Use of Ground Penetrating Radar, Hydrogeochemical Testing, and Aquifer Characterization to Establish Shallow Groundwater Supply to the Rehabilitated Ni-les’tun Unit Floodplain: Bandon Marsh, Coquille Estuary, Oregon, USA

Fluvial-tidal wetlands in the Ni-les’tun Unit (~200 hectares) of the Bandon Marsh, Coquille Estuary, Oregon, were analyzed for shallow aquifer conditions that could influence surface water-qualities in reconstructed marsh, pond, and discharge/tidal channels. The wetlands were surveyed for pre-historic channel features, depth to groundwater surface (GWS), and subsurface salinity intrusion by ground penetrating radar (GPR) in 50 profiles, totaling 11.1 km in track line distance. Only small flood-discharge/tidal channel features (<10 m width and 1–2 m depth) were recorded in the interior floodplain areas. GWS reflections were observed at 0.5–2.0 depth, where the GPR signal was not obscured by localized salinity intrusion (~0.5 km landward distance) from the adjacent Coquille Estuary channel. Top-sealed piezometers (1.5–2.0 m depth) were installed at 10 sites, where in-situ groundwaters were monitored for temperature (8.5–16.5° C), conductivity (<100–18,800 μS cm-1), and pH (2.5–7.8) on a seasonal basis. Dissolved oxygen was semi-quantitatively measured (ChemSticks) at some sites, and all sites were monitored (fall, winter, summer) for GWS level. Low dissolved oxygen (DO <1 ppm) at four sites was of particular concern for potential discharge into small channels that were to be constructed for juvenile salmonid nursery habitat. The horizontal and vertical asymmetries of conductivity (salinity), used as a conservative groundwater source tracer, and measured GWS elevation trends (gradients) led to a four-part flow model for shallow groundwater supply in the Ni-les’tun floodplain. Freshwater supplied, in part, by hillslope discharge contributes to low pH and low DO water quality in the shallow aquifer. Saline water, supplied by subsurface salinity intrusion and evaporative capillary rise, could introduce salinity toxicity to isolated (stagnant) surface ponds. Following construction of a dense channel network (2009–2011) by the Bandon Marsh National Wildlife Refuge, selected Ni-les’tun channel waters (13 sites) were monitored (2011-2012) for resulting water-quality. The tidally-connected channels generally showed improved water-quality relative to groundwater in some nearby piezometer sites. However, low-quality groundwater supply compromised some channel reaches (DO ~2.0–4.7 ppm) that depended on groundwater recharge from hillslope discharge during either summer or winter conditions.


Introduction
In this article, hydrogeomorphic analyses of the rehabilitated Ni-les'tun Unit of the Bandon Marsh National Wildlife Refuge (Figure 1) (USFWS, 2019a) are used to demonstrate relations between shallow groundwater conditions and surface wetland habitats. Wetlands in fluvial-tidal floodplain settings, such as the Ni-les'tun Unit, can be dependent on groundwater supply during periods of limited surface water recharge (Winter, 1995;Winter, 1999;Moore, 1999;Custodio, 2000). Though unseen from above, such shallow aquifer conditions can be important to supporting wildlife habitats in marshes, groundwater window ponds, and incised drainage/tidal channels (Krause et al., 2007;Meyer and Posey, 2009). Rehabilitation planning for previously diked, drained, and/or buried floodplains can benefit from understandings of the shallow aquifer and associated groundwater conditions that could influence the reconstructed floodplain habitats (Carol et al., 2013). In early planning for rehabilitation of the Ni-les 'tun Unit (2003), ground penetrating radar (GPR) profiling and ground-truthing by auger gouge-coring were conducted throughout the study area (~200 hectares) to establish 1) latest-prehistoric channel morphology, 2) subsurface depth of ground water surface (GWS), and 3) landward extents of subsurface salinity intrusion. Seasonal measurements of GWS depth and groundwater hydrogeochemistry (2005)(2006) were obtained from sealed piezometers, which were strategically positioned in the Ni-les'tun floodplain. The porosity and hydraulic conductivity (permeability) of the shallow aquifer were constrained by permeameter analyses of remoulded geoprobe core samples. Rehabilitated drainage/tidal channels in the Ni-les'tun Unit floodplain were monitored for water-quality after initial channel network construction and storm/tide gate removal at tidal creek entrances (Beard, 2013). The depth of sandy soil saturation and corresponding standing water elevations in reconstructed drainage/tidal channels are locally dependent on the GWS elevation trends that extend under the Niles'tun floodplain surface. The marshes and reconstructed channels in the Ni-les'tun Unit are influenced by recharge from hillslope runoff/interflow and subsurface saline water intrusion from the Coquille Estuary channel, as well as by direct precipitation on the floodplain and by Coquille River/Estuary overbank flooding. In this article, we focus on the integrated groundwater studies that were used to constrain the subsurface hydrogeomorphic conditions that could impact wetland habitats in the Ni-les'tun Unit floodplain. Such integrated groundwater studies should be applicable to other fluvial-tidal floodplains that are undergoing wetland rehabilitation for wildlife habitat, particularly in view of potential coastal wetland response to future climate change (Erwin, 2009;Herrera-Pantoja et al., 2012).  (Peterson et al., 2012) and (Part B) the Ni-les'tun Unit (boxed) in the lower Coquille River Estuary. Bedrock (Br) contacts (dashed lines) (DOGAMI, 2019a) are shown landward of uplifted Quaternary marine terraces and overlying late-Pleistocene eolian dune-sheet deposits (Qt) (Peterson et al., 2007). Modern flood plain widths (dotted lines) reflect widths of the ancestral Coquille River valley, which is controlled, in part, by north-south aligned fold belt structures, as is also the case for Coos Bay tidal marshes to the north (Briggs, 1994).

