Groundwater Hydrology

The elements of the groundwater hydrology of Johnson Creek basin include the basic inflow (recharge) and outflow (discharge), and the change in storage manifested by fluctuations in water levels. Understanding of these elements provides a foundation for discussion of the groundwater-related events that motivated this study: extremes in streamflow of Crystal Springs Creek, the emergence of the water table in a low-lying area of the basin (Holgate Lake (map number 43)), and spatial and temporal trends in groundwater discharge to Johnson Creek.

Recharge

Recharge to the Johnson Creek basin is through the infiltration of precipitation, although in areas of urban development, the diversion of stormwater runoff into UIC systems and infiltration of effluent from onsite waste-disposal (septic) systems are additional sources of recharge. Recharge to the Portland basin was estimated using data from the late 1980s by Snyder and others (1994, p. 30). Within the Johnson Creek basin recharge ranged from 0 to 48 in/yr, with a mean of about 22 in/yr. Recharge from UIC systems and septic systems contributed about 11 and 12 percent of the total, respectively. Irrigation-return flow and losing streams also may constitute locally important sources of seasonal recharge; however, they were not quantified by Snyder and others (1994, p. 3).

The addition of UIC systems and the removal of septic systems since the study by Snyder and others (1994) require a recalculation of recharge based on 2008 conditions using the methods of Snyder and others (1994). The addition of UIC systems increases recharge by routing stormwater runoff onsite to the subsurface, whereas stormwater runoff was previously routed out of the basin through a combined sewer system. Removal of septic systems decreases recharge by routing sewage to the sewer system that had previously been infiltrated onsite. For this study, it was assumed that the pattern of spatial distribution and rates of precipitation have not changed much since the study by Snyder and others (1994). Although impervious area is expected to increase because of infill development in the urban areas and new development in the rural areas, it was beyond the scope of this study to evaluate the effect of these changes on the amount of recharge from the infiltration of precipitation. Therefore, the infiltration of precipitation calculated by Snyder and others (1994) was considered to be representative of conditions for this study. The number of UIC systems has increased substantially and nearly all septic systems in the areas used in that analysis have been removed as part of an extensive sewer improvement project in central Multnomah County. Recharge from septic systems in rural areas of the Johnson Creek basin was not quantified by Snyder and others (1994, p. 26) or by this analysis because these areas generally are less densely populated than the rest of the basin. However, septic systems do remain in some unincorporated urbanized areas but they are the focus of efforts for decommissioning (Andrew Swanson, Clackamas County’s Water Environment Services, written commun., 2008). The source of the water supply used for septic systems can be either internal or external to the Johnson Creek basin; however, for this study, no distinction was made regarding the source of water for septic systems. Recharge is possible from overbank floodwaters of Johnson Creek, particularly in the area just east of I-205. This recharge might affect groundwater levels and the quantity of discharge at locations in the study area and downgradient of the recharge area for short periods of time.

The analysis by Snyder and others (1994, p. 23) estimated there were about 3,700 UIC systems owned by public agencies in Multnomah County and an undetermined number of privately owned UIC systems. Estimates of the number of UIC systems in the Portland metropolitan area in 2008 include about 11,000 publicly owned and 25,000 to 35,000 privately owned UIC systems, most are in Multnomah County (Snyder, 2008, p. 6-7). The area drained by the newly installed, publicly owned UIC systems in the Johnson Creek basin is an additional 2,100 acres. These new UIC systems are estimated to contribute recharge at a rate of 10.6 in/yr, the same mean rate of recharge for areas with UIC systems already identified within the Johnson Creek basin (Snyder and others, 1994, pl. 1B), for about 1,800 acre-ft/yr of additional recharge to the Johnson Creek basin. This additional recharge represents a 22-percent increase in recharge from UIC systems to the Johnson Creek basin since the analysis by Snyder and others (1994), which estimated recharge from UIC systems to be about 8,100 acre-ft.

The recharge from areas with septic systems within the Johnson Creek basin identified by Snyder and others (1994, pl. 1C) totaled about 7,400 acre-ft/yr. The installation of sanitary sewer systems in these areas has resulted in the decommissioning of nearly all these septic systems. For this calculation, the quantity of recharge previously attributed to septic systems has been diverted to the sewer systems and no longer contributes to recharge in the Johnson Creek basin.

The 2008 annual recharge for the Johnson Creek basin is equal to the recharge estimate of 66,600 acre-ft as reported by Snyder and others (1994), supplemented by 1,800 acre-ft of recharge from new UIC systems, and reduced by 7,400 acre-ft because of the decommissioning of septic systems resulting in an estimated annual recharge of 61,000 acre-ft for 2008 in the Johnson Creek basin. The change in recharge resulting from the 2008 configuration of UIC systems and septic systems by about 5,600 acre-ft/yr represents about an 8-percent reduction since the analysis by Snyder and others (1994) to the Johnson Creek basin. This change in recharge was almost entirely within the intervening drainage basin area between the Milwaukie and Sycamore sites. The recharge to the drainage basin area upstream of the Sycamore site is unchanged from the analysis of Snyder and others (1994) and was estimated at 34,500 acre-ft/yr. These estimates of recharge assume that the infiltration of precipitation is identical to that of the previous analysis by Snyder and others (1994) and does not take into consideration the substantial but unknown amount of recharge contributed from privately owned UIC systems or from the installation of new septic systems outside areas where sanitary sewers are available. These changes to recharge may result in changes in the discharge to rivers, streams, and springs or may be reflected in changes in groundwater levels.

Groundwater Discharge

Groundwater in the Johnson Creek basin discharges to springs, to the stream by leakage through the streambed, to wells, by evapotranspiration, and by subsurface flow to other basins. The existence of streamflow during the late summer, a period of little or no precipitation, indicates groundwater discharge. Groundwater withdrawals by wells are mostly restricted to the eastern area of the basin, where wells are used for domestic supply and for irrigation (Collins and Broad, 1993). Municipal sewage is not discharged to Johnson Creek. Evapotranspiration is not a major source of groundwater discharge because the depth to groundwater typically exceeds plant root depth. Evapotranspiration losses probably are greatest in areas of shallow groundwater adjacent to Johnson Creek and in wetland areas just east of I-205 at Beggars Tick Marsh (pl. 1).

Groundwater discharge, which contributes to flow of Crystal Springs and Johnson Creeks, was estimated in several ways: (1) using measurements of springs (and predominantly spring-fed streams) available since the 1970s, (2) using seepage measurements made along Johnson Creek from 1988 to 2006, (3) comparing streamflow in Johnson Creek at the streamflow sites during the low-flow period from 1989 to 2006, (4) comparing base-flow separation results for different time periods at the Johnson Creek streamflow sites, and (5) assessing qualitatively the groundwater discharge to Johnson Creek using environmental tracers (chemical and thermal characteristics of the stream).

Springs and Seepage

Johnson Creek is a perennial stream, where flow is sustained during the dry summer and autumn by groundwater discharge. Groundwater discharge occurs as springs, where a distinct flow emanates from an area, typically as a small stream forming a tributary to Johnson Creek, or as seepage through the streambed. Seepage may be visible if it discharges along the banks above stream level. Relative to streamflow, springs and seepage studied in this report are groundwater contributions to the stream. Most spring and seepage discharge to Johnson Creek is downstream of RM 5.5. The predominance of springs and seepage in the lower basin results in relatively high summer flow compared to the flow upstream of RM 5.5.

Crystal Spring (fig. 1) is the largest spring in the Johnson Creek basin, and, as referred to in this report, consists of several complexes of springs that discharge to Crystal Springs Creek. The individual spring locations are described in McFarland and Morgan (1996, p. 8, p. 24-27) and in McCarthy and Anderson (1990, p. 31). Crystal Springs Creek is about 2.5 mi long, flows westward from the Reed College campus, southward through a golf course, Westmoreland Park, and residential areas, entering Johnson Creek at RM 1.0. Crystal Springs Creek provides most of the summer streamflow of Johnson Creek downstream of RM 1.0. Spring flow originates in two general areas. The upper spring complex is on the Reed College campus, where flow emanates from the bluffs surrounding a lake and wetland area. The lower spring complex is about 1 mi from the mouth of the creek, where flow emanates at several locations from the bluffs to the east, flowing into a constructed pond, and then into the main channel of Crystal Springs Creek.

