The remaining populations of the endangered dwarf wedgemussel (DWM)
(
The sediment–water interface is an important ecotone that harbors many
organisms evolved to live in this dynamic environment. Areas where
groundwater discharge to rivers is focused commonly are far less dynamic,
particularly with regard to temperature and sediment saturation, and some
organisms rely on this stability to survive (Hayashi and Rosenberry, 2002;
Smith, 2005). Such may be the case for the endangered dwarf wedgemussel
(
Previous studies investigated whether the riverbed in these locations would
become dry during low-flow events. Although low-flow conditions dewatered
much of the riverbed, areas populated by DWM remained wetted as long as river
discharge exceeded 15.8 m
Quantification of exchange between groundwater and surface water is particularly difficult in coarse-grained fluvial settings (e.g., González-Pinzón et al., 2015) due to spatial and temporal heterogeneity and multiple scales of flow that complicate distinction between hyporheic exchange and larger-scale groundwater discharge. Instruments are difficult to install where cobbles and boulders are present at and beneath the bed. Hydraulic gradients commonly are very small and difficult to resolve. Water chemistry is often a good method for distinguishing groundwater from surface water but, unfortunately, the chemistry of groundwater and surface water in the upper reaches of the Delaware River were found to be virtually identical.
Fortunately, use of several substantially different methods can minimize uncertainty and provide redundancy where some installations are difficult to impossible or when results based on a single method are inconclusive. The hydraulic head in water-table wells near the riverbank can be compared to river-surface elevations to evaluate the potential for lateral groundwater discharge on a reach scale. Electromagnetic-induction methods can indicate changes in streambed geology over many kilometers, particular in areas where stream water and groundwater are of similar electrical conductivity (Ong et al., 2010). These methods can be used in combination with point-scale measurements to obtain a more comprehensive, process-based understanding of DWM habitat. Point-scale physical methods, such as in-river piezometers and seepage meters (Rosenberry et al., 2008), indicate the direction and magnitude of flow across the sediment–water interface at specific locations. Streambed vertical temperature profiles can also be used to determine seepage direction and rate, and can extend point-in-time measurements of water flux to month-long time series of sub-daily flux estimates using automated analytical (e.g., Irvine et al., 2015; Gordon et al., 2012) and numerical (e.g., Koch et al., 2015; Voytek et al., 2014) 1-D models. Other temperature-based methods can be used to separate groundwater discharge from superimposed hyporheic flow. Thermal infrared (TIR) and fiber-optic distributed temperature sensing (FO-DTS) methods are used to collect large field-of-view (100 s of m) or extensive longitudinal (km) water-temperature measurements (Hare et al., 2015). TIR does not penetrate the water surface, whereas FO-DTS measures temperature along the sediment–water interface.
We used the above-listed methods to investigate the occurrence and
distribution of groundwater discharge along three reaches of the upper
Delaware River. Along each reach, which we refer to as sites, we compared
results where DWM were present with results where they were absent.
Specifically, we pursued three main goals:
determine the spatial distribution of the rate and direction of water
exchange across the sediment–water interface related to the distribution of
DWM populations; evaluate temperature dynamics at the sediment–water interface during
warm, summer low-flow periods to investigate larger-scale groundwater
discharge distributions, and determine whether areas populated by DWM may
serve as cold thermal refugia; investigate the geology of the riverbed and relate groundwater–surface-water exchange to potential geologic controls.
During the course of the investigation, we discovered a relatively large
spring within an area populated by DWM and studied in detail the thermal
influence on adjacent and downstream water (Briggs et al., 2013). Here we
expand the scope more broadly to address the three goals listed above with
data collected at all three DWM-populated reaches of the Delaware River.
Delaware River reach (highlighted) on the border between New York and Pennsylvania between Hancock and Callicoon.
