in other Group 1 wells. Even though the linear decline dominates the pattern of response for Group 1 wells, seasonal and slight interannual trends are apparent, and these variations commonly are reflected in more than one well (fig. 17). 


A smaller group of wells with similar behavior (Group 2 in figs. 9 and 10) is clustered immediately to the south of the Rocky Prairie thrust fault. Group 2 wells have lower initial water levels, but these wells are also steadily declining and appear to be trending towards a similar final hydraulic head. Although the thrust fault is mapped to the north of the wells, the part of the aquifer system affected by faulting likely extends towards these wells. Water levels in these wells are interpreted as being driven by the same physical processes as Group 1 wells, but having a reduced response due poor hydraulic connection within the fault-affected zone. 


No other well hydrographs have the persistent steep linear declines exhibited by Groups 1 and 2. Water levels in upgradient wells (for example well 453845121191401, fig. 10A) also exhibit declines (fig. 18), although the rate of decline since the mid-1980s is typically smaller than for Group 1 wells. However, comparing the 1978 and 1985 groundwater levels from well 453845121191401 (fig. 18) indicates that water levels dropped about 24.5 feet over 6.5 years, indicating a rate of decline of 4 ft/yr or greater may have occurred during periods since onset of Group 1 declines in the 1970s. 


Water-level elevations in all upgradient wells range from a few hundred to more than 1,700 ft higher than Group 1 wells. The data were sufficiently sparse for upgradient wells prior to 1984, that significant groundwater level responses during the 1970s are poorly documented. However, there is a general trend of apparent steeper groundwater level declines in several upgradient wells during the 1970s followed by flattening of the hydrographs in the 1980s. Crude, two-point estimates of average decline during the 1970s are between 4 and 8 ft/yr for upgradient wells.


Seasonal Variation in Groundwater Levels 


Water levels in wells fluctuate seasonally in response to changes in recharge, evapotranspiration, groundwater pumping, and streamflow. Beginning in autumn and continuing through mid-spring, water levels rise as recharge from precipitation to the groundwater system exceeds discharge to evapotranspiration, groundwater pumping, and streamflow. From mid-spring until autumn, water levels decline as water drains or is pumped at increased rates from the aquifer system, and during a period when recharge from precipitation is much lower and evapotranspiration is higher.


The seasonal water-level variation ranged from a negligible amount to about 50 ft. Wells with the greatest water‑level ranges were in the OWRD administrative area, with smaller seasonal water-level ranges above the administrative area and to the north of the Rocky Prairie thrust fault. Most seasonal water-level changes ranged between 10 and 25 ft in the administrative area (fig. 17).


Decadal Variations in Groundwater Levels 


A comparison between precipitation at Hood River and groundwater levels in the OWRD administrative area wells reveals that a part of the groundwater level changes in the study area is likely related to decadal-scale wet and dry periods (fig. 19). The annual total precipitation by water year is strongly correlated between Hood River to the west and The Dalles to the east, indicating either precipitation record can be used as a surrogate for wet and dry periods in the study area. Average precipitation at Hood River has been approximately 31 inches per water year since 1950, with more persistent decadal-scale wet and dry periods of precipitation after 1970. 


To examine the response to decadal variation in precipitation, a shallow and deep pair of Group 1 basalt wells was selected and de-trended. First, the seasonal patterns were removed by selecting late-winter water levels, followed by removal of the linear trend. For both wells, the best-fit slope for the post-1974 data was 3.9 ft/yr of decline. The shallow well (454031121215701) is open to the uppermost basalt aquifer, and water levels follow the wet and dry periods closely (fig. 19). The nearby deeper well (454031121224001) is completed in an aquifer that is several basalt aquifers below the shallow well, and water levels show a decadal scale trend; however, the response is attenuated and lagged by as much as 10 yrs with respect to the shallow well response. The deep well response is more typical of hydrographs for basalt wells in the OWRD administrative area (fig. 9), indicating the typical climate-driven variation of water levels for wells in the deeper basalt aquifers in this area is approximately 10 ft. 


Regularly, the decadal-scale relative high groundwater level of the shallow well corresponds to a decadal-scale relative low of the deep well (fig. 19). When comparing the water-level measurements of these two wells since 1979 (fig. 9), the hydraulic head difference between the well pair has varied between –12 and +67 ft (negative value indicates a downward gradient) with a typical 14 ft upward gradient (computed as the difference between the hydrograph trend lines). The larger variation associated with the shallow well is likely a localized phenomenon, associated with an aquifer of limited areal extent. 