An increase in the height of the water surface of Crystal Springs Creek, noted by local residents beginning in July 1997, prompted inquiry by the city of Portland (Chestnut, 1997). Streamflow measurements were made at three locations: (1) on the Reed College campus at RM 1.8 (map number 38), (2) at RM 1.0 (map number 39), and (3) at the mouth of the creek (map number 40) (pl. 1). The streamflow measurements characterized the peak in 1997, tracked a multiyear recession, and by 2001 the streamflow had receded to the rate observed a decade earlier (fig. 2).

Measurements of streamflow of Crystal Springs Creek in the Reed College area began in 1988 when the combined discharge of several distinct springs was 1.6 ft³/s (McFarland and Morgan, 1996, p. 26). In this study, discharge of the springs, determined by streamflow measurements of Crystal Springs Creek in the Reed College area, ranged from a maximum of 6.3 ft³/s in September 1997 to a minimum of 0.16 ft³/s in September 2005 (fig. 2). Two measurements in 1999 indicated greater streamflow; however, surface runoff or release of water from a beaver dam may have temporarily increased the flow of Crystal Springs Creek. Streamflow measurements made in 2001, 2003, and 2004 indicated that the flow was about 1 ft³/s. The low streamflow observed in 2005 was followed by an increase to 1.31 ft³/s in 2006.

Streamflow measurements made prior to 1997 (primarily during the summer and autumn) at or near the mouth of Crystal Springs Creek indicate that the collective discharge of the springs was between 10.5 and 14.8 ft³/s. Concurrent streamflow measurements in 1997 and 1998 at RM 1.0 and at the mouth of the creek indicate that the difference in streamflow usually is within the range of measurement uncertainty, so data from either site can be used for comparison. The earliest known streamflow of Crystal Springs Creek was 12.4 ft³/s in 1872, when the City of Portland was considering the feasibility of tapping Crystal Springs Creek as a source for public water supply (McFarland and Morgan, 1996, p. 26). The high streamflow of Crystal Springs Creek, as measured in 1997, seems to have begun the previous year. Although Crystal Springs Creek was not measured in 1996, streamflow was inferred from streamflow of Johnson Creek at Milwaukie (map number 41) by using the relation of late summer streamflow of Crystal Springs Creek to that of Johnson Creek at Milwaukie (fig. 3). The streamflow measurements of Crystal Springs Creek used for this analysis were made during the period from August to October 1989 and August to October of every year from 1997 to 2006. These streamflows were related to the annual 7-day minimum streamflow of Johnson Creek at Milwaukie for the same period. Based on this relation, the streamflow of Crystal Springs Creek in late summer 1996 was estimated to be 19.5 ft³/s, close to that measured in July 1997.

The streamflow of Crystal Springs Creek at the mouth (map number 40) decreased from 1997 to 2001, decreased less rapidly from 2002 to 2005, and similar to the upper spring area, increased slightly in 2006 (fig. 2). The streamflow reached 20.1 ft³/s in September 1997, which is about 60 percent greater than measured or estimated values between 1987 and 1995. Although the maximum streamflow is in December 1999, this high flow probably was a result of surface runoff or beaver activity. Streamflow in July 2005 was 9.2 ft³/s, lower than any previous measurement of Crystal Springs Creek.

Errol Spring, a complex of individual springs, emanates from a bluff about 0.3 mi from Johnson Creek and forms what is known locally as Errol Spring Creek (map number 34) (pl. 1). Errol Spring Creek passes through a wetland and enters Johnson Creek at RM 3.1.

Another spring discharges to Johnson Creek through a pipe (map number 35) a few feet downstream of the mouth of Errol Spring Creek. This spring emanates from the bluffs near Errol Spring and flows under industrial properties nearby, incorporating street drainage along its length, and is piped to Johnson Creek. Measurements of this spring were made in late summer of 1988, 1999, 2000, and 2006, and indicated fairly similar discharge to Errol Spring Creek.

Errol Spring Creek was measured in August and September 1988 when the streamflow was 0.56 and 0.39 ft³/s, respectively (fig. 2). Errol Spring Creek was first measured for this study in 1998 when the streamflow was 0.97 ft³/s, or about twice the streamflow measured a decade before. Streamflow of Errol Spring Creek remained relatively high in 1999 and 2000. Measurements from 2001 to 2004 indicated a steep decrease compared to streamflow observed in 1998 to 2000. Flow decreased to 0.19 ft³/s in 2005 and increased to 0.37 ft³/s in 2006. The general pattern of fluctuation in streamflow of Errol Spring Creek from 1988 to 2006 was similar to that of Crystal Springs Creek.

Numerous springs emanate from bluffs along the southern side of Johnson Creek between RM 3.2 (map number 33) and RM 2.2 (map number 36). These springs have not been measured directly by the USGS; however, streamflow measurements made on Johnson Creek at RM 3.2 and 2.2 in 1988, 1998, 1999, 2000, and 2006 indicate groundwater inflows (including Errol Spring) of between 0.73 and 2.40 ft³/s (table 2).

Several springs discharge near the mouth of Johnson Creek in Milwaukie. Spring Creek enters Johnson Creek at RM 0.4. The streamflow of Spring Creek was 1.2 ft³/s in 1988 (McFarland and Morgan, 1996, p. 57). Another spring once used to fill a public swimming pool in downtown Milwaukie is piped to Johnson Creek at RM 0.1. Collectively, the discharge of springs to Johnson Creek between the Milwaukie streamflow site (map number 41) and the mouth of the creek (map number 42) was 4.3 ft³/s in 1988 and 4.1 ft³/s in 2006.

Comparison of the decrease in flow of springs in the high-flow years of 1996–99 to the low flows of 2005 indicates that streamflow of Crystal Springs Creek in the Reed College area and Errol Spring Creek decreased much more rapidly than Crystal Springs Creek at the mouth. By 2005, the flow of Crystal Springs Creek in the Reed College area was about 3 percent of the peak streamflow, and Errol Spring Creek decreased to about 20 percent of that observed in 1998. Streamflow of Crystal Springs Creek at the mouth decreased to 50 percent of the peak, indicating that the saturated thickness of the groundwater contributing area to the upper Crystal Springs and Errol Spring Creeks is much less than that of the lower Crystal Springs Creek because of the higher topographic position of the upper Crystal Springs and Errol Spring Creeks. Therefore, changes in the water-table elevation have a proportionately larger effect on groundwater discharge to upper Crystal Springs Creek and Errol Spring Creek relative to the groundwater discharge contributing to lower Crystal Springs Creek. The springs contributing streamflow to the lower area of Crystal Springs Creek additionally may be the result of regional groundwater flow, which may be less sensitive to short-term climatic variation.

Seepage is the exchange between groundwater and surface water at the streambed forming the boundary between a stream and an aquifer system. Rates and locations of seepage to and from Johnson Creek were determined by evaluating the difference (gain or loss) in measured streamflow between measurement locations. Although some recharge to the system is from losing reaches of Johnson Creek, most of reaches are gaining, receiving discharge from the groundwater system. Seepage measurements were made each summer from 1997 through 2000 and in 2006 as part of this study and compared to seepage measurements made in 1988 (fig. 4, table 2).

Precipitation in water years when seepage measurements were made varied widely (fig. 5). Although the average (WY 1911–2006) precipitation at the Portland Airport is about 37 in., precipitation was several inches less than average in WY 1988 and 2000, and several inches more than average in WY 1998, 1999, and 2006 (Oregon Climate Service, 2007). Precipitation in WY 1997 was the highest since recordkeeping began in 1911, and was about 21 in. more than average.

The individual measurements and general conditions for quantifying gains and losses to the stream were good overall; however, the measurements for the 2006 study caused considerable uncertainty because of multiple beaver dams between RM 17.2 and 1.2. The inability to measure the streamflow at several of the sites measured previously because of ponded conditions and the buildup and subsequent wash-out of these dams raised uncertainty about the stability of flows during the 2006 study.