Sites 1
The three study sites containing DWM are within the 43 km reach of the upper Delaware River between Hancock and Callicoon, NY (Fig. 1). Prior to collection of data for this study, these sites were surveyed in 2012 by biologists familiar with DWM to determine the riverbed areas currently occupied by DWM. Each site encompassed areas where DWM were found as well as similar adjacent or nearby areas where DWM had never been found. Previous studies at these same three sites investigated minimum flows and temperature stability (Cole et al., 2008) and modeled shear stress related to occurrence of DWM (Maloney et al., 2012). Site 1 extends along the right (descending) side of a mid-channel island (Fig. 2a). Site 2 extends along a straight reach of the river where a single channel exists, and is centered above and below an ephemeral stream that enters the river on the right bank, approximately separating the known DWM area (M) and non-mussel area (N) (Fig. 2b). The known M area at site 3 is situated along the south side of a mid-channel island, while the N area is on the north side (Fig. 2c). At site 3, DWM were found at various times over a 10-year period along the entire reach of the channel south of the mid-channel island. However, during the 2012 field season, DWM were found only along a 200 m reach at the downstream end of this channel (Fig. 2c). Because DWM were found upstream of the M reach prior to 2012, the upper portion of the channel was deemed inappropriate to serve as the N reach. Therefore, the N reach at site 3 was located across the mid-river island in the river channel to the left (north) of the island where current velocity was reduced. All locations are deliberately obscured to protect the endangered animals (Fig. 2).
Discharge (
Median daily river discharge (
Grain-size distribution of the bed surface was determined using the Wolman (1954) pebble-count method over an approximately 100 m distance within each of the M and N reaches. River depth and flow velocity were measured at every M and N location at approximately the same time. River-surface slope was surveyed along and beyond each M and N reach; combined with measurements of water depth, this provided a reach-averaged estimate of shear stress for each M and N reach. Shields stress, a dimensionless term that relates shear stress to the size of sediments on the bed (e.g., Buffington and Montgomery, 1997), was calculated and compared to critical Shields stress to determine the likelihood that the sediment bed was mobilized based on water depths measured during site visits.
Samples for water-quality analysis were collected from piezometers installed at each M and N location at sites 2 and 3, from water-table monitoring wells, from several seeps along the river bank, from the large spring/seep at site 2 (Fig. 2b), and from the river at each site. Groundwater and river chemistry were found to be universally similar; therefore, these results are not discussed in further detail.
Discharge of groundwater to the river was visually evident at all three sites. Flowing water either discharged along the bank just above the river surface (sites 1 and 2) or was visible as it discharged rapidly enough to suspend sediment just beneath the river surface (sites 1 and 3). A handheld TIR camera (FLIR T620, FLIR Systems, Inc., Nashua, NH) was used to locate and measure surface-temperature anomalies related to cold groundwater seepage near the streambank. TIR data were used to quickly discern between actively flowing seeps and other bank areas that were simply wet. TIR imagery represents only the temperature at the water or land surface; therefore, the cameras were most useful for identifying seeps at and landward of the shoreline and unmixed plumes of groundwater that reached the river surface (e.g., Hare et al., 2015). A bucket and stopwatch were used to quantify spring/seep discharge where conditions allowed.
A water-table monitoring well was installed adjacent to the right (southwest) bank of the river at each site to determine the hydraulic gradient between the water table and the river. Wells were installed using an auger to depths beneath land surface of 2.85 and 2.81 m at sites 2 and 3, respectively. At site 1, a monitoring well could not be installed because boulders in the bank were too large and densely distributed to auger a hole. However, a 0.46 m deep hole was dug by hand to below the water table at a distance of 7.1 m from the river shoreline. This allowed a single measurement of horizontal hydraulic gradient at site 1 (Fig. 2a).
Discharge of groundwater to the river was calculated using the standard
Darcy equation:
EMI quadrature data at 33 030 Hz converted to bulk conductivity
for
A single-well slug test (Bouwer and Rice, 1976; Bouwer, 1989) was conducted
in each monitoring well to estimate
Seepage meters directly measure water flow across an approximately 0.25 m
Similar to the riverbank monitoring wells, streambed piezometers can
determine the potential for the direction of flow by measurement of
hydraulic head within the streambed compared to the surface-water stage, but
on a vertical axis. Streambed piezometers were installed directly adjacent
to seepage meters at all M and N locations, except where installations were
impossible due to buried boulders or where locations were so close together
that one piezometer could represent both locations. Piezometers
consisted of a stainless-steel pointed screen (30 mm diameter and 85 mm
screened interval) connected to 27 mm diameter galvanized pipe. Piezometers
were driven to approximately 0.5 to 0.6 m depth beneath the riverbed.