Conclusions from Analysis of Groundwater Levels


The following are the primary conclusions from the analysis of groundwater levels:


1. Most basalt wells in the OWRD administrative area have seasonal variations of 10–25 ft, decadal oscillations of approximately 10 ft, and persistent linear declines of about 4 ft/yr. A few wells, likely representing shallow aquifers of limited extent, have larger water level fluctuations, but are still declining at a rate of approximately 4 ft/yr. 

2. Groundwater levels outside the administrative area are highly variable, with many exhibiting declines and oscillatory behavior, although documented rates of decline generally are significantly less than 4 ft/yr. Few data are available for most wells upgradient of the OWRD administrative area prior to 1984, although limited data support water-level declines in several wells during the 1970s at rates of 4–8 ft/yr with significantly lower rates of decline after this period. 

3. Groundwater levels respond to decadal precipitation patterns. Because the post-1970 period has higher average precipitation than for the 20 years prior, groundwater levels should be rising rather than falling as observed. Therefore, the persistent groundwater declines in the study area cannot be attributed to changes in precipitation.


Conceptual Model of Changes in Groundwater Storage


Changes in groundwater levels are a result of the combination of pumping, commingling, and varying recharge. Reduction in recharge is an unlikely contributor to the persistent groundwater level declines beginning in the 1970s, but the relative contributions of pumping and leakage from the aquifer system due to commingling are more difficult to distinguish. To understand the groundwater conditions and the relative contributions of pumping and commingling to groundwater declines, a groundwater-flow simulation model was developed to incorporate the available data and to represent the complex flow paths within the aquifer system. 


The geologic model was used to develop a groundwater‑flow simulation model geometry that satisfies the conceptual model of groundwater flow direction (see section “Conceptual Model of the Flow System”). In addition, a conceptual model of storage changes was also developed to aid in the flow simulation model analysis. This conceptual model is illustrated for a single groundwater level in a single basalt aquifer in the OWRD administrative area (fig. 20). Under pre-development conditions, the groundwater system is in dynamic equilibrium (or steady state) with groundwater levels varying seasonally, but exhibiting no long-term trends. This period is represented by the constant groundwater level prior to 1950, after which wells were drilled and pumping begins. Between 1950 and 1972, a few wells were drilled into the upper aquifers in the lower watershed, including the OWRD administrative area. Because only upper aquifers were penetrated, commingling was negligible, and pumping resulted in groundwater declines, with water levels stabilizing at a few tens of feet lower than under pre-development conditions (early-time steady state in figure 20). Starting in the early 1970s, additional wells were installed (fig. 12), increasing the amount of pumping (fig. 13) and the number of aquifers potentially commingled (fig. 16). The resulting groundwater level response to the combined pumping and commingling is much more pronounced, with a significantly lower final water level that has not yet been reached (late-time steady state in figure 20).


Group 1 well water-level data support this conceptual model; although measured groundwater levels are declining at much more linear rate than shown for the conceptual model (compare figs. 9 and 20). Figure 20 emphasizes that if the well configuration, pumping, and recharge remain constant, the system will eventually approach a new equilibrium condition. Under these constant conditions, the rate at which the system approaches the new equilibrium is controlled by the properties governing storage change in the aquifer system, but the final steady-state groundwater levels only depend on the amount of groundwater flowing through the system and the groundwater flow paths (including pumping and commingling). In other words, the magnitude of the declines provide the most information about the relative effects of pumping and commingling, and the rate of decline provides information about the storage and release of water from the aquifer system. For this reason, the groundwater-flow simulation methods employed to identify the principal causes of the large declines in groundwater levels (see section Separation of Pumping and Commingling Effects) emphasize representation of the magnitude of the declines rather than the rate of the declines.


Practices that would restore groundwater levels are reductions in pumping and repair of commingling wells. If all pumping was ceased in the study area, then water levels will recover (fig. 21). If commingling is negligible, then groundwater levels will recover to pre-development conditions. If commingling is not negligible, then groundwater levels will recover to some lower value. The difference between the recovered value and the pre-development steady‑state value can be attributed to the effect of commingling. The difference between the recovered value and the theoretical late-time steady-state value can be attributed to the effect of pumping. This relation formed the foundation of the groundwater-flow simulation model analysis of relative effects of pumping and commingling.


First posted March 1, 2012