The Johnson Creek basin was divided into three reaches for this analysis based on seepage characteristics. The upper reach extends from RM 17.2 to 10.2 (pl. 1). The slope of the stream in the upper reach is 20 ft/mi and the elevation of the channel decreases from 370 to 228 ft. The drainage basin area increases by 14.3 mi² in this reach (Tanner and Lee, 2004, p. 7). Much of the surface-water contributing area (the area in a drainage basin that contributes water to streamflow) of this reach consists of Troutdale Formation gravels and steeply sloped volcanic buttes. These materials are less permeable than the deposits north and west of this reach. The middle reach is from RM 10.2 to 5.5 (pl. 1). The slope of the stream channel in the middle reach is 8 ft/mi and the elevation decreases from 228 to 190 ft, as a result the stream channel of the middle reach is less steep than the upstream reach. The drainage basin area increases by 17.6 mi² in this reach. The contributing area to the middle reach on the southern side is similar to that of the upstream reach, except at the lower end of the reach, which is flat and composed of unconsolidated sediments. The contributing area on the northern side of the creek is relatively flat and consists of permeable Pleistocene flood deposits. The lower reach extends from RM 5.5 to 0.7 (pl. 1). The channel slope (35 ft/mi) is greater than that of both reaches upstream. Elevation decreases from 190 to 24 ft. The drainage basin area in this reach increases by 8.8 mi². The stream channel in this reach intersects a relatively large thickness of unconsolidated sediments of the terrace deposits, contributing to increased seepage. This reach includes Errol Spring and Crystal Springs Creeks and numerous other spring inflows.

Seepage measurements in the upper reach indicated relatively small gains during each study (fig. 4, table 2). All gains in the upper reach were greater than the measurement uncertainty of 5 percent. Streamflow in this reach was not measured in 1997. During the September 1988 seepage run, the flow at the upstream end of this reach was only 0.08 ft³/s. Because the drainage basin size upstream of this location is 12.5 mi², this relatively low streamflow is an indication of minimal groundwater contribution to streamflow. The least gain in streamflow along this reach was in 1988 (0.56 ft³/s). The greatest gain was in 2006 (1.82 ft³/s). The gains decreased from 1998 (1.71 ft³/s) to 2000 (1.28 ft³/s), but were still more than twice the gains recorded in 1988.

Seepage measurements in the middle reach indicated slight gains and losses; however, in 4 of the 6 years that measurements were made, the gains or losses were less than the measurement uncertainty. The only gain that was greater than the measurement uncertainty was in 1988 when streamflow gained 0.28 ft³/s. The only loss greater than the measurement uncertainty was in 2000 when the flow decreased by 0.46 ft³/s. These gains and losses are small considering the reach length and substantial increase in drainage basin area.

Gains in flow in the lower reach were much greater than in either of the two reaches upstream. The lower reach is characterized by diffuse seepage to the stream and by distinct spring inflows, the largest of which is Crystal Springs Creek. The least gain was in 2006 (11.5 ft³/s). The greatest gain was in 1997 (16.2 ft³/s), and gains generally decreased from 1997 to 2006. All gains in streamflow in this reach were greater than the measurement uncertainty. Streamflow in the lower reach indicated gains at each subsequent measurement location, except in some years between RM 2.2 and 1.2 (fig. 4). Losses between RM 2.2 and 1.2 may be attributed to recharge to groundwater through the streambed, as the stream gradient decreases and the stream channel intersects permeable deposits near the confluence with Crystal Springs Creek.

In summary, spring-flow and seepage measurements indicate that groundwater discharge to Johnson Creek is low in the upper and middle reaches, and high in the lower reach. In 1998, when spring flow (fig. 2) and seepage (fig. 4, table 2) were high, streamflow of Johnson Creek in the 11.7 mi (encompassing an increase in drainage basin size of 31.9 mi²) from RM 17.2 to 5.5 increased only 1.62 ft³/s. In contrast, in the relatively short (4.8 mi) distance and small (8.8 mi²) increase in drainage basin size from RM 5.5 to 0.7, even in 1988, a relatively dry year, the stream gained 12.5 ft³/s.

Long-Term Streamflow Data

Long-term records of daily streamflow were used to quantify groundwater discharge to the stream in two ways. First, analyses were made of differences in low flows of Johnson Creek each year at the Gresham, Sycamore, and Milwaukie streamflow sites (map numbers 26, 30, and 41, respectively). Second, base-flow separation techniques were used to estimate the groundwater discharge component of streamflow at the Sycamore and Milwaukie sites. Comparison of the recharge and groundwater discharge derived from base-flow separation provides insight into the spatial distribution of groundwater discharge in the Johnson Creek basin.

Similar to the results of seepage measurements, analysis of low-flow data at the streamflow sites indicates differences attributable to groundwater discharge. Although the seepage measurements provided detailed spatial definition of groundwater discharge to the creek in specific reaches, and only during the years the seepage measurements were made, analyses of low flows from the streamflow sites provided a year-by-year indication of seepage to the stream.

The difference in low flow is shown in fig. 6 between the Gresham and Sycamore sites (from 1998 to 2006), and between the Sycamore and Milwaukie sites (from 1989 to 2006). At each site, the minimum flow is during late summer. The analysis was based on the same 7-day period at each site each year and was indexed to the days of minimum streamflow at the Sycamore site. For example, the 7-day low streamflow in 2005 at the Sycamore site was from September 21 to 27, and was compared to the 7-day streamflow for those days at the Gresham and Milwaukie sites. Although the annual 7-day minimum streamflow at the Gresham and Milwaukie sites may not coincide with that of the Sycamore site, differences in streamflow between the common period and the annual 7-day low flow at each site were usually within the uncertainty associated with the streamflow records. For each year, the 7-day flow from the upstream site was subtracted from that of the downstream site, resulting in the gain during the low-flow time of each year. The relatively large (about 1 ft³/s) groundwater discharge from Gresham to Sycamore in 1998 through 2000 was followed by a decrease to less than 0.4 ft³/s beginning in 2001. In 2003, the stream lost about 0.3 ft³/s. Although only an indication of a loss, considering measurement uncertainties, the net groundwater discharge to this area of Johnson Creek during the summer of 2003 could be characterized as near zero. From 1989 to 2006, gains in flow from groundwater discharge to Johnson Creek between the Sycamore and Milwaukie sites were greater, ranging from 8.5 to 25.9 ft³/s. As discussed previously, most of the groundwater discharge is downstream of RM 5.5 and Crystal Springs is the largest contributor. The average summer gain in flow for 1996–99 was 22.4 ft³/s, compared to the 1989–95 average of 14.8 ft³/s and the 2000–2006 average of 12.2 ft³/s. Use of low-flow data provides a minimum estimate of seasonal groundwater discharge to Johnson Creek. Groundwater discharge during the remainder of the year is higher because of increased infiltration of precipitation (recharge) and resulting increases in discharge to the stream from the surrounding aquifer.

For years when seepage measurements were made (1997 to 2000, and 2006), differences in streamflow measured at the Gresham, Sycamore, and Milwaukie sites were similar to differences in streamflow derived by comparison of 7-day low flows. The seepage measurements were made at low flow, but not necessarily at the lowest flow of the year. For this reason, groundwater discharge from seepage was usually slightly greater than determined from the 7-day low flows.

Several inherent factors contribute to uncertainty in estimation of seepage to the stream from groundwater discharge using the low-flow data. Streamflow is computed using a lookup table, also known as a “stage-discharge rating.” (“Stage” is the stream level relative to an arbitrary datum.) At low flow, small changes in stream level result in relatively large changes in flow. Although streamflow measurements define the stage-discharge relation at a streamflow site, the relation can be imprecise and is affected by impermanent factors such as leaves, algae growth, and beaver dams. Similar to methods used in compiling seepage measurements, subtracting one small number from another also introduces uncertainty. Other factors that were not quantified include differences in evapotranspiration losses and possible water withdrawals. The lowest 7-day flow of the year at a given site additionally was not necessarily at the same time at all three sites because of the recession characteristics and differences in response to small rainfall events during the summer.

Base-flow separation also was used to quantify groundwater discharge to Johnson Creek, and to provide comparison between recharge estimates. This analysis is made using multidecade-scale, average streamflow values, and draws a comparison to long-term average recharge. Although recharge and discharge vary from year to year, these relatively short-term variations are dampened by the decade-scale analysis. Snyder and others (1994) estimated recharge for the period from 1949 to 1983. The streamflow records at the Sycamore site begin in WY 1941, and at the Milwaukie site in WY 1989. The base-flow component of streamflow at the Sycamore site was calculated for the period covered by the Snyder and others (1994) analysis, the period of record at the Sycamore site, and from WY 1990 to 2006, the period of concurrent record at the Milwaukie site. The difference in the average groundwater discharge between these periods was less than 10 percent. For comparison, the concurrent period of record of the Sycamore and Milwaukie sites (WY 1990–2006) was selected.