Completion depth was less than 0.5 m if, after several attempts, buried
cobbles or boulders prevented deeper installation. In some locations where
vertical head difference was very small, the piezometer was driven to a
greater depth to create a measured head difference larger than the
measurement error. Insertion depths ranged from 0.42 to 1.15 m. In-river
piezometers can indicate rates of exchange at the sediment–water interface
if a value for
Surface temperature variations propagate downward into streambed sediments due to the sum of conduction and advection; if the conductive properties of the bed are measured or assumed, vertical advection can be determined using 1-D analytical or numerical models (Constantz, 2008). Thermistor dataloggers (iButton Thermochron DS1921Z, Maxim Integrated, San Jose, CA) were installed at depths ranging from 0 to 0.4 m in 14 of the piezometers to provide temperature profiles with depth over time. These temperature records were collected for approximately 3 to 7 days. Strong, upward groundwater flow often reduces measurable diurnal signal penetration to less than 0.2 m (Briggs et al., 2014); therefore, at least 1 short complementary temperature profiler designed specifically to measure upward seepage was installed within the M zone at all three sites. These short profilers were constructed with four thermistor dataloggers (iButton Thermochron DS1922L) positioned at depths of only 0.01, 0.04, 0.07, and 0.11 m beneath the riverbed. One such profiler was installed at site 1 in close proximity to an observed bankside seep, three profilers were installed at site 2 adjacent to seepage meters, and two were installed adjacent to seepage meters at the site 3 M reach (Fig. 4; locations indicated as 1-D temperature). Temperature records were collected for approximately 25 days.
Streambed-temperature time-series data were analyzed with the VFLUX program (Gordon et al., 2012) run in Matlab (Mathworks, Natick, MA). Diurnal signals were extracted from field data using VFLUX and applied to the amplitude–attenuation analytical model (as described by Hatch et al., 2006) because this model has been shown to be reliable in determining upward flow rates (Briggs et al., 2014). Error associated with sediment-property uncertainty was determined using Monte Carlo analysis and adjusting sediment thermal properties within expected ranges (Briggs et al., 2012b). This method of analysis provides the ability to resolve temporal patterns of vertical seepage at sub-daily time steps over the period of temperature-data collection.
A Sensornet Oryx (Sensornet House, Elstree, Hertfordshire, UK) fiber-optic distributed-temperature-sensing system (FO-DTS) was deployed on the riverbed at sites 2 and 3 to collect continuous temperature data in space and time along linear cables (e.g., Selker et al., 2006). The stainless-steel-reinforced fiber-optic cables were distributed across 585 m of the streambed at site 2 and across 944 m of the streambed at site 3. The deployment at site 2 (21–24 July 2012) encompassed adjacent M and N reaches, while the site 3 installation (25–27 July 2012) only covered the M reach due to length limitations of the cable. FO-DTS data were analyzed to identify locations of anomalously cold temperature and small thermal variance that may correspond with focused groundwater seepage to the river (e.g., Briggs et al., 2012a), and thermal refuge for the DWM.
FO-DTS data were collected at 4 and 10 min intervals and calibration for
thermal drift was performed using a continuously mixed ice bath monitored
dynamically by a Sensornet thermistor-type thermometer. Approximately 30 m
of cable were placed in the calibration ice bath. The standard deviation of
the recorded FO-DTS temperatures in the ice bath, determined to be 0.07
In addition to the spatial coverage provided by the linear FO-DTS cables,
manual point (snapshot) measurements of streambed temperature were
collected at 0.05 m sediment depth using a high-precision (0.01
Bedrock and unconsolidated materials have characteristic electrical-conductivity properties that can be sensed remotely with a variety of geophysical tools. Multi-frequency electromagnetic-induction (EMI) data were used to make inferences about underlying geologic structure of the streambed; EMI has been used previously to better constrain exchange between groundwater and surface water at landscape scales (e.g., Ong et al., 2010). These data were collected at all three sites using a portable digital, multi-frequency, electromagnetic conductivity sensor (GEM-2; Geophex, Inc., Raleigh, NC) that measures the bulk apparent subsurface electrical conductivity (or magnetic susceptibility). Variance in electrical conductivity provides information about groundwater quality (e.g., salinity) or substrate properties, such as porosity. Larger conductivity values correspond to more conductive subsurface materials, such as shale bedrock or near-surface materials with a higher silt or clay fraction, whereas smaller conductivity values may indicate sandstone bedrock or coarser-grained surficial deposits. GEM-2 can be used to estimate streambed characteristics at depths up to approximately 12 m depending on streambed composition.