Geomorphology of The Coquille Fluvial-Tidal Estuary
The sinuous nature of the incised (ancestral) lower Coquille River valley is controlled by north-south striking upperplate structures and associated bedrock formation contacts, as developed in the Cascadia subduction zone fold/fault belt (Peterson et a., 2012) (Figure 1A). Incised valley widths and associated modern floodplain widths in the lower Coquille River valley are similarly influenced by the fold belt structural controls ( Figure 1B). Freshwater creeks in fold axes valleys and groundwater from incised ancestral valley hillslopes discharge into the landward margins of the Coquille River/Estuary floodplains. The lowermost reaches of the Coquille River valley are incised into uplifted marine terraces (Baldwin et al., 1973) and overlying late-Pleistocene eolian dune deposits (Figure 2A), locally reaching several tens of meters in thickness (Peterson et al., 2007). Holocene dunes overlap the seaward-most extent of late-Pleistocene dune sheet, where they abut against the northwest corner of the Niles'tun Unit. The Ni-les'tun Unit floodplain is therefore bounded by upland Quaternary deposits to the north and the Coquille River/ Estuary channel to the south and east. The head of tide in the Coquille Estuary occurs about 60 km upriver from the river mouth. Elevated salinity levels extend about 20 km upriver during dry summer months, but are constrained to the lowest estuarine reaches by major river floods in winter months. Such floods deliver lithic-rich sub-angular river sand to the river mouth. The abundant river sediment supply has largely infilled the submerged ancestral Coquille River valley, leading to the extensive floodplain development throughout all but the lowermost-estuarine reaches. Latest-Holocene (3-0 ka) vertical accretion of the fluvial-tidal floodplains ( Figure 2B) has kept pace with net sea-level rise (~1.0 m ka -1 ), as recorded in core site 521. However, alternating deposits of mud and muddy-peat in most of the older floodplain vibracore sections reflect neotectonic cycles of regional coastal interseismic uplift and coseismic subsidence (0.5-1.5 m paleo-tidal level change) in the southcentral Cascadia subduction zone (Briggs, 1994;Peterson et al., 2012). The muddy floodplain deposits are underlain by basal sand deposits, originating from lateral channel migrations in the upper-and middle-estuarine reaches (sites 511 and 521) and from tidal marsh progradation over tidal sand flats in the lowermost-estuarine reaches (site 508). The hydro-morphostratigraphy of the Ni-les'tun Unit floodplain is described in detail in later sections of this article.  (Peterson et al., 2007) are shown relative to the Ni-les'tun Unit (boxed). Part B. Logs of recovered vibracores from selected floodplain sites in the lower Coquille River valley. Estimated coseismic subsidence events (arrows) range from 0.5-1.0 m in the lowermost river valley sites (508, FAHY) to 1.0-1.5 m in the more landward sites (521, 523) (Peterson et al., 2012). Redrafted vibracore logs are from Peterson et al. (2012). The calibrated 14 C date (cal) in FAHY is projected from an auger gouge-core site (A15) in the central marsh area, as shown in Figure 6.

Historic Alterations of the Coquille River Estuary Floodplain Wetlands
Using historic wetland data (1857-1872), Patricia Brenner mapped as much as 10,100 hectares of floodplain wetlands, dominated by timbered swamp or wooded bottomlands, in the lower Coquille River valley (Ecotrust, 2019). A recent survey (1982) of the floodplains yielded only 5,700 hectares of wetlands, including emergent grasses, or upland prairies, swamps, and tidal/flood-discharge channels. The large decline in wetland area was largely due to extensive diking, draining, and filling for agricultural purposes. The natural floodplain wetlands in the lower Coquille River valley provide feeding habitat for migrating waterfowl and nursery habitat for endangered juvenile anadromous fish species, including chinook, coho, and steelhead, among others (ODFW, 1979;1998). Only about five percent of the inter-tidal wetlands in the Coquille Estuary remain undiked (USFWS, 2019a). Efforts are currently underway to rehabilitate available altered wetlands back to conditions favorable for wildlife habitat, such as at Winter Lake near Coquille, Oregon (The Nature Conservancy, 2019) ( Figure 2A). However, the Bandon Marsh and Ni-les'tun Unit restoration projects (2002)(2003)(2004)(2005)(2006)(2007)(2008)(2009)(2010)(2011) were amongst the first in the Coquille Estuary, after the Bandon Marsh National Wildlife Refuge was established by the U.S. Congress in 1983(USFWS, 2019a. Alterations of the Ni-les'tun Unit floodplain were made, beginning in the early 1900s, to improve seasonal pasturage for dairy and beef cattle grazing. Coquille Estuary channel-bank dikes (dredge-fill) and storm flood/tide gates, respectively, were installed along the Coquille Estuary channel bank and small tidal creek outlets ( Figure  3). Semi-diurnal tidal range (MHHW-MLLW) is ~2.0 m in the adjacent shallow (vertically-mixed) Coquille Estuary channel (~5 m thalweg depth) (ODFW, 1979). The smallest drainage channels in the central Ni-les'tun marshes were filled-in and rectilinear drainage ditches were constructed to divert flood waters to the floodgatecontrolled drainage ditches (USFWS, 2019b). A fill-dike was constructed across a northern portion of the floodplain marsh to permit vehicle access along the North Bank Ln roadway. The Ni-les'tun floodplain (~200 hectares surface area) is shown including the Ni-les'tun Unit and the Fahys Creek marsh, located on either side of the North Bank Ln roadway (bold line). Floodplain alterations include 1) in-filled drainage channels (lines), 2) riverbank dike (bold dashed line), 3) constructed drainage ditch (bold line), and 4) storm flood gates at tidal creek entrances (solid squares). The FAHY vibracore site (solid circle) is shown in the Fahys Creek marsh. The rehabilitated Ni-les'tun Unit is bounded to the north by hill-slopes and the North Bank Ln dike and to the south and east by the Coquille River estuary channel. Seasonally flooded cranberry bogs (top of map) are developed in late-Pleistocene dune sheet deposits. Holocene dune ridges (left side of map) barrage Fahys Lake, a window lake in the upland dune sheet aquifer. Headward eroded gullies (bold lines) locally drain the hill-slope aquifer (southward) to the adjacent Ni-les'tun floodplain. Lidar base map (2009) is from DOGAMI (2019b).

Preliminary Surveys for Wetland Restoration
Restoration planning for the Ni-les'tun Unit included vegetation and surface feature surveys (2002) to establish latest-prehistoric floodplain conditions. Low marsh (salt tolerant) vegetation (Frenkel et al., 1981), such as pickleweed (Salicornia), was found to occur locally in the central marsh area, some 300-400 m landward of the diked channel bank/tidal creek entrances ( Figure 4A). Perennial rush and sedge swamps (freshwater) were found to occur along the marsh back-edges that are adjacent to hillslopes and narrow gullies, and within the Fahys creek drainage outlet. These perennial swamps are developed in supratidal sites (>2.5 m elevation), which occur above the reach of the highest summer tides, but they are within reach of Coquille River/Estuary overbank floods ( Figure  4B). To reconcile these unexpected results, several hydrogeomorphic studies of the Ni-les'tun floodplain were initiated in 2003 and 2005, as outlined below. These studies concluded with preliminary water-quality testing of reconstructed floodplain channels in 2011 and 2012 (Beard, 2013).

Methods
Subsurface hydrogeomorphic studies in the Ni-les'tun Unit began in 2003 with GPR imaging of shallow floodplain deposits, including 1) remnant latest-prehistoric discharge/tidal channels, 2) approximate depths of GWS reflections, and 3) distinctive signal interruptions by subsurface salinity intrusion. GPR has been used in other peaty wetlands to establish depth to the GWS (Lowry et al., 2009). Fort this study, a Sensors and Software™ pulseEKKO 1000v GPR system was used with 100MHz antenna to collect GPR profile data, with ≥ 4 digital stacking at 0.5 m step intervals ( Figure 5A). Data post-processing with AGC gain control (Jol and Bristow, 2003) and calibrated mud and muddy-sand signal velocities of (0.07 m ns -1 ) yielded signal return penetrations of ~4 m depth subsurface. Ground-truthing of buried channel cut-and-fill features was performed with continuous auger gouge-coring to basal channel lag (woody debris) deposits. Auger gouge-core site elevations are from Lidar 2009 bare ground DEM (reported vertical error ± 0.1 m) (DOGAMI, 2019b). Part A. Photograph of the ground penetrating radar (GPR) system with 100 MHz antennae, as mounted on a carbon fiber trailer. Operators, riding on the all-terrain-vehicle, record GPR data and line segment positions from postprocessed global positioning system (GPS) (± 2 m estimated error). An odometer wheel triggered the GPR at rolling shot point step distances of 0.5 m. Part B. Photograph of a piezometer well, as used for monitoring 1) shallow groundwater water level (phreatic surface) and 2) water quality, including temperature, conductivity, pH, and dissolved oxygen (DO). Piezometers were top-sealed to assure that well waters were in equilibrium with insitu groundwater conditions during testing.
Following preliminary GPR mapping of GWS gradients and apparent subsurface salinity intrusions, a total of 10 bottom-screened point-tipped piezometers (1.5-2.0 m depth subsurface) were driven into tight-fitting auger holes (2005) for seasonal monitoring of shallow groundwater conditions ( Figure 5B). The piezometer exterior collars were sealed with mud/grout slurry. Piezometer tops were sealed with rubber clamp gaskets. Piezometer well heads were leveled-in to the NAVD88 datum by surveyors (vertical error ± 1 cm). Piezometer water levels were recorded by electronic water meter (phreatic surface depth ± 1 cm). Shallow groundwater hydrogeochemistry in sealed piezometer wells was measured with a micro-pump and constant through-flow cell (~15 ml min -1 ), using a ThermoOrion 635 MultiProbe and HP Pocket PC PDA data logger. The Orion multiprobe was used to measure temperature, conductivity, and pH. The multi-probe pH sensor was calibrated with two known pH buffer solutions of pH 4 and pH 7 at each piezometer site prior to well sampling. Dissolved oxygen was qualitatively estimated (minimum thresholds) with CHEMets DOK-7501 kits in the field. Late winter sampling occurred in early March due to site accessibility (overbank flooding) issues.
To better understand the potential shallow groundwater flow in the Ni-les'tun Unit floodplain, deeper core sampling was undertaken with a track-mounted geoprobe (model 6620DT). Geoprobe site elevations are from Lidar 2009 bare ground DEM (reported vertical error ± 0.1 m). Continuous core sections were recovered and archived in core liners (1.5 m length, 2.5 cm internal diameter) to logged depths of 5-8 m subsurface. Representative geoprobe core sections were sampled and analyzed for grain size, bulk density, porosity, and hydraulic conductivity (permeability) or 'K' (Dingman, 1994). Porosity and hydraulic conductivity were measured with a 1000 ml cylindrical constant-head permeameter (ASTM, 2006), using geoprobe samples that were homogenized and then remolded to initial core sample (in-situ) compaction or unconfined shear strengths.
Coincident with floodplain channel construction and Coquille Estuary channel bank dike and flood/tide gate removal 2009-2011(USFWS, 2019b, selected channel sites were tested for water quality parameters in 2011 and 2012 (Beard, 2013). Water quality testing was performed at 0.3 m water depths (near surface) with Vernier ™ sensors, including temperature, conductivity, pH, and dissolved oxygen. Channel bank site elevations are from Lidar 2009 bare ground DEM (reported vertical error ± 0.1 m).

Results
In this section we describe 1) morphostratigraphic studies of the latest-prehistoric floodplain deposits and incised discharge channels, 2) spatial variability of shallow aquifer hydrogeologic conditions, and 3) spatial variability of shallow groundwater hydrogeochemical parameters in the Ni-les'tun Unit floodplain.

GPR Profiling and Auger Gouge-Coring in The Ni-les'tun Unit Floodplain
The scale and distribution of latest-prehistoric (buried) channel features in the Ni-les'tun Unit were investigated by continuously-towed GPR profiling ( Figure 5A). The combined GPR track line distance (11.1 km), in 50 orthogonal profiles, yielded a total of ~22,200 shot points (0.5 m step distance) ( Figure 6). Signal return penetrations of ~4 m depth subsurface were sufficient to identify shallow channel cut-and-fill features. Analyzed channel parameters included 1) channel positions (distances in georeferenced track lines), 2) channel 'apparent' widths, relative to GPR line orientations, and 3) channel depths, as based on basal high-amplitude reflections. Paleo-tidal channels, located near the Coquille River bank (at least 6 in number), reached up to 35 m in width and at least 2.5 m in basal depth ( Figure 7A). No such larger channel features were found in floodplain interior areas. Smaller channel crossings (5-10 m in 'apparent' width and 1.0-2.5 m depth subsurface) were present ( Figure 7B), though uncommon (n=15 identified), throughout the floodplain interior ( Figure 6). The greater size of the paleotidal creek entrances along the Coquille channel bank are not unexpected, as modern tidal creek entrances typically widen and deepen at confluences with major tidal channels. In summary, the central interior of the Ni-les'tun Unit likely contained several dendritic networks of small, i.e., narrow and shallow, paleo-drainage/tidal creek channels prior to historic alterations. Interpreted GPR profile data include 1) shallow paleo-channel width (box length) and basal depth (m), 2) identified sites of GWS reflection with estimated subsurface depth (m), and 3) maximum landward extent of apparent signal truncation by subsurface salinity. Selected GPR targets were ground-truthed by shallow auger gouge-coring. GWS elevations on the Coquille Estuary channel levees, at the south ends of GPR lines 48 and 32, were from saturated sand depths in auger-gouge cores. Subsurface salinity intrusions in the vicinity of those lines prevented GWS reflection imaging by GPR. Two small paleo-channels near auger gouge-core site A7 were mapped in exposed Coquille Estuary cutbank sections, as salinity intrusion prohibited signal penetration in GPR line 9. In addition to ground-truthing GPR profile interpretations, auger gouge-coring was performed to constrain the stratigraphic development and relative age of the Ni-les'tun Unit. Auger gouge cores reached 1.0-3.0 m depth subsurface ( Figure 8). Recovered core deposits included mud and sand size fractions. Peaty mud or slightly rooted mud dominated the upper 0.2-0.5 m of most recovered core sections. These wetland topsoils were mottled by FeOHx mineralized root casts. The redoxy-morphic mottling (Birkeland, 1999) is indicative of variable GWS levels in the seasonally submerged floodplain setting. Underlying deposits included sandy mud, muddy sand, or sand. Odors of decomposing organics in some of the water-saturated basal sandy sections in sites A2, A3, A11 and A13, suggested locally reducing groundwater conditions in the shallow floodplain aquifer.
Unlike the protected marsh setting represented by vibracore site FAHY in the Fahys Creek marsh (Figures 2B and 3), the broader Ni-les'tun Unit floodplain is generally characterized by one preserved buried marsh (peaty-mud) horizon (Table 1). This shallowest buried peaty horizon is 14 C dated to ~290 cal intercept at 0.8 m depth in site A15, which corresponds to the last Cascadia megathrust regional rupture at AD1700 (Satake et al., 1996). The 1 st buried peaty mud, between 0.3 and 1.1 m depth in auger gouge-core sites, is widely distributed (n=13 sites) in the Ni-les'tun Unit. The widespread record of the AD1700 subsided marsh establishes a latest-prehistoric floodplain surface, of >0.3 ka in age, throughout the Ni-les'tun Unit floodplain. Only three core sites, A3, A8, and A9, include a next older buried peaty horizon, or 2 nd buried marsh, presumably corresponding to regional rupture events at either 1.1 ka or 1.3 ka (Darienzo and Peterson, 1995;Atwater et al., 2003). Therefore, most of the Ni-les'tun floodplain wetlands are relatively young (< 1.5 ka in age), leading to the modest record of channel dissections, as indicated by the GPR profiling ( Figure 6).

Figure 8. Floodplain elevation contours and selected gouge core logs in the Ni-les'tun Unit
Marsh surface elevation contours (0.5 m intervals) are averaged from bare ground Lidar 2009 DEM using the NAVD88 vertical datum in the Ni-les'tun Unit. The NAVD88 0 m datum is ~1.0 m below mean tide level (MTL) in the study locality. Auger gouge-core logs show that peaty mud generally overlies laminated mud and sandy mud, which both overlie muddy sand or fine sand. Basal channel lag consisted of sand, shell fragments and/or woody debris, as shown in core logs from sites A2, A4, A6, and A13. Saturated sandy deposits occurred below ~1 m depth in sites A3, A13, A14, and A15, and below ~1.5 m depth in sites A7, A8, and A11, thereby limiting sediment recovery by auger gouge-coring at greater depths. Strong orders of decomposed organics occurred at basal sand depths in sites A2, A3, A11, and A13. A buried peaty deposit at 0.7-0.8 m depth in site A15 was 14 C dated to 240±40 conventional and 290 cal intercept (see Table 1 for laboratory details).  Figures 6 and 7C). The shallowest GWS depths (~0.5 m) subsurface (n=9) were grouped towards the northern side of the floodplain, as bounded by the ancestral valley hillslopes and Fahys Creek marsh. The deepest GWS depths 1.5-2.0 m (n=4) occurred in the southern and southeastern sides of the floodplain, which are bounded by the Coquille Estuary channel. Summer GPR profiling revealed one other surprising relation of shallow groundwater conditions in the Ni-les'tun Unit. Abruptly terminated signal penetrations in the southeast margins of the Ni-les'tun floodplain ( Figure 7D) were attributed to subsurface salinity intrusions from the adjacent Coquille Estuary channel. Such abrupt signal terminations were tentatively traced to the central marsh areas in GPR lines 48, 32, and 39 ( Figure 6).

Piezometer Well Water Monitoring
A total of 10 sealed piezometer wells were installed in the Ni-les'tun Unit (Figure 9). The piezometers were monitored (2005)(2006) for shallow aquifer hydrogeochemical conditions that represent potential groundwater sources and/or water quality. Standing water depths (phreatic surfaces) were also measured in the piezometers to represent local depth of GWS in the floodplain (Dingman, 1994). Piezometer floodplain surface elevations ranged 1.91 m to 3.31 m elevation, and screened bottom depths range from 1.0 to 2.2 m depth subsurface ( Table 2). The Coquille Estuary (channel) and Fahys Creek, were also monitored for their hydrogeochemical conditions. During the winter sampling period (2006), standing or ponded waters on the floodplain surfaces were tested for comparable hydrogeochemical conditions at six of the piezometer sites (P3, P5, P6, P8, P9, and P10).  Piezometer, estuary channel, and Fahys Creek channel analyses included temperature, conductivity, pH, and minimum thresholds of dissolved oxygen (DO) ( Table 3). Well water temperature values ranged from 11.3-15.6° C (summer) to 8.5-16.5° (winter). Groundwater temperatures (moderate year-round) were not considered to be conservative tracers, so were not measured in all sites. Summer well water conductivities (n=4) showed a much wider range of values (2-16,000 μS cm -1 ) than expected, leading to a denser sampling strategy for the following winter (2006) survey. Winter water well conductivities (n=9) were 14-18,000 μS cm -1 . Well water pH values ranged from 3.10-7.80 (summer) to 2.53-6.80 (winter). The low ends of the groundwater pH values (pH ≤4.0) in piezometer sites P3, P8, and P9 and low dissolved oxygen (< 1 ppm) in sites P1, P2, P8, and P9, could represent potential stressors to juvenile salmonids (Geist et al., 2006) if those waters were to be intercepted by surface channel flows. The distributions of low-dissolved oxygen and pH in the northernmost piezometer sites point to origins of the 'low-quality' groundwater from hillslope runoff, interflow, and/or groundwater, as potentially discharged from the upland dune sheets. The dune sheet aquifers in the region are generally associated with reducing or low oxygen conditions (Brown and Newcomb, 1963;Easterly, 2005).  (Figures 8 and 9) received their salinity entirely from subsurface salinity intrusion and capillary rise of the saline groundwater in the fully enclosed (diked) central floodplain area. Therefore, the saline topsoil and seasonally ponded brackish water conditions in the supratidal settings of the central marsh do not represent submergence to intertidal levels, but rather subsurface salinity intrusion and capillary rise (summer evaporation) to supratidal elevations. Standing or 'locally-ponded' water depths on the floodplain surface were less than 0.25 m depth during the winter survey. Winter surface water conductivity values were 14-1,210 μS cm -1 , substantially lower than the piezometer groundwater conductivity levels measured in some piezometers (up to 18,880 μS cm -1 ). Winter surface water pH values were 5.09-6.51, clearly lacking the very-low pH values found in some of the piezometer groundwater analyses. Several of the winter surface water sites showed > 1 ppm dissolved oxygen, as expected, because surface water in full contact with the atmosphere should contain ≥8 ppm dissolved oxygen. In summary, shallow groundwater in the Ni-les'tun Unit is not derived from direct vertical infiltration of surface water, but rather from a variety of subsurface sources that differ greatly in conductivity, pH, and dissolved oxygen. Locally, the shallow groundwater in the Ni-les'tun Unit could compromise surface feature water-quality if sluggish channel flows or stagnant window ponds were to become recharged by low-quality groundwater, particularly during summer periods of low surface water replenishment.
To help establish the timing of shallow aquifer recharge in the Ni-les'tun Unit, the installed piezometers were monitored for seasonal variation in GWS depths (subsurface), which were converted to elevation (m NAVD88) using the piezometer ground surface elevations (Tables 2 and 4  Notes. Piezometer water heights, relative to the well head, were measured with an electronic water meter (± 1 cm error), then converted to the NAVD88 datum from the surveyed piezometer elevations. See Table 2 for piezometer positions. Due to marsh surface flooding, and corresponding inaccessibility of some piezometer sites, the fall sampling period extended into latest fall and the winter sampling period occurred during early March.

Characterization of Shallow Aquifer Hydrogeologic Units
To characterize the shallow aquifer under the Ni-les'tun Unit, a track-mounted geoprobe was used to recover continuous core samples for hydrogeologic analyses (Figures 10 and 11). Geoprobe site elevations (2.21-3.84 m) and logged bottom depths (5.0-8.0 m) are presented in Table 5. Rig mobilization was aided by elevated earth-fill roads and dikes in the marshy floodplain setting. Drive depths and core lengths were compared for each core interval (successive 1.5 m lengths) to account for loose sand compaction and/or rodding within some sections. Recovered core sections in 1.5m drive lengths were stored in sealed plastic core liners, prior to core logging and subsampling.  (Table 1). Representative core samples, 30-50 cm in length, were homogenized and then remolded for hydrogeologic lab testing. Locations of geoprobe sites (G9 and G3) are shown below in Figure 11.
The shallow aquifer under the Ni-les'tun floodplain surface is generally characterized by 1) muddy units in the uppermost 1.0-1.5 m of depth, 2) underlying units of either sandy mud or muddy sand, 3) basal sand at 3-8 m depth, and 4) non-contiguous layers of mud at depths below 3 m subsurface in several sites (G1, G2, G5, and G11) ( Figure 11). The presence of the deeper mud lenses and an absence of sand/mud accretionary bank laminations, argues against floodplain accretion by major subtidal channel migration. Rather the floodplain appears to have developed from 1) the vertical accretion of a subtidal sand flat to an intertidal mud flat, and then 2) the progradation and episodic vertical accretion of a supratidal floodplain. Shallow buried marsh deposits in sites G3, G5, and G12 confirm the emergence of the barren tidal flat deposits to vegetated wetlands prior to 0.3 ka, or the burial age of the 1 st buried marsh at AD1700 (Satake et al., 1996). Only one geoprobe site (G12) contained evidence of a 2 nd buried marsh, presumed to be ~1.1 or ~1.3 ka in age, as suggested above in Section 4.1. jgg.ccsenet.org Journal of Geography and Geology Vol. 12, No. 1; 2020 Figure 11. Representative geoprobe sites and core logs in the Ni-les'tun Unit Representative geoprobe site positions are shown in the Ni-les'tun Unit. Corresponding core logs, reaching 5-8 m depth subsurface, demonstrate sediment textures or dominant grain sizes in the shallow aquifer. Sand grain-locking with core liner in some geoprobe cores lead to rodding and missing sections (core breaks). Recovered sections were used for hydrogeologic testing, including analyses of grain-size fractions, bulk density, porosity, and hydraulic conductivity (K). Mud is defined as clay and silt (<0.062 mm) of at least 90% volume fraction, as calibrated by in-lab rapid settling/decanting of sand fraction volume. Gravel is defined by wet sieving at > 2mm size fraction. Intermediate textural compositions are approximated by major and minor volume components, where the adjective is the minor component (Folk, 1980). Notes. Geoprobe coordinates are in UTM 10N meters (m). Ground surface elevations are in m NAVD88. Basal core depths (m) are to logged depths, as corrected for compaction.
Representative core sample analyses of grain size fractions, bulk density, porosity, and hydraulic conductivity in the Ni-les'tun Unit shallow aquifer are shown in Table 6. Gravel is rare in the shallow aquifer, which is dominated by sand (average 81 wt %, n=14). The mud size fraction (<0.062 mm screen size) makes up a significant minor component (average 14 wt %). The shallow aquifer has a large porosity (average 41 volume %) and moderate hydraulic conductivities (K average=29 cm hr -1 ). However, near surface mud layers, such as in G4 at 0.4-0.9 m depth and in G1 at 0.2-0.6 m depth, result in very-low hydraulic conductivities (K=0.002-0.004 cm hr -1 ). These impermeable topsoil mud layers could account for the observed localized ponding of surface water during wet jgg.ccsenet.org Journal of Geography and Geology Vol. 12, No. 1; 2020 39 winter periods. Relatively high values of hydraulic conductivity (K~20-140 cm hr -1 ) in underlying sandy units should not substantially restrict lateral groundwater flows in the shallow aquifer, as shown in sites G1 at 2.1-2.6 m depth and 6.4-7.0 m depth, G2 at 4.6-5.1 m depth, G3 at 5.4-5.9 m depth and 6.3-6.8 m depth, and G11 at 2.3-2.5 m depth. Notes. Sediment core samples from geoprobe sites (G) at subsurface depths (cm) were analyzed for grain size fractions by wet sieving (weight %), bulk density in ring samples (g cm -3 ), and porosity (%) and hydraulic conductivity or K (cm hr -1 ) in 1000 ml cylindrical constant head permeameters.

Discussion
In this section, relations are presented between 1) shallow aquifer development, 2) sources of groundwater supply, 3) patterns of groundwater flow, and 4) initial analyses of rehabilitated channel water quality in the Ni-les'tun Unit floodplain.

Hydrogeomorphology of The Ni-les'tun Unit Floodplain
The complexities of shallow aquifer dynamics in the Ni-les'tun Unit floodplain derive from 1) multiple recharge sources, 2) stratification of shallow hydrogeologic units, and 3) incised drainage channels that intercept the shallow hydrologic units. In addition to the expected Coquille River/Estuary overbank flooding and infiltration of meteoric water during wet winter periods, the floodplain aquifer is also supplied with water from adjacent hillslope discharge and subsurface intrusion of estuarine saline water (Table 3). The hillslope fresh-water supply is concentrated at the Fahys Creek drainage, which enters the floodplain at the Fahys Creek marsh ( Figure 12A). Localized sediment accumulations from the Fahys Creek distributaries yield a topographic mound reaching ~1.5 m above the central marsh surface (Figure 8). Subsurface intrusion of saline water from the Coquille Estuary channel occurs along the southeast margin of the floodplain, where the main estuary channel (~ 5 m thalweg depth) intercepts the floodplain aquifer. A relatively rapid vertical accretion and progradation of the Ni-les'tun Unit floodplain from a sandy tidal flat to a muddy top soil occurred between 0.3 and ~1.1 ka (Figure 11), which is in accordance with reported fish weir 14 C dates of 0.6-0.9 ka on the Coquille Estuary channel banks of the Ni-les'tun Unit floodplain (Byram, 2002). The general lack of deeper subtidal paleo-channels in the Ni-les'tun Unit likely results from the young age of the vertically accreted floodplain. Whereas the near-surface muds (0-1.5 m depth) in the Ni-les'tun Unit have very low hydraulic conductivities (K<0.1 cm hr -1 ), the underlying muddy sand/sand units (2-6 m depth) yield moderate hydraulic conductivities (K~10-100 cm hr -1 ) ( Table 6, Figure 12B). The very-low hydraulic conductivities of near surface mud deposits likely limit or delay vertical infiltration, leading to increased floodplain discharge through surface runoff and/or drainage channels, during wet winter months. The moderate hydraulic conductivities of the underlying sandy hydrogeologic units permit lateral groundwater flows, extending 1-2 km distance (north-south) under the floodplain surface. The subsurface salinity intrusion is likely forced at the Coquille Estuary-floodplain margin under high-tide-level hydraulic heads (Carol et al., 2012), but continues landward by density flow or by lower GWS levels in the central marsh, as lowered seasonally by summer discharge/evaporation (Table 4). The bottoms of small paleo-channels (~2 m depth subsurface) could have intercepted the sandy hydrogeologic units, assuring exchanges between groundwater and surface channel water. Prior to historic alterations of the Ni-les'tun Unit, the paleo-channels crossing the floodplain interior could have 1) been recharged by groundwater during dry summer periods, but then 2) served as sources of localized surface water recharge to the underlying hydrogeologic units during wet winter months of abundant surface water supply.

GWS Surface Trends, Salinity Tracers, And Predicted Flow Asymmetries
GWS elevations from piezometers (