The average base-flow component of streamflow from WY 1990 to 2006 at the Sycamore site was about 20,500 acre-ft/yr and about 35,500 acre-ft/yr at the Milwaukie site. At each site, groundwater discharge (the base-flow component of streamflow) represents about 60 percent of estimated recharge. The remaining 14,000 acre-ft of recharge at the Sycamore site is assumed to leave the drainage basin upstream of the site by groundwater flow paths leading toward other points of discharge either downstream of the site or outside the Johnson Creek basin. The remaining 25,300 acre-ft at the Milwaukie site likely discharges to areas outside of the Johnson Creek basin. Possible discharge of water that recharges to the Johnson Creek basin to areas outside of the basin is discussed further in the section “Groundwater Flow Direction.”

Use of Environmental Tracers to Identify Groundwater Discharge

A qualitative assessment of seepage was based on observations of specific conductance and continuous records of stream temperature. In this context, specific conductance and temperature were used as tracers to identify the presence of groundwater discharge in a given stream reach, providing independent confirmation of trends identified by flow measurements (Winter and others, 1998; Stonestrom and Constanz, 2003).

The specific conductance of groundwater typically is higher than surface runoff because of prolonged contact with and dissolution of minerals in the subsurface environment. For example, the specific conductance of Johnson Creek during high flows is usually less than 100 µS/cm compared to specific conductance values of greater than 200 µS/cm in Crystal Springs Creek, a groundwater fed stream (Edwards, 1992, p. 28). Increasing specific conductance from one measurement location to the next downstream generally is an indicator of groundwater discharge to the stream.

Specific conductance measurements were made at most locations during the 1998, 1999, and 2006 seepage studies. Specific conductance generally increased less than 30 µS/cm from RM 17.2 to the lower end of the middle reach (RM 5.5) (fig. 4, table 2). In contrast, specific conductance increased more than 50 µS/cm between RM 5.5 and 2.2. These results were consistent with the small gains in flow observed upstream of RM 5.5 and the much greater gains between RM 5.5 and 2.2. The absence of change in specific conductance from RM 2.2 to 1.2 is consistent with the absence of gain in streamflow indicated by the streamflow measurements. A large gain in streamflow is observed without a large increase in specific conductance between RM 1.2 and 0.7 because the gain is a result of inflow from Crystal Springs Creek, and at low flow, the specific conductance of Johnson Creek is similar to that of Crystal Springs Creek.

Stream temperature is another indicator of presence or absence of groundwater discharge. Seepage measurements are usually made during the summer, when air temperature is higher than the stream temperature. Groundwater discharge to the stream is cooler than either the stream temperature or the summer air temperature. During the summer, the temperature of Johnson Creek fluctuates on a diurnal cycle, warming several degrees during the day and cooling at night. In contrast, groundwater temperature is much cooler and fluctuates little. For example, the temperature of two springs (spring identification numbers 12 and 15 as shown in McCarthy and Anderson, 1990, table 2) in the Crystal Springs area was a constant 12.6°C from May to August 2000, despite seasonably warm air temperatures.

Groundwater discharge is identified by an overall cooling of the stream in the summer and moderation of the magnitude of the diurnal fluctuation in temperature. In the absence of groundwater discharge to the stream in a given reach, the diurnal pattern of stream temperature is repeated at subsequent locations downstream, generally warming because of increased exposure of the stream to the warm air and to solar radiation along many of the reaches. Groundwater discharge to the stream is identified by comparing the temperature at the upstream and downstream end of a given reach.

Temperature recorders were installed in the stream at each streamflow location during the seepage studies of 1998 and 2000 for several days before and after the day of streamflow measurements. An example of the stream temperature is shown in figure 7 at three locations on Johnson Creek in 2000, and the results shown are similar to those of 1998. The stream temperature was fairly similar between RM 7.8 and 5.5, indicating negligible groundwater discharge to the stream, and substantiated by seepage measurement, which indicated little change in streamflow (fig. 4, table 2). In contrast, streamflow increased from RM 5.5 to 3.2 and was accompanied by an overall decrease in stream temperature.

Groundwater Flow Direction

An understanding of groundwater flow direction is useful in evaluating the source areas (referred to as the contributing areas) for features that may receive groundwater discharge such as streams, lakes, wetlands, springs, and wells. If the location and extent of the contributing areas for discharge features is identified then information can be provided about how natural and man-made changes in climate, water use, or land use, for example, can influence the quantity and quality of water that may discharge. Therefore, knowledge of the direction of groundwater flow can be used to improve our understanding of the hydrology of the Johnson Creek basin, which will facilitate effective management of the water resources.

The direction of groundwater flow in an aquifer is dependent on the hydraulic head, an indicator of the total energy available to move groundwater through an aquifer (Taylor, and Alley, 2002, p. 3). Hydraulic head differs from the water level which describes the position of the groundwater surface usually relative to some datum such as sea level. For the water table, the hydraulic head is equal to the water-level elevation. For a vertical position in the water column below the water table or under confined conditions, the hydraulic head is the sum of the elevation of the position and fluid pressure at that point (the contribution to the hydraulic head from kinetic energy as a result of flow typically is assumed to be negligible) (Freeze and Cherry, 1979, p. 18-22, 39; Taylor, and Alley, 2002, p. 3). The hydraulic head at some point in an aquifer, such as a narrow open interval for a well, typically is measured as the height (water level) to which a column of water will stand above some datum. Groundwater flows from areas of high hydraulic head to areas of low hydraulic head. The direction of groundwater flow is used to identify the presence of a groundwater flow divide (a ridge in the groundwater surface from which groundwater moves away in both directions perpendicular to the ridge line), which is useful in inferring the extent of contributing areas to groundwater-discharge features such as streams and springs. However, groundwater moves in complex, three dimensional patterns that change with time and consist of vertical and horizontal components of flow (Franke and others, 1998, p. 12). Water-level maps are useful in determining the approximate horizontal direction of groundwater flow but should be used with caution and the knowledge that a vertical component of flow also is present and that the direction of lateral flow may vary with depth below the water surface.

Groundwater flow directions in the Johnson Creek basin were inferred from water-level maps determined directly from measurement of groundwater elevations (McFarland and Morgan, 1996) and groundwater flow modeling that simulates the physical properties of the groundwater flow system and the stresses applied to the system to calculate hydraulic heads (Morgan and McFarland, 1996). A USGS study of the depth to groundwater and configuration of the water table in the Portland metropolitan area by Snyder (2008) resulted in a map of the estimated water-table elevation. The water-table map utilized much of the groundwater data collected as part of this study and can be used to infer the direction of groundwater flow on a local scale (Snyder, 2008, p. 5). The results of the analysis by Snyder (2008, p. 28, pl. 2) (see fig. 8) and analyses by McFarland and Morgan (1996, pls. 2-3) and Morgan and McFarland (1996, pls. 3-4) indicate that in many areas of the Johnson Creek basin, the direction of groundwater flow is not toward Johnson Creek, but instead is out of the drainage basin towards the Sandy, Columbia, or Willamette Rivers. Upstream of about RM 21, the direction of groundwater flow in the Johnson Creek basin generally is northward to the Sandy River. Between RM 21 and 15, the direction of groundwater flow is towards Johnson Creek. Between RM 15 and 3, groundwater flow may be towards Johnson Creek in the immediate vicinity of the creek. However, between RM 15 and 3, groundwater may flow north to the Columbia River in the eastern reach, with flow becoming more westerly towards the Willamette River in the western part of this area. Downstream of RM 3, in the lower part of the Johnson Creek drainage basin, the direction of groundwater flow at the water table is more consistently toward Johnson Creek, indicating local groundwater discharge.