Multi-frequency EMI data were collected at all three sites. A fixed land location was established at each site and visited at the beginning and end of each survey to correct for instrument drift. The instrument was suspended about 1 m above the water surface using non-metallic PVC pipe secured inside an inflatable raft. A kayak and drogue were used to position the raft to provide areal coverage of the riverbed. All GEM-2 land locations and surveys were geo-referenced with an on-board GPS unit.
Median values for parameters measured at each installation location at sites 2 and 3.
Median water depths measured at site 2 during the June 2012 field visit were
0.58 and 0.59 m for M (DWM present) and N (DWM not present) locations,
respectively. Median depths at site 3 for M and N locations were 0.41 and
0.44 m, respectively (Table 1). Median river velocities were virtually
identical between M (0.18 m s
Reach-averaged shear stress was nearly identical at the M and N locations at site 2, primarily because the slope of the river surface (0.00037) was the same at both reaches. The M reach at site 3 also had virtually the same slope (Table 1). The slope at the N reach at site 3 was nearly twice as large at 0.00065. Therefore, shear stress at site 3, reach N, was more than double that of any of the other reaches. Shields stress (Table 1) at all reaches was well below commonly assumed critical values of 0.03 to 0.06 required for bed mobility (e.g., Shvidchenko et al., 2001).
River slope and water depths were not measured at site 1. Maloney et
al. (2012) indicated that water depth, current velocity, and shear stress at
site 1 are similar to site 2 during river discharge (less than about
100 m
Walking along the site 1 riverbank within, above, and below the reach where DWM have
been identified revealed 10 bank-side seeps on both sides of the channel
southwest of the mid-channel island (Fig. 2a). Small wetland areas of
approximately 10 to 30 m
The spring area at site 2 (Figs. 2b and 5) included two areas approximately
0.1 m in diameter and separated by about 0.5 m that discharged copious
amounts of water and is described in detail in Briggs et al. (2013). The
smaller spring discharged 12.9 L min
At site 3, the riverbank immediately southwest of the riverbed area where DWM
have been located did not contain any obvious seeps, but the 10 to 15 m wide
bench immediately adjacent to the shoreline was soft and wet in areas. Two
seeps were identified at the shoreline next to the mid-channel island,
approximately equidistant from the two westernmost M locations (Fig. 2).
These seeps discharged water both above and below the shoreline and suspended
sand where the discharge point was submerged. Five other colder seeps were
located upriver along the right bank (south side) of the channel. The
discharge point at all of the right-bank seeps was 0.1 to 0.2 m above the
river surface. Although difficult to measure, several of the seeps were
discharging at approximately 0.5 to 2 L min
The single measurement of hydraulic gradient (
The median value for
River stage, water-table elevation, and hydraulic gradient at
sites 2 and 3. Legend in panel
The median value for hydraulic gradient based on the monitoring well at site
3 was 0.05 (Fig. 6c). Other than during high-discharge events, the median
value was remarkably stable during the period of record. The gradual decrease
in hydraulic gradient from late July until early September 2012, is likely a
return to hydrostatic conditions following well installation and indicative
of the low
Relative river stage and water level in the adjacent groundwater monitoring
well are plotted in Fig. 6 for sites 2 and 3. Values are adjusted so river
stage approximately equals the water depth at each in-river pressure
transducer. As the river stage rose, the shoreline moved laterally and the
distance between the shoreline and the monitoring well decreased.
Calculations of
During the 1-year period from July 2012 through June 2013, the hydraulic
gradient at both sites 2 and 3 reversed and became negative 7 times in
response to a rising river stage that preceded and exceeded a corresponding
rise in the water table at the monitoring wells. This effect is displayed in
Fig. 6b for the largest rise in river stage at site 2 on 19 September. The
rapid increase in river stage from 03:00 to 06:00 LT was substantially larger than
the responding increase in head at well WT1, resulting in a reversal of
hydraulic gradient that exceeded
Seepage was generally small at all but a few M and N measurement locations.
Median values of seepage were upward at 8 of 10 M locations and at 6 of
10 N locations. Both M reaches had positive (upward) median values (0.43 and
3.55 cm d
Median values of seepage flux. Error bars indicate maximum and
minimum measured values. Median value for site 3 N2 is 84 cm d
Seepage at locations S1 and S3 at site 2 was only slightly to moderately faster relative to seepage measured at nearby M locations (Table 1). Although larger rates of seepage were expected within this cold-water spring area, detailed temperature measurements indicated that most of the seepage, and the source of the cold water at the streambed, originated landward of the shoreline (Briggs et al., 2013).
Vertical hydraulic gradients (
Calculated
Vertical seepage rates determined with VFLUX from the thermal records
collected in piezometers varied substantially depending on which pair of
thermometers was used to calculate
Seepage rates determined with VFLUX and Delaware River discharge determined over 25-day period from 28 June–23 July.
Riverbed temperatures indicated by snapshot thermal surveys
(shaded riverbed areas) and FO-DTS at site 2 (panels
Comparison of time-averaged VFLUX seepage values and median seepage-meter values at select locations.
Seepage determined with temperature data from the shallow profilers designed
to capture upward flow averaged 16 cm d
Slow seepage-flux estimates in the range of
Average FO-DTS temperatures collected over 4 days at site 2 ranged from
14.0 to 22.5
Temperatures measured with the snapshot streambed thermal surveys at sites 2 and 3 are generally similar to patterns shown in the FO-DTS data. However, the snapshot data indicated several discrete cold zones near the island at site 3 that were missed with nearby FO-DTS cables. Discrete cold patches were found at sites 2 and 3 along the M zones but not in the N zones (Fig. 9). The cold anomalies make up a relatively small percentage of the overall surveyed area at both M reaches. The largest cold anomaly is located at the site 2 spring area and indicates a plunging plume of cold water, as discussed earlier. The areal extent of this anomaly is approximately twice as large as the plume footprint measured within the water column, likely indicating an influence from more diffuse groundwater upwelling through the streambed, as detailed in Briggs et al. (2013). Cold riverbed areas were better detected with the discrete snapshot method than with the continuous FO-DTS method, likely because the snapshot measurements were made at 0.05 m depth and the fiber-optic cable was resting on top of the bed and influenced to a greater extent by surface-water temperatures. The snapshot method also provided better lateral distribution of data collection.
Consistent spatial patterns of streambed electrical conductivity were observed in multiple adjacent and overlapping EMI lines, but there was no apparent relation between riverbed electrical conductivity and occurrence of DWM (Fig. 4). For example, DWM areas at sites 1 and 3 are located above more conductive material, whereas corresponding N reaches are generally less conductive. Conversely, DWM at site 2 are found over the least-conductive material, whereas the opposite side of the river and N reach are both more conductive.
Some of the individual methods for characterizing rates and spatial
distribution of groundwater discharge produced inconclusive results,
indicative of the difficulty presented by such a challenging setting.
Collectively, however, they lead to the conclusion that groundwater discharge
is related to occurrence and distribution of DWM in the upper Delaware River.
Listed from strongest to weakest, the evidence stacks up as follows:
Easily visible seeps and springs were present at or just upriver of all
three M reaches but not at the N reaches. Large lateral hydraulic gradients
toward the river indicate the potential for substantial groundwater discharge
at all three sites. Upward seepage through the riverbed measured with seepage
meters was much faster and more consistently upward at reaches populated by
mussels. Median upward vertical hydraulic gradients were 3–9
times larger at M reaches than at N reaches. Seepage based on vertical
temperature profiles measured with two different methods of instrumentation
was upward, circumneutral, and downward, at 2, 5, and 1 of 8 M-reach
locations, respectively, whereas temperature-profile-based seepage was
downward at all three N-reach locations. Riverbed temperature based on FO-DTS
and snapshot streambed thermal surveys was slightly colder in the M reaches
than in the N reaches; bed temperature was particularly cold in discrete
patches that were better captured with the bed temperature snapshot surveys.
Combined results from the first four methods, in particular, provide compelling evidence that groundwater discharge is substantial in areas populated by DWM. Somewhat surprisingly, neither geophysical nor chemical methods were related to presence or absence of DWM. Although patterns were evident, geophysical data showed no clear correlation with M or N reaches. Chemistry of groundwater, water removed from in-river piezometers, and surface water was virtually identical, rendering chemistry, often a good indicator of water source, of little use to distinguish M from N reaches.
It is also clear that groundwater discharge is not evenly or universally distributed across the M or N areas. Hyporheic exchange superimposed on broader-scale groundwater discharge exerts a highly complex flow path distribution that results in variable rates of upward, largely horizontal, and downward seepage across the riverbed. This local-scale variability in seepage direction and rate did not appear to be related to locations of individual DWM. Three pairs of seepage meters and streambed piezometers were installed nearby, and within 1 m of an individual DWM, without evidence of anomalously strong seepage at specific DWM locations. These three paired observations, along with substantial heterogeneity in seepage rate and direction within each M reach, indicate that DWM do not require focused groundwater discharge precisely where they are located, but instead rely on the existence of substantial groundwater discharge within or just upstream of their populated area.
Other studies that have investigated the effect of groundwater discharge on benthic invertebrates have yielded mixed results. One study indicated a direct correlation between rate of groundwater discharge and abundance and taxonomic richness (Hunt et al., 2006), while another showed little correlation (Schmidt et al., 2007). Few studies have related groundwater discharge with mussel abundance and species richness. A study conducted in a river with similarly coarse sediment indicated a relation between mussel population density and upward seepage rate (Klos et al., 2015), but upward seepage in that setting was primarily driven by hyporheic exchange. The net upward seepage at DWM sites in the Delaware River, although clearly influenced by hyporheic exchange, is primarily the result of area-wide groundwater discharge as evidenced by substantially faster reach-averaged upward seepage and also colder water along M reaches relative to N reaches.
Obtaining direct measurements of groundwater discharge is difficult in
settings such as the upper Delaware River where large boulders up to 1 m or
more in diameter are common. Distinguishing hyporheic exchange from
groundwater discharge in coarse-grained fluvial settings can also be
challenging (e.g., González-Pinzón et al., 2015; Menció et al.,
2014; Ward et al., 2013; Bhaskar et al., 2012), hence the multiple lines of
evidence pursued for this study. Therefore, few studies of exchange between
groundwater and surface water have been successfully conducted in such
coarse-grained sediments. Compared to those that have (e.g., Rosenberry et
al., 2012; Fritz et al., 2009; Klos et al., 2015), values for point
measurements of seepage exchange at these three sites on the Delaware River
were not particularly large. This indicates that hyporheic exchange is
perhaps smaller than would be expected along M reaches, given the coarseness
of the bed. And, just as was inferred regarding smaller-than-expected
Cold-water anomalies were detected along all M reaches, but never along an N reach. At site 2, mussel-location data from 2010 and 2012, in particular, indicated a strong clustering of animals directly adjacent to and downstream from the main spring described here and by Briggs et al. (2013) (J. Cole, unpublished data). DWM indeed may be present in these areas due to relatively stable and cold groundwater discharge that serves as a refuge for these animals during periods of lowest river stage. Additionally, mussel surveys have only been taken at these locations during summer months; groundwater discharge also may offer benefits for mussel survival during cold winter extremes that are not apparent based on these data collected during the summer.
Data indicating flow in opposite directions across the riverbed are initially puzzling (Table 2). Hyporheic flow paths in substantially heterogeneous and highly transmissive sediment, a common situation in a cobble-bed river, are predominantly horizontal with small upward and downward flow components. Seepage meters quantify the upward or downward component of flow across the sediment–water interface whether the flow is vertical or largely horizontal. Because piezometers and vertical temperature profilers are installed vertically, thermally derived interpretations of seepage assume vertical flow through the sediments, often a poor assumption in hyporheic settings. It is not uncommon for seepage meters to indicate flow in one direction while hydraulic gradients indicate the opposite (Rosenberry and Pitlick, 2009; Rosenberry et al., 2012; Angermann et al., 2012; Käser et al., 2009). Locations with discordant data are indicative of flow across the sediment–water interface that was largely driven by hyporheic processes, which are superimposed on larger-scale groundwater-discharge patterns (Rosenberry et al., 2012). Hyporheic flow appeared to dominate exchange at the site 3 N reach. Furthermore, substantial changes in the vertical component of hyporheic flow were indicated at most of the locations where temperature was measured at multiple depths in the riverbed (Table 2), also indicative of hyporheic exchange that is reduced or transitions to horizontal flow with increasing sediment depth (e.g., Briggs et al., 2012b).
The first four methods listed at the beginning of the discussion section all provide strong evidence for groundwater discharge to the river, although the scale of those measurements was not the same. Large lateral hydraulic gradients based on data from bankside monitoring wells consistently indicated potential for substantial groundwater discharge at all three study sites on a site-wide scale, but those results could not distinguish between groundwater discharge at M vs. N reaches. The other three methods were more local in scale, which allowed comparisons of M and N reaches. Direct observation of seeps and springs at M but not at N reaches, faster upward seepage at M than N reaches, and larger upward vertical hydraulic gradients at M not N reaches all indicate greater groundwater discharge in and near areas populated by DWM. However, for in-river seepage and hydraulic-gradient measurements, these conclusions can only be reached when data are aggregated within each M or N reach. Otherwise, local-scale hyporheic exchange greatly confounds the interpretation.
Hyporheic exchange also made it difficult to obtain clear interpretations from temperature-based data. Because shallow hyporheic flow paths in coarse-grained sediments are primarily horizontal and temperature-profile methods assume primarily vertical flow, it should not be surprising that data might be difficult to interpret. Other studies in finer sand-bed streams have obtained conclusive and consistent results using these methods (Rosenberry et al., 2016; Hatch et al., 2010). Excluding the questionable temperature-based data in Table 2, when all other location-specific data are aggregated over an entire M or N reach, the conclusion is consistent; greater groundwater discharge occurs at M than at N reaches.
Given the substantial hyporheic exchange that results in upward flow across the bed with basically the same temperature as the river water, it also is not surprising that the two thermal-reconnaissance methods would not show a strong difference between M and N reaches. However, the manually measured snapshot temperature survey still identified colder areas of the sediment bed, but in M reaches only. The efficacy of the snapshot method was a pleasant surprise, likely because measurements were made at 0.05 m sediment depth and better indicated the temperature of discharging groundwater.
Only the methods based on geophysics and chemistry provided data of little value. As suggested previously, geologic controls on distribution of seepage may have been of a scale that was impossible to resolve with these geophysical tools. Regarding chemistry, it may be that groundwater flow paths were neither sufficiently long nor groundwater sufficiently old, in this headwaters area for groundwater chemistry to have developed a water chemistry distinguishable from river water.
Substantial groundwater discharge clearly occurs at areas populated by DWM,
and no areas of focused discharge were identified immediately upstream or
downstream of these three DWM-populated areas. However, is this prodigious
discharge greater than what is typical along the upper reaches of the
Delaware River? Fortunately, river discharge can be compared between two USGS
gaging stations: Lordville (USGS station number 01427207;
Rates of measured or calculated groundwater discharge.
In conclusion, the collective lines of evidence indicate that DWM are situated in or directly downstream of areas of substantial groundwater discharge to the river. The work presented here and in Briggs et al. (2013) may be the first to demonstrate the importance of groundwater discharge to unionid species. Additional work is needed to better understand the linkages between groundwater discharge and presence of DWM as well as geological controls that focus groundwater discharge in these areas.
Data on geomorphic parameters and groundwater–surface-water exchange are available upon request to Donald Rosenberry. Temperature data, seepage rates determined from measurements of temperature, and geophysical data are available upon request to Martin Briggs.
We thank Jeffrey Cole for advice and instructions related to river and riverbed logistics, and Heather Galbraith and Carrie Blakesley for mussel identification and location, all from the USGS Northern Appalachian Research Branch. Don Hamilton from the National Park Service Upper Delaware Recreational and Scenic River, and Joseph Markos, Richfield, MN, are thanked for their field assistance and logistical support. Jason Halm's (University of Colorado-Boulder) exceptional support before, during, and following field work is greatly appreciated. This work was funded by the US Fish and Wildlife Service. Use of trade names is for identification purposes only and does not constitute endorsement by the US Geological Survey. Edited by: C. Stamm Reviewed by: two anonymous referees