The heat tracing of water can be used to map a distribution of spatially focused, or “preferential”, groundwater-discharge zones throughout surface water systems at times of contrast between the surface and groundwater temperature. The measurement of the water temperature from the reach to watershed scale is now possible using thermal infrared and fiber-optic distributed temperature sensing (FO-DTS) methodology (Dugdale, 2016; Hare et al., 2015; Steel et al., 2017). Remote infrared data collection throughout the river corridor has been enabled by handheld cameras, piloted aircraft, and the rapidly evolving capabilities of unmanned aerial systems. Researchers are capitalizing on the ongoing refinement of these technologies to identify zones of focused groundwater seepage to streams in order to map potential discrete preferential cold-water fish habitats such as summer thermal refugia (Dugdale et al., 2015). However, surface thermal surveys alone do not indicate groundwater flow path dynamics or the suitability of an interface aquatic habitat (Briggs et al., 2018a).

For example, dissolved oxygen (DO) concentration must be sufficiently high for cold groundwater seepage to provide support for fish life processes at the direct point of discharge to surface water (Ebersole et al., 2003), which is not apparent from thermal analysis alone. During warm summer periods in systems with suboxic groundwater, cold-water fish species such as salmonids can face a tradeoff between occupying discrete zones of preferred water temperatures with near-lethal DO levels and stream sections that are too warm for long-term survival (Matthews and Berg, 1997). The use of groundwater upwelling zones as thermal refugia is further complicated by competition with aggressive invasive species (to the northeastern USA) such as brown trout, which compete with native trout for resources (Hitt et al., 2017). Streams at higher elevations may support the persistence of reach-scale cold-water habitats where point-scale thermal refugia are not needed under current climatic conditions, serving as vital “climate refugia” against rising air temperatures (Isaak et al., 2015). In systems with reliably cold channel water in summer, which can also exist at low elevations when heavily influenced by discharging groundwater, salmonid fish may directly use groundwater-seepage zones for spawning rather than thermal refuge.

Brook trout (Salvelinus fontinalis) are a species of char that are native to eastern North America, from Georgia to Québec (MacCrimmon and Campbell, 1969). Populations have been stressed by warming temperatures and reduced water quality, particularly in low-elevation areas (Hudy et al., 2008). Stream network-scale tracking of fish has indicated that the brook trout directly utilize stream confluence mixing zones and preferential groundwater discharge to survive warm summer periods (Baird and Krueger, 2003; Petty et al., 2012; Snook et al., 2016). Additionally, brook trout spawn in the fall, and eggs deposited in redds develop over the winter before hatching in spring (Cunjak and Power, 1986). Oxygen use by the shallow buried embryos increases over the period of development (Crisp, 1981); therefore, DO concentration is a critical parameter of the pore water in which the eggs are bathed. Several studies have demonstrated the importance of hyporheic downwelling in increasing shallow oxygen concentrations, including for salmonid redds, where deeper stream-bed pore water is generally reduced in DO (e.g., Buffington and Tonina 2009; Cardenas et al., 2016; Harvey et al., 2013). Fine sediments can reduce the efficacy of hyporheic DO exchange in spawn zones (Obruca and Hauer, 2016) and are actively cleared by trout during the spawning process (Montgomery et al., 1996).

The importance of hyporheic exchange to salmonid spawning may be limited in the lowland streams that are expected to harbor native cold-water species in the 21st century, namely those with strong groundwater influence. Groundwater upwelling reduces the penetration of the hyporheic flow from surface water (Cardenas and Wilson, 2006) and may shut down hyporheic flushing in redds (Cardenas et al., 2016). While hyporheic exchange introduces oxygenated channel water into the shallow stream bed, the downward advection of heat associated with near-freezing surface water in winter will also cool stream-bed sediments (Geist et al., 2002), potentially impairing egg development. Coaster brook trout, a life-history variant of native brook trout exhibiting potadromous migrations within the Great Lakes, have been shown to specifically prefer groundwater-discharge zones for building redds (Van Grinsven et al., 2012). The development of trout in winter has been found to positively correlate with warmer stream water temperatures as influenced by groundwater seepage (French et al., 2017). Therefore, spatially discrete groundwater-discharge zones with adequate DO may form preferred brook trout spawning habitats (Curry et al., 1995).