The greater relative streamflow observed in the lower reaches of the Johnson Creek basin compared to the upper and middle reaches is likely explained by discharge from intermediate and regional groundwater flow systems in these areas. The general direction of shallow and intermediate depth groundwater flow in the middle and eastern areas of the Johnson Creek basin (containing the middle and upper reaches of Johnson Creek) generally is toward the Columbia River north or northwest, away from Johnson Creek (fig. 8; McFarland and Morgan, 1996, pls. 2-3; Snyder, 2008, pl. 2). Vertical hydraulic gradients between the water table and the underlying hydrogeologic unit and between the water table and deeper hydrogeologic units as simulated by Morgan and McFarland (1996, p. 32-35) additionally indicate downward movement of water throughout most of the eastern area of the Johnson Creek basin. In the middle areas of the basin, the vertical gradients are mixed. In the western area of the Johnson Creek basin, the general direction of shallow and intermediate depth groundwater flow is northwest changing to west towards the Willamette River and the mouth of Johnson Creek in the westernmost area of the basin. Vertical hydraulic gradients in the westernmost area of the basin indicate upward gradients that may denote the discharge of intermediate and regional groundwater flow to the lower reaches of Johnson Creek, tributaries and springs in the area, or to the Willamette River.

The apparent absence of groundwater discharge to some reaches of Johnson Creek is further indicated by a particle-tracking analysis (the simulation of hypothetical particles of water through the groundwater flow system using the results of a groundwater flow model) by Hinkle and Snyder (1997). The results of Hinkle and Snyder (1997) indicate that groundwater that recharges in areas of the Boring Hills flows vertically down and north beneath Johnson Creek through the regional groundwater flow system and likely discharges to the Columbia River (Hinkle and Snyder, 1997, p. 20 and pl. 1 A, B, and C).

These observations and analyses indicate that the surface-water divides delineating the Johnson Creek drainage basin are not necessarily coincident with groundwater divides particularly along the northern boundary of the basin. The surface-water divides along the southern boundary of the basin generally are coincident with the groundwater divides with possible exceptions in the vicinity of Happy and Pleasant Valleys, where areas of the valleys may be within the groundwater contributing area of the Johnson Creek basin although they are within the surface-water basins for Mt. Scott and Rock Creeks, respectively. As a result, recharge to the groundwater system within some areas of the Johnson Creek surface-water drainage basin may not discharge to Johnson Creek and instead may flow and discharge to regional groundwater discharge areas such as the Columbia and Willamette Rivers. Discharge of groundwater to areas outside the Johnson Creek basin appears to be more prevalent in the eastern and northern areas of the basin. This conclusion is further supported by the observations based on seepage measurements and streamflow data that indicate relatively small gains to Johnson Creek in the upper and middle basin compared to the lower basin.

Variation in Groundwater Levels and Storage

Understanding the factors that affect the location, timing, and magnitude of variations in groundwater levels and storage will aid managers in development of guidelines and regulations to protect the groundwater and surface-water resources of the Johnson Creek basin. Water-level fluctuations in wells represent variations in groundwater storage and are a result of many factors, including aquifer properties, the rates of recharge or discharge, the direction and magnitude of groundwater flow, and the construction of the well. In the Johnson Creek basin, the fluctuation of groundwater levels and storage is of particular concern because of the influence of these factors on the magnitude and timing of spring discharge, streamflow, and the surface expression of the water table in low-lying areas.

Changes in the rates of recharge or discharge cause changes in groundwater storage, which are represented by water-table fluctuations. The water table rises because of increased groundwater storage when the rate of recharge exceeds the rate of discharge and declines when these conditions are reversed (Veeger and Johnston, 1996, p. 28). The water table rises in response to increased groundwater storage that can result from recharge because of precipitation, losing streams, stormwater runoff into UIC systems, infiltration from septic systems and excess irrigation, or increased inflow of groundwater from adjacent areas. The water table declines in response to decreased groundwater storage that can result from discharge to gaining streams, springs, pumpage, evapotranspiration, or the outflow of groundwater to adjacent areas. In the Portland area, the quantity and timing of precipitation typically exerts the greatest influence on water-level fluctuations, and because of the seasonal nature of precipitation, a seasonal response in water-table level results (Snyder, 2008, p. 24-25, 29-30). Groundwater levels rise following precipitation of sufficient intensity and duration to exceed evapotranspiration and soil moisture deficits and the residual infiltration results in recharge. Water levels decline as groundwater is removed from storage during periods when discharge to rivers, streams, springs, pumping wells or evapotranspiration, is greater than recharge. Time scales for water-table fluctuations range from hours (in the response to high intensity precipitation events or changes in stream level) to years or decades (in response to long-term changes in climate or land-use practices).

Well construction also can influence water-level fluctuations in wells. The well depth, the open-interval length and its proximity to the water table, and if the well is open to a water-table (unconfined) aquifer or confined aquifer can all contribute to water-level fluctuations in wells.

Changes in groundwater storage indicated by fluctuations in groundwater levels are of special interest in two areas within the Johnson Creek basin: (1) the groundwater contributing area to Crystal Springs (which provides the flow to Crystal Springs Creek (map number 40) that consequently provides most of the late summer flow to Johnson Creek downstream of the confluence with Crystal Springs Creek), and (2) the vicinity of Holgate Lake (map number 43) that is subject to flooding when groundwater levels exceed land-surface elevation in low-lying areas.

The surface-water contributing area to Crystal Springs Creek is insufficient to supply the observed discharge based on estimates of groundwater recharge (Dames & Moore, 1998, p. 59). The groundwater contributing area for the flow to Crystal Springs, which supplies most of the discharge to Crystal Springs Creek, may extend to areas well beyond the surface-water contributing area of the Crystal Springs Creek basin perhaps as far east as Powell Butte according to observations of regional and local groundwater flow directions (McFarland and Morgan, 1996, pls. 2-3; Snyder, 2008, pl. 2). However, an accurate delineation of the groundwater contributing area for Crystal Springs will require the use of an updated three-dimensional groundwater flow model with sufficient discretization and data to adequately represent the springs and gaining and losing reaches of Johnson Creek. The change in groundwater storage in this area may have a direct relation to the timing and quantity of streamflow of Crystal Springs Creek.

Analysis of the groundwater flow system was focused on the area extending from Powell Butte westward which includes the areas of Crystal Springs Creek and Holgate Lake, two concerns to the community. A monitoring network of wells was established in 1998 to increase understanding of the groundwater flow system in the area. Initially, the wells were measured monthly, changing to bimonthly in 2000, and to quarterly beginning in 2005. Hydrographs from these wells were used to analyze the spatial and temporal variations in groundwater levels in the western area of the Johnson Creek basin (fig. 9; pl. 1; table 1). In 1998 a continuous water-level recorder with real-time telemetry was installed near Holgate Lake (map number 9). In 2001, the city of Portland completed construction of seven monitoring wells between I-205 and Westmoreland Park. Continuous water-level recorders were installed in five of these wells (map numbers 19, 20, 21, 23, and 24). The records from the continuous water-level recorders at Holgate Lake and between I-205 and Westmoreland Park provide the most detailed information on the magnitude and timing of groundwater fluctuations in the area west of Powell Butte (figs. 10 and 11). This information was used to evaluate how groundwater storage changes from day to day, season to season, or year to year in response to precipitation events, seasonal variations, or long-term trends, respectively.

Water levels in the four easternmost wells with continuous water-level recorders (fig. 10) fluctuated in response to precipitation (fig. 11). The general pattern of water level in these four wells is a sinusoidal fluctuation that varies seasonally. From east to west, the magnitude of the response is increasingly attenuated and the time between precipitation and the corresponding rise in water level is increased. The annual water-level fluctuation (fig. 10) ranges from as much as 10 ft in the easternmost well (map number 9) to less than a foot in the westernmost well of this group (map number 21). In a given year, the water level in the easternmost well responds quickly to the autumn rains, rising rapidly following distinct rainy periods, and usually peaks in March or April, coincident with the end of the rainy period and the beginning of substantial uptake of water by plants and evaporation. The rise in water levels illustrated by this set of the four easternmost wells with continuous recorders is increasingly muted and delayed with distance westward (from east to west: map numbers 9, 19, 20, and 21), such that the response from autumn rains in the westernmost well of these wells does not usually begin until March or April of the following year, a 6-month difference. In 2001 and 2005, precipitation was decreased or fell later than normal (fig. 11) and seems to have contributed less to groundwater recharge due to decreased infiltration or uptake by soil moisture storage and evapotranspiration compared to other years as indicated by a lower than normal rise in the groundwater level at the easternmost well (map number 9), and little or no increase in groundwater levels in the other wells shown in figure 10, until the following season’s recharge from precipitation. The increase in the attenuation and delay of the rise in the level of groundwater in response to the autumn and winter precipitation in an east-to-west direction may be attributed to the characteristics of the unsaturated zone, such as thickness and composition. The unsaturated zone thickness is one of the factors affecting the travel time of precipitation from the land surface to the water table and greater thickness contributes to longer and more diverse travel times. The median daily thickness of the unsaturated zone is about 44 ft in the easternmost well, and is about 120, 131, and 100 ft in the other three wells from east to west, respectively (fig. 10). Unsaturated zone travel time also is a function of several factors, including the composition of the unsaturated zone including rock type, particle size/texture (clay, silt, sand, and gravel), sorting, consolidation, cementation, homogeneity, and moisture content. The unsaturated zone materials consist primarily of unconsolidated deposits from the Missoula Floods. These deposits tend to be finer grained to the west as distance increases from the western end of the Columbia River Gorge (Trimble, 1963, p. 63; Hogenson and Foxworthy, 1965, p. 26). However, these materials can be variable and discontinuous vertically and laterally (Trimble, 1963, p. 58–62). The presence of poorly sorted finer grained materials may further increase the time and diversity of travel times through the unsaturated zone. Differences in effective porosity of these materials also could result in differences in the magnitude of the response to precipitation events, such that greater effective porosity results in greater storage available within the aquifer and a smaller change in the water-table position resulting from a change in a given volume of water when compared to an aquifer with a lesser effective porosity (Snyder, 2008, p. 24-25, 29-30).

Alternatively or additionally, there may be a groundwater wave (hydraulic pulse) in the saturated zone moving from east to west when recharge enters the groundwater flow system and moves laterally. A groundwater wave is defined as “a high in the water table that moves laterally, with a wavelike motion, away from a place where a substantial quantity of water has been added to the saturated zone within a brief period” (Wilson and Moore, 2003, p. 97). A groundwater wave originating in the area on the west side of Powell Butte and moving westward could explain the pattern of water-level fluctuations in the wells from Powell Butte extending west to the area just east of Westmoreland Park. The pattern displays increased attenuation and delay of the rise in the level of groundwater as distance increases from the source area. A groundwater wave is an expression of a hydraulic (pressure) pulse that can move more rapidly through an aquifer when compared to the actual movement of groundwater, which would be expected to be much slower. The existence of a groundwater wave may be resolved by further analysis of existing or new groundwater level data in this area.

Understanding of the pattern of water-level fluctuations in the two Westmoreland Park wells is needed to assess the possibility of a relation of groundwater levels in this area to the level and streamflow of Crystal Springs Creek in the Westmoreland Park area. The patterns of water-level fluctuations in the two wells in Westmoreland Park (a shallow well (map number 23), and a deeper well (map number 24); fig. 11) are fairly similar, but differ greatly when compared to the four continuous-recorder wells toward the east (fig. 10). The different trends in water levels are a result of two factors. First, the Westmoreland wells are completed in permeable deposits of a former stream channel that could be either of the Clackamas River or of the Willamette River (Hogenson and Foxworthy, 1965, p. 10-11, 28). The channel facies deposits consist of interlayered silts, sands, and gravels, which in some places are locally covered by bog or pond sediments (Beeson and others, 1989). These deposits may exert a strong influence on the response of water levels within the groundwater system. Second, proximity to Johnson Creek and the Willamette River indicates that changes in the hydraulic head of these streams (as represented by their surface-water elevations) may be transmitted through the permeable channel facies deposits and affect water levels in the Westmoreland Park wells as indicated from the hydrograph characteristics for the two wells (see discussion below). Water levels in the two Westmoreland Park wells are relatively close to land surface, where depth to water from 2001 to 2006 was about 3 to 12 ft below land surface.

The water level in both Westmoreland Park wells rises quickly in response to precipitation, and also rises and declines on an annual cycle (fig. 11). Following the end of a given precipitation event, the water level in these wells declines rapidly. This rapid decline in water level in the Westmoreland Park wells is unlike the wells to the east, where the effect of precipitation typically is cumulative within a given year, and water levels begin to decline slowly in response to decreased precipitation intensity and groundwater flow towards discharge areas (fig. 10). Of the two wells in Westmoreland Park, the shallow well (map number 23) responds slightly more quickly, and the rise generally is of greater magnitude than the deep well (map number 24). Response time of both wells to precipitation is within 1 to several days. Most of the water-level rise associated with a typical several-day winter storm with 2 to 3 in. of rainfall is often about 2 ft in the shallow well and about 1 ft in the deep well. The short-term rise and decline in water level because of discrete rainfall events is superimposed on an annual, relatively steep rise in water level in autumn and winter and an attenuated recession during the spring and summer periods. Water levels in these two adjacent wells at Westmoreland Park were compared to precipitation at the Portland Airport (NWS site number 356751) (Oregon Climate Service, 2007). Weekly precipitation totals show a moderate correlation with the weekly change in groundwater levels at the shallow and deep well, with a correlation coefficient (R) of 0.60 and 0.65, respectively. Although rises in water level in both wells are on a several-day time scale, these rises are followed by declines on the same time scale. These declines in water level may account for the relative weakness in this correlation. Water levels in these wells show stronger correlations with stream levels in Johnson Creek and the Willamette River. The hydrographs of the two Westmoreland Park wells bear a striking similarity to the river-level hydrographs of Johnson Creek and the Willamette River (fig. 11) —similar shape and distinct storm and annual sinusoidal responses. Daily water levels in the shallow well (map number 23) are closely correlated (R = 0.85) to the daily mean stream level at the Johnson Creek at Milwaukie site (map number 41), and are at least in part affected by tidally induced fluctuations of the Willamette River. Although the well is about 1.5 mi north of the Johnson Creek stream monitoring site, the closest approach is only about 0.7 mi north of Johnson Creek. There seems to be a subsurface hydraulic connection between the shallow well and Johnson Creek because the lag time between a change in stream level and a corresponding change in the water level at the well is less than 1 day as determined using a cross-correlation analysis. At about RM 2.0, where Johnson Creek flows out of an incised canyon, the streambed consists of recent alluvial deposits that overlie the channel facies deposits at the mouth of the canyon (Beeson and others, 1989). At this location, the elevation of Johnson Creek is several feet above the elevation of the water table in the low-lying drainage basin of Crystal Springs Creek. Changes in stage in Johnson Creek likely are transmitted through the shallow, semi-confined channel facies deposits, to the Westmoreland Park wells. Potential recharge to the aquifer from Johnson Creek has been indicated by seepage measurements, which often indicate small losses along this reach to just upstream of the confluence with Crystal Springs Creek (fig. 4, table 2).

Backwater from tidal fluctuations of the Willamette River, about 0.8 mi to the west, also affects the water level in the shallow Westmoreland Park well. Backwater is the effect on the water-surface elevation at a given stream location from fluctuations of a downstream water body. The Willamette River fluctuations of several feet per day (as observed at the site in downtown Portland (map number 44) about 3 river miles downstream) are transmitted to the shallow well through permeable channel deposits, resulting in a fluctuation of a few hundredths of a foot, daily and sometimes twice daily.

The daily mean water level in the deep well (map number 24) at Westmoreland Park also exhibits a strong correlation with the daily mean level of the Willamette River. The lag times in response to changes in level of the Willamette River and subsequent changes in the water level in the deep well at Westmoreland Park are short, perhaps less than 1 day as determined through cross-correlation. This strong correlation (R = 0.81) and short lag time may be the result of a hydraulic connection between the Willamette River and the channel facies deposits within the former river channel that extend to the Willamette River and to which the deep well is open. As indicated above, a minor relation is apparent in the connection of tidal fluctuation in the Willamette River to the shallow well in Westmoreland Park. The deep well responds more reliably to the twice-daily tidal fluctuations, where two highs and two lows during many days affect the water level in this well. Similar to the shallow well, the amplitude of these fluctuations is several hundredths of a foot. The difference in the tidal effect observed in the shallow and deep wells may be attributed to a coarsening of subsurface materials at depth. The driller’s logs of these two wells indicated a silty layer about 20 ft below land surface near the open interval of the shallow well, which may inhibit movement of water from the otherwise coarse alluvial deposits to the shallow well.

The tidal influence and backwater from the Columbia River affect the Willamette River and these fluctuations also affect the water level in the Westmoreland Park deep well (fig. 11). Although the high-flow period of the Willamette River (and associated high stream level) usually is in winter, the spring melt in the headwaters of the Columbia River basin (several hundred miles to the northeast) in May and June of each year raises the Columbia River in the Portland area, backing water up in the Willamette River in the Portland area. The backwater creates a “hump” in the otherwise (usually) smooth recession of the Willamette River at Portland (map number 44). This rise is not associated with precipitation in the area and provides clear differentiation between precipitation and river-level induced changes in water level in the Westmoreland Park deep well. Evaporative losses additionally are high during late spring and effectively reduce recharge from precipitation in the May–June time period. The backwater effect is not the same each year, and depends primarily on the magnitude and duration of high flow of the Columbia River. For example, the presence of backwater was almost nonexistent in 2001, when May and June flows of the Columbia River were at a near-record low, but was present in 2006, an average flow year for the Columbia River.

Response of the Groundwater System to Precipitation Extremes

The response of the groundwater system to extended periods of either unusually high or low precipitation is manifested by changes in groundwater levels and discharge torivers, streams, and springs. The period of study, 1997–2006, encompassed relative extremes in annual precipitation which were accompanied by changes in groundwater levels, streamflow, and spring discharge. The analysis of data collected during this period enabled the development of relations to forecast the effect of these precipitation extremes on groundwater levels and the resulting effect on surface-water features. These relations can be used to anticipate changes in discharge of spring-fed streams or to estimate the expected inundation of low-lying areas as a result of groundwater levels rising above land surface.

High and Low Groundwater Discharge: Crystal Springs Creek

Crystal Springs Creek receives flow from several springs, flows in a channel through park and residential areas of Southeast Portland, and enters Johnson Creek at RM 1.0. Historically, increased discharge of these springs has raised concerns about the stream level and potential flooding of areas adjacent to Crystal Springs Creek. Flooding was reported in park and residential areas adjacent to the creek in July 1997 (Chestnut, 1997). In 2005, low flow of Crystal Springs Creek in the Reed College area prompted concern about fish survival under these conditions.

The channel of Crystal Springs Creek has been modified. Although the creek is a natural feature, the mostly predevelopment landscape, as mapped by the General Land Office in 1852, consisted of a distinct stream channel in the upper area near the present location of Reed College campus, tributaries corresponding to the location of several springs in 2008, and an extensive wetland area covering much of what is now Westmoreland Park (University of Oregon, 2006). The wetland narrowed to a single stream channel near the confluence with Johnson Creek. The creek was channelized through much of its length in 1936 during construction of Westmoreland Park. The present configuration is an approximately 10-ft-wide natural bottom, concrete-walled channel, where the creek flows nearly bankfull year-round. The stream channel opens into a pond in Westmoreland Park and narrows again toward the mouth of the creek.

Flooding along Crystal Springs Creek was first noticed in the summer of 1997 (Chestnut, 1997; Dillin, 1997). The creek overtopped the banks and flowed into park and recreational areas, onto streets, and into basements of creekside properties. The cause of the flooding was increased discharge from springs and decreased capacity of the channel. Reductions in the channel capacity between 1996 and 1997 probably was a factor because the streamflow of Crystal Springs Creek was nearly as high in 1996 when no flooding was reported as it was in 1997 when flooding was reported (fig. 2). An assessment of the channel capacity, specifically the effect of aquatic growth or sediment in the stream channel was beyond the scope of this study. Removal of aquatic weeds in August 1997 (when the streamflow was still relatively high) lowered the level of Crystal Springs Creek in the Westmoreland Park area by about 1.5 ft, which caused a near-bankfull rather than an overbank condition (Sellwood Bee, 1997).

The cumulative impact of consecutive years of higher-than-average precipitation and the resulting increase in groundwater discharge was the primary factor contributing to the high flow of Crystal Springs Creek. Precipitation records indicate that Crystal Springs Creek probably did not have an increased streamflow such as recorded from 1996 to 1999 since the channel was modified in 1936, and possibly not since 1911. The near-record precipitation in WY 1996 (54.7 in.) was followed by the highest annual precipitation on record in WY 1997 of 58.7 in. (Oregon Climate Service, 2007) (fig. 5) The 3-year moving average annual precipitation depicts the unprecedented surge in groundwater recharge and subsequent discharge, and the multiyear recession observed in the late 1990s. The 3-year moving average precipitation in WY 1997, 1998, and 1999 was several inches greater than any year after recordkeeping began. By WY 2000, the 3-year moving average annual precipitation was close to the long-term average, and flow on Crystal Springs Creek had receded as well.

The elevated level of Crystal Springs Creek in the Westmoreland Park area probably is not associated with the rise of local groundwater levels and instead is the result of increased discharge of the numerous springs upstream of the Westmoreland area. Comparison of the hydrographs for the shallow and deep wells at Westmoreland Park (fig. 11) indicates a consistent downward gradient (where the head in the deeper well is less than the head in the shallow well) in this area. The absence of fluctuation in level of Crystal Springs Creek in the Westmoreland Park area (compared to 1 to 3 ft fluctuations in the shallow water table) and concurrent measurements along Crystal Springs Creek that showed little gain or loss indicates an absence of connection with the aquifer. A survey of measuring point elevations for the wells and along Crystal Springs Creek would help determine if the stream lies above the water table and is therefore disconnected from the groundwater system.

Low flow of Crystal Springs Creek in the Reed College area (map number 38) in July 2005 prompted curtailment of irrigation withdrawals. The low flow in this area was confirmed by streamflow measurements made in July, September, and October 2005 (fig. 2). The average of the three measurements made in 2005 was about 0.3 ft³/s, less than one-third of the flow measured in the previous 3 years, and less than 5 percent of the highest flow measured in 1997. The likely cause of decreased spring discharge was a decline in water levels and therefore groundwater storage of the aquifer upgradient of the springs. The monitoring well at Berkeley Park (map number 21), 0.8 mi southeast of the streamflow measurement location at Reed College, indicated there was little or no recharge in 2005, in contrast to previous years (2002 through 2004) that indicated an annual fluctuation in response to precipitation (fig. 10). Although 2005 was not a particularly dry year, precipitation came late in the season. In this condition, recharge for a given amount of precipitation may be less because of increased losses to evapotranspiration and soil moisture storage.

Use of Hydrologic Data to Predict Streamflow of Crystal Springs Creek

Because flow in Crystal Springs Creek is predominantly from the springs, and the spring flow is related to groundwater levels, flow of Crystal Springs Creek can be predicted using groundwater levels. The high flows of Crystal Springs Creek from 1996 to 1999 were the result of the prolonged effect of consecutive years of high precipitation in 1996 and 1997. Precipitation in 2005 was decreased or fell later (fig. 11), contributing little to groundwater recharge, which may have led to unusually low flows that summer. The groundwater level in the terrace just east of the headwaters of Crystal Springs Creek, represented by the Berkeley Park well (map number 21), fluctuates in response to precipitation (recharge), which is reflected in the subsequent discharge to the creek from the springs.

The Berkeley Park well has a continuous water-level recorder providing a useful data record for comparison with the streamflow of Crystal Springs Creek. However, in order to describe the relation between the streamflow of Crystal Springs Creek and groundwater levels, an intermediate step was required to estimate what the groundwater level was in the Berkeley Park area before the well was drilled in 2001. The water level in the Berkeley Park well relates closely to a well nearby that has a longer record of periodic measurements (map number 22) (pl. 1; fig. 9; table 1). Concurrent water-level measurements at these two wells from 2001 to 2006 were used in a regression to estimate the water level at the Berkeley Park well from October 1998 through August 2000, the period prior to the installation of the well (fig. 12). The extent of the data and the linear regression is scaled in fig. 12 to show the extent of extrapolation. Further data collection at higher water levels than those collected during the period of study may provide a better understanding of this relation.

The flow of Crystal Springs Creek at the Reed College site and at the mouth of the creek can be predicted based on groundwater levels at the Berkeley Park well (fig. 13). The relation of water level in the Berkeley Park well and flow of Crystal Springs Creek is valid over the range of concurrent water-level measurements and estimates, and streamflow measurements. The water level in the well ranged from 92.3 to 103.9 ft. The streamflow of Crystal Springs Creek encompassed by this relation was from 0.2 to 5.9 ft³/s and from 9.2 to 19.1 ft³/s for the locations at Reed College and at the mouth of the creek, respectively.

As discussed previously, groundwater level in the area of the Crystal Springs Creek basin in Westmoreland Park is affected by the stream level of Johnson Creek and Willamette River. Although the summer stream level of Johnson Creek varies little from year to year, the Willamette River was particularly high in the summer of 1997, which was coincident with the high flow of Crystal Springs Creek. Because the wells in Westmoreland Park were not drilled until 2001, the relation of nearby river levels to groundwater level and streamflow in 1997 could not be explored. During June 1997, the stream level of the Willamette River was about 16 ft, about 6 ft higher than during the same period from 1998 to 2006. Previous discussion referred to an absence of connection between Crystal Springs Creek in the Westmoreland Park area and the shallow aquifer. Continued monitoring of nearby stream levels and groundwater levels in wells in Westmoreland Park may provide further indications of stream–aquifer relations in this area.

Flooding from High Groundwater Level–Holgate Lake

Holgate Lake (map number 43) is an ephemeral lake in Southeast Portland near 136th Avenue and Holgate Boulevard (pl. 1). The origin of the lake and recurrent flooding and damage to nearby properties has been the subject of concern for local residents and municipal officials, prompting focused study in this area. Although the history of residential development of this area is uncertain, the first homes were probably built in the early 20th century. Until then, land use was primarily small farms. The lake is in a topographic depression in a residential area, and has flooded residences numerous times over the past several decades. Based on the present configuration of streets and residential development in the area, flooding begins whenever the lake level exceeds about 191 ft. Analysis of nearby groundwater level data and precipitation data has led to the conclusion that the occurrence of the lake is the result of groundwater flooding. Groundwater flooding begins in low-lying areas whenever the water table rises above the land surface (Jones and others, 2000). Precipitation and groundwater level data may be used to anticipate the extent of flooding in the low-lying area that has been designated as “Holgate Lake.”

The first record of flooding in the area now known as Holgate Lake was in March 1949, indicating the lake had risen 6 ft in 1 week, flooding several homes on the south side of the lake. The water-surface elevation of the lake was inferred from a photograph, indicating an elevation of between 190 and 195 ft. The cause was attributed to seepage from the surrounding slopes. The water level observed in 1949 was the highest since about 1943 (Gresham Outlook, 1949).

The next reports of flooding were in 1969. The elevation of the lake was about 199 ft, covered an area of about 50 acres (although the present interpretation of the flooded area is about 30 acres), and the western extent of the lake was near 128th Avenue. These observations were made in January 1969 and included a groundwater level measurement and reference to the land-surface elevation at a well near 122nd Avenue and Holgate Boulevard. The emergence of the lake was attributed to a rise in the water table (William S. Bartholomew, State of Oregon, written commun., 1969). A local newspaper attributed the flooding to unusually high water table and springs, following the heaviest precipitation since 1881. Flooding was reported by residents in 1943, 1948, 1956, and 1964 (Gresham Outlook, 1969).

Flooding from Holgate Lake began again in February 1996, lasting at least a month. The city of Portland pumped about 5 Mgal/d for a month from the lake to the sewer system, resulting in no decrease of the lake level. Flooding began again in December 1996, lasting at least until February 1997 (Oregonian, 1997). An aerial photograph of Holgate Lake in January 1997 was used to estimate the elevation of the lake at 195 ft (Jennifer Antak, city of Portland, written commun., 2008).

Observations of the level of Holgate Lake by the USGS began in February 1999 and a water-level recorder was installed in March of that year (fig. 14). The recorder operated through 2000 and periodic water-level observations were made until May 2006. Except for the absence of the lake in 2001 and 2005, the typical pattern observed during the period of study was emergence of the lake in early winter each year and a return to a dry lakebed by midsummer.

The maximum lake level observed from WY 1999 to 2006 was 192.3 ft on March 5, 1999, corresponding to a depth of about 9 ft at the deepest part of the lake. The lake again extended over the street on its south side. The surface area of the lake was 1.2 acres on April 13, 1999 when the lake level was 190.3 ft (about 2 ft lower than the maximum level observed the previous month). The lake perimeter was defined using a Global Positioning System (GPS), and then mapping software was used to calculate the area. At the time of maximum level on March 5, 1999, the lake area was estimated to be about 3 acres.

Groundwater level monitoring near Holgate Lake began in July 1998 (fig. 14). The primary observation well (map number 9) is about 0.2 mi southeast of the lake (table 1). A recorder (with telephone telemetry) provided real-time water-level data. In 1999, a hand-driven piezometer (map number 11) was temporarily placed in the lake bed to monitor the water level beneath the lake (fig. 14). In April 2000, a well was drilled on the western side of the lake (map number 10), replacing the piezometer, which was inaccessible much of each year.

The fluctuation of water level in the primary observation well closely matches that of Holgate Lake (fig. 14). The lake level and the groundwater level in the wells in and on the western side of the lake agree closely. When water levels rise each winter, the lake level and groundwater level are similar. When groundwater and lake levels decline in spring and summer, the lake level declines more rapidly than the level in the primary observation well. By early winter, prior to the beginning of the recharge period, the groundwater level beneath the lakebed can be as much as 2 ft lower than at the primary observation well.

Use of Hydrologic Data to Anticipate Flooding from Holgate Lake

Observations dating back to 1949 indicated that the cause of flooding of Holgate Lake was closely related to the water level in the shallow aquifer near the lake. This study confirms this hypothesis by comparison of lake level to water levels in nearby wells. The maximum level of Holgate Lake corresponds closely to the maximum groundwater level in the observation well (map number 9) nearby, and the groundwater level generally follows trends in cumulative precipitation each year (fig. 14). The primary observation well, equipped with telemetry, provides the water-table elevation to the public through the world-wide web at http://waterdata.usgs.gov/nwis/dv/?site_no=452912122312801&agency_cd=USGS&referred_module=sw. The groundwater data from this well and precipitation data can be used in a general way to anticipate the emergence of the lake and the potential for flooding. Heavy precipitation, particularly in the autumn and early winter, leads to a rapid rise in groundwater levels, which in turn leads to emergence of Holgate Lake. Precipitation after about March and into the summer has less effect than early season precipitation, probably because of the increased evaporation, uptake by vegetation, and soil-moisture storage. For example, in WY 2001 and 2005, less than 20 in. of precipitation fell between October and March, resulting in no emergence of Holgate Lake. In 2005, substantial precipitation came later in the spring and summer, totaling about 30 in. for the water year; however, losses, probably because of evapotranspiration and soil-moisture uptake, reduced recharge to the shallow aquifer and resulted in no emergence of the lake.

The water level of Holgate Lake corresponds to precipitation amounts and timing, when, after satisfying a soil-moisture storage threshold, groundwater level rises several feet for each foot of precipitation (figs. 14 and 15). The annual groundwater level rise from WY 1999 to 2006 ranged from 1.8 ft (2001) to 9.6 ft (2003). Although the net rise observed in 2003 (but no flooding) was greater than observed in 1999 (and flooding present), the antecedent groundwater level prior to the autumn rainy period was 4.6 ft higher in 1999 than in 2003. The higher antecedent groundwater level and higher precipitation in 1999 than in 2003 contributed to the 1999 flooding. Although no lake-level measurements were made at the beginning of WY 1999, the groundwater level at the site of the lake likely was near the lakebed, based on the assumption of relatively high groundwater levels in the area in the years following the high precipitation of 1996 and 1997. As indicated by the scatter of data points in figure 15, the response of groundwater level to precipitation is not predictable with certainty. The uncertainty may be a result of other factors not easily measured, including antecedent soil moisture, variations in porosity of the unsaturated zone at different depths, evapotranspiration, air temperature, precipitation timing and intensity, and basin-scale groundwater flow. Despite these uncertainties, the relation of water-level rise to precipitation may be useful to local residents and municipal agencies to assess flooding risks in the Holgate Lake area.