Multiscale physical and biogeochemical factors influence temperature and DO concentrations along groundwater flow paths. In river valleys, discharge to the surface water of locally recharged groundwater is expected to emanate from more shallow, lateral flow paths controlled by the local topography (Modica, 1999; Winter et al., 1998). Shallow groundwater flow paths, particularly those within approximately 5 m of the land surface, will be more sensitive to annual air temperature patterns and long-term warming trends due to strong vertical conductive heat exchanges (Kurylyk et al., 2015b). The distance of seeps from upgradient groundwater recharge zones will also affect seepage temperature dynamics and associated aquatic ecosystems due to future changes in surface and recharge temperatures (Burns et al., 2017). Therefore, characterizing the hydrogeochemical attributes of discharging groundwater flow paths is critical in understanding the thermal stability of current and future point-scale preferential brook trout habitats (Briggs et al., 2018a). The complimentary methodology of geophysical remote sensing, geochemical sampling, and vertical bed temperature time series can indicate the physical and chemical properties of groundwater flow paths that source preferential discharge zones utilized routinely by fish for spawning.

Coarse-grained mineral-dominated aquifers with little fine particulate organic matter and low dissolved organic carbon supplies tend to result in generally oxic groundwater conditions (Back et al., 1993). The sandy surficial aquifer of Cape Cod, where our investigation took place, is a classic example of a mineral soil-dominated flow system (Frimpter and Gay, 1979). The flow of groundwater through near-stream organic deposits, however, can result in inverted redox gradients toward the upwelling interface, such that groundwater discharged to surface water is reduced in DO (Seitzinger et al., 2006). In sandy glacial terrain with superimposed peatland deposits, the specific flow patterns of groundwater to surface water in relation to buried peat will influence the groundwater-discharge biogeochemistry. Krause et al. (2013) found that stream-bed groundwater seepage was strongly reduced in DO in zones with peat deposits, likely due to an increase in both near-stream residence time and localized sources of dissolved organic carbon.

Interdisciplinary collaborations between physical and biological scientists are useful to better understand how cold-water species utilize the stream habitat influenced by groundwater discharge and the larger landscape-scale controls on discharge characteristics. While previous hydrogeological research in the coastal stream used for this study had focused on locating and quantifying discrete groundwater discharge (e.g., “cold anomalies”, Hare et al., 2015; Rosenberry et al., 2016), here we endeavor to understand the hydraulic and biogeochemical controls on seepage zone distribution utilized directly by native brook trout. In this groundwater-dominated stream (e.g., likely climate refugia), brook trout do not need to occupy discrete inflows for summer thermal refugia but do favor certain upwelling zones for fall spawning. We compare over a decade of visual survey and electronic fish passive integrated transponder (PIT)-tag dropout data regarding repeat brook trout spawning locations to a comprehensive physical and chemical characterization of groundwater-seepage zones across 2 km of stream in order to do the following: 1.

identify repeat brook trout spawning locations and determine if they are directly associated with the preferential discharge of groundwater through interface sediments, and

Cape Cod is a peninsula in southeastern coastal Massachusetts, USA, composed primarily of highly permeable unconsolidated glacial moraine and outwash deposits. The largest of the Cape Cod sole-source aquifers occupies a western (landward) section of the peninsula (LeBlanc et al., 1986) and is incised by several linear valleys that drain groundwater south to the Atlantic Ocean via baseflow-dominated streams. Strong groundwater discharge to one such stream, the Quashnet River, supports a relatively stable flow regime that has averaged 0.49±0.15 (SD) m3 s-1 from 1986 to 2015 (Rosenberry et al., 2016). The lower Quashnet River emerges from a narrow sand and gravel valley into a broader area with well-defined lateral floodplains. Historical cranberry farming practices, abandoned in the 1950s, have modified the stream corridor (Barlow and Hess, 1993). Primary modifications included the straightening of the main channel (reducing natural sinuosity), installation of flood-control structures, incision of shallow groundwater drainage ditches in the lateral peatland floodplain, and widespread application of sand to the floodplain surface. The current bank-full width of the main channel averages approximately 4 m.

The Quashnet River has long been recognized as a critical habitat for a naturally reproducing population of native sea-run brook trout (Mullan, 1958) with a genetically distinct population (Annett et al., 2012). Efforts to restore trout habitats by the group Trout Unlimited and others have been ongoing for over 40 years (Barlow and Hess, 1993). These efforts include the removal of flood-control structures, the planting of trees along the main channel, and the addition of wood structures to stabilize banks and provide cover from airborne predators. Furthermore, the Commonwealth of Massachusetts purchased 12.5 ha in 1956 and an additional 146 ha along the lower Quashnet River in 1987 and 1988 to protect the area from development. The Massachusetts Division of Fisheries and Wildlife has been monitoring trout populations since 1988 and their movement since 2007.

The groundwater influence on stream temperature is pronounced, particularly over the 2 km reach above the U.S. Geological Survey gage no. 011058837, below which the stream stage is tidally affected. Ambient regional groundwater temperature is approximately 11 ∘C (Briggs et al., 2014), and strong conductive and advective exchange with the proximal aquifer maintains the surface water temperature well below the lethal threshold for brook trout (maximum weekly average temperature >23.3 ∘C, Wehrly et al., 2007). Therefore, point-scale thermal refugia are not a current concern in this system, as the stream supports a system-scale cold-water habitat that is likely to persist into the future and serve as warming “climate refugia” (Briggs et al., 2018a). In winter, seepage zones can be located as relatively warm anomalies, increasing and buffering surface water temperatures from ambient atmospheric influence.

Previous work has measured relatively large net gains in streamflow over the lower Quashnet River (Barlow and Hess, 1993; Rosenberry et al., 2016), which are attributed to groundwater discharge through direct stream-bed seepage and the harvesting of groundwater from the floodplain platform via relic agricultural drainage ditches. Deployments of fiber-optic temperature sensing (FO-DTS) cables along the thalweg stream-bed interface indicate that the greatest density of focused seepage zones occurs along the broader valley area, approximately 1 km upstream of the U.S. Geological Survey gage (Fig. 1). This zone coincides with the largest gains in net streamflow (Hare et al., 2015). Based on the stream-bed interface temperature data presented by Rosenberry et al. (2016), Fig. 1 shows how temperature-sensitive fiber optic cables have been used to pinpoint possible groundwater-discharge zones based on an anomalously cold mean temperature and/or reduced thermal variance. A focused evaluation of FO-DTS anomalies with physical seepage meters and vertical temperature profilers confirmed localized, meter-scale seepage zonation along the streambed where discrete colder zones indicated through heat tracing showed approximately 5 times the groundwater-discharge rate of adjacent sandy bed locations only meters away (Rosenberry et al., 2016). The active heating of wrapped FO-DTS cables deployed vertically within an open-valley stream-bed seepage zone indicated the true vertical flow to at least 0.6 m into the bed sediments (Briggs et al., 2016), an expected characteristic of a more regional groundwater discharge (Winter et al., 1998), rather than that of a flow driven by the valley topography local to the river. Hyporheic exchange in the lower Quashnet River system is superimposed on the general upward hydraulic gradient to the stream, therefore being reduced to a thin, shallow hyporheic exchange zone (e.g., <0.1 m depth) along the thalweg by these competing pressures (Briggs et al., 2014). Vertically compressed hyporheic zones such as these have been simulated for similar stream systems (e.g., Cardenas and Wilson, 2006).

A combination of fish tagging and visual spawning observations, heat tracing, geophysical surveys, and focused pore-water sampling was used to investigate the interplay between the locations of preferential brook trout spawning and the local hydrogeology. For consistency between varied methods and years of data collection, all sample locations are spatially referenced as downstream channel distances from the fish ladder river crossing at the upper end of the study reach (Fig. 2).

The hydrogeochemical characterization of observed repeat trout spawning zones and other major stream-bed groundwater-discharge zones are contrasted below.

Heat tracing reconnaissance technologies, such as FO-DTS and thermal infrared, offer an efficient means to comprehensively characterize preferential groundwater-discharge points at the reach to watershed scale (Briggs and Hare, 2018). Using the groundwater-fed Quashnet River as an example, Rosenberry et al. (2016) showed that cold stream-bed interface anomalies in summer indeed correspond to discrete zones of particularly high groundwater discharge through stream-bed sediments. This spatial characterization of discharge points alone is not sufficient to understand the physical and chemical drivers of a niche habitat, but can efficiently guide additional data collection, as was done here. Compared to more randomly distributed stream-bed field parameter surveys and larger spatial scale evaluations of net groundwater discharge made with differential gaging, the comprehensive spatial mapping of groundwater discharge using heat is a great advance in the context of understanding groundwater-dependent ecosystems. However, in fast flowing streams, FO-DTS cable placement on the stream bed will likely impact which specific groundwater-discharge zones are captured with FO-DTS, as shown here by applying cables along opposite banks through the Spawn 3 area (Fig. 4). The largest seepage zones may have a spatial footprint that encompasses the stream-bed area from bank to bank (e.g., the Spawn 3 cold anomaly), but a subset of more discrete seepage zones are bound to be missed with a single linear cable deployment. We did not capture Spawn zones 1 and 2 in early FO-DTS field efforts (Fig. 1), but fish tracking indicated their importance in regards to trout spawning behavior. Therefore, in studies of niche stream habitats as influenced by preferential groundwater discharge, a combination of heat tracing and biological observation may be needed to both identify major discharge points and discern which points are directly used by the biota of interest (e.g., brook trout).

In a study of the regional Cape Cod aquifer condition, Frimpter and Gay (1979) state that groundwater is typically near DO saturation, except in the case of the downgradient of peat or river bottom sediments, where consumption of DO allows the mobilization of natural iron and manganese. Visible observations along the open-valley section, in addition to stream-bed sediment coring (Briggs et al., 2014), revealed the widespread coating of shallow stream-bed sediment grains with metal oxides, consistent with the conceptual model of organic material influence on near-surface groundwater (Fig. 9). Aquifer recharge passing through upgradient groundwater flow-through kettle lakes (e.g., Stoliker et al., 2016) may also serve to decrease the DO content of the regional flow paths that discharge vertically through the bed of the Quashnet River, although we hypothesize that localized peat deposits may be the primary control on both seepage zone distribution and chemistry.

Out of the dozens of preferential groundwater-discharge zones located along the lower Quashnet with heat tracing, most were suboxic to anoxic (Table 1). Brook trout consistently prefer three areas for fall spawning, all along meander bend cutbanks into the sand and gravel valley wall. Zones of locally enhanced seepage, likely controlled by subtle differences in sediment hydraulic conductivity, can lead to the groundwater sapping of fines, reduction in bank stability, and consequent slumping of bank material into the river; this process was observed in real time at the Spawn 3 meander in February 2016 (Fig. 3c). Slumping effectively forms seepage-driven alcoves outside of the main flow and are more suitable for redd placement, along with forming a more favorable coarse sand and gravel substrate (Bowerman et al., 2014; Hausle and Coble, 1976; Raleigh, 1982).

In other systems, trout have been observed to occupy microhabitat around and within groundwater-discharge zones, even being segregated by fish size and desirable temperature range (e.g., Fig. 2.4.1.2 in Torgersen et al., 2012). Here, real-time observation and visual imagery show trout clustering tightly against the bank in Spawn 3 (Fig. 3d, Supplement) where pore water was found to be more oxygen rich and lower in SpC. The month-long time series of vertical groundwater-discharge rates are reduced considerably from the near-bank to the near-channel areas at all spawning zones (Fig. 6), indicating in part a reduction in stream-bed hydraulic conductivity as influenced by peat deposits under the main channel and as observed in GPR data (Fig. 5). The evidence of higher near-bank vertical groundwater flux rates and DO combined with lower SpC indicates limited interaction between the shallow groundwater flow paths and peat against the meander bend cutbanks. As observed in other systems, it appears that even short travel distances through organic deposits toward the center channel at Spawn 1 and 2 may be sufficient in increasing total dissolved solids, depleting DO (e.g., Levy et al., 2016), and rendering upwelling zones undesirable for redd construction. Therefore, near-surface channel sediments may need to be specifically characterized in preferential groundwater-discharge zones, as net chemical reactivity over the last ∼1 m of transport may dominate net chemical change of the discharging groundwater.

The alcove seepage features utilized by trout in this study are apparently similar to the numerous cold-water alcove patches observed in another stream system by Ebersole et al. (2003). In that study of preferential salmonid habitats, alcoves were often located where streams converged on valley walls and were the most abundant type of discrete cold-water habitat type identified. Conversely, valley-wall alcoves were the least common type of seep morphology observed along the Quashnet River. It is likely that the artificial reduction in channel sinuosity along the Quashnet River by farming practices has reduced the number of natural higher-quality spawning locations.

Other bank and alcove features with strong groundwater discharge found along the open-valley section (Fig. 3b) were highly influenced by organic material deposition and did not apparently support spawning habitats. Our research indicates that in lowland systems with organic-rich floodplain sediments, valley-wall alcoves alone create a favorable brook trout spawning habitat via local mineral soil-dominated groundwater-discharge flow paths, as shown in conceptual Fig. 9. This finding might help inform future ecologically based stream restoration practices in using the natural landscape to predict desirable preferential groundwater-discharge points, as was recently done by Hare et al. (2017) to inform the engineering of a large-scale cranberry bog restoration.

The pore-water SpC, Cl-, and DO data alone do not definitively show that seepage at the cutbank spawn sites is derived from more localized groundwater recharge, as opposed to regional groundwater that is unadulterated by buried peat lenses. However, the hydrodynamic data derived from long-term vertical temperature profiling in seepage zones does offer additional insight. In general, groundwater-discharge rates are more variable at cutbank spawn zones than in the open-valley stream-bed zones (Fig. 6), and this variability may be tied to shorter-term changes in local river stage and/or water table depth, impacting the local hydraulic gradient. The relatively stable patterns of open-valley groundwater discharge may be controlled by the regional gradient, where the flow path length term dominates the Darcy relation and is therefore relatively insensitive to local changes in river stage and water table fluctuations. Furthermore, the stable water isotope data display evaporative signatures at the open-valley stream-bed discharge sites, indicating regional groundwater that has passed through one or more upgradient flow-through lakes (Table 2). In contrast, the Spawn sites all show isotope signals that fall along the local meteoric waterline and therefore likely represent recharge to the hillslopes more local to the river. These localized groundwater flow systems would be expected to be less influenced by regional groundwater contamination, which is widespread in the regional Cape Cod aquifer (Walter and Masterson, 2002).

Groundwater drainage-ditch data collected along the river corridor indicate that low SpC/Cl- conditions exist for the majority of ditches throughout the lower Quashnet River riparian areas (Fig. 7). The hillslope piezometer in sand and gravel at the down valley wall has a similar chemical signature along with high DO. This similarity further indicates that low-SpC groundwater discharges even to the lower portion of the river corridor but is chemically modified by travel through near-stream organics. The relic drainage ditches allow the discharging groundwater to effectively short-circuit the valley floor peat deposits and remain high in DO, similar to the natural valley-wall springs and cutbank alcoves. Future restoration strategies that seek to actively enhance groundwater discharge (e.g., Kurylyk et al., 2015a) may consider capitalizing on this short circuit behavior, possibly by auguring through buried stream-bed peat or through the movement of the stream channel toward the valley wall to create more desirable brook trout aquatic habitat.

The three repeatedly utilized discrete spawning zone locations that have been identified for over a decade of observation have coupled strongly discharging groundwater with high DO concentration. A conceptual diagram of the hydrogeochemical setting of spawn zones vs. other non-favorable stream-bed locations of groundwater discharge is shown in Fig. 9. Spawn zones are located exclusively in side alcoves of the channel created by bank slumps along meanders, where the river cuts into steep hillslopes along the glacial sands and gravel valley wall. In the alcoves at the base of the cutbanks, hillslope groundwater with high DO concentrations is discharged through the stream bed without appreciable loss of oxygen. Just a few meters away toward the main channel, however, groundwater consistently discharges at lower rates, reduces in DO, and increases in SpC. The lowest oxygen concentrations in groundwater are associated with water emerging from the stream bed adjacent to the wide riparian areas that flank the Quashnet in the open-valley section of the study reach, even though groundwater-discharge rates were also relatively high. In the open valley, where the stream is not near the valley walls, proximity to the stream bank does not seem to control seepage chemistry, and GPR data indicated thick zones of discontinuous stream-bed peat. In this and other groundwater-dominated streams that are expected to serve as climate refugia for future native trout populations, hyporheic exchange will be limited by a strong upward hydraulic gradient. Therefore, preferential spawning habitat in such lowland valley systems may be primarily supported by discrete zones of oxic groundwater upwelling at the meter to sub-meter scale, as has been indicated by previous work (e.g., Curry et al., 1995).

In systems where all groundwater discharge is universally anoxic, preferential salmonid spawning zonation may be controlled by points of downwelling hyporheic water where shallow sediments remain high in DO (Buffington and Tonina, 2009; Cardenas et al., 2016). However, these hyporheic areas will deliver cold surface water to shallow sediments during winter, which may impair the overwintering of brook trout eggs (French et al., 2017). Here and in many other coastal systems, groundwater temperature is expected to range from approximately 10–12 ∘C, which is an ideal range for brook trout egg development (Raleigh, 1982). Points of oxic groundwater upwelling devoid of near-stream buried organics, combined with a recirculating side alcove and favorable sand and gravel sediments, may provide an ideal and unique and preferential spawning habitat for native trout.

Stream surface or stream-bed interface heat tracing of groundwater discharge offers an efficient means to locate discrete seepage zones but offers only limited insight into source groundwater flow path hydraulics and geochemistry. A combined toolkit that also includes spatially informed (using heat tracing) geochemical and isotope sampling and geophysical imaging can be used to trace groundwater flow paths back into the source aquifer, and develop a robust hydrogeochemical characterization. Additionally, as digital elevation models become more refined and combined with infrared data derived from unmanned aerial systems, the remote identification of relatively small features such as the seepage alcoves described here should be possible. A comprehensive and process-based characterization of a niche stream habitat can be used to guide a stream ecological restoration design that directly incorporates the local preferential groundwater-discharge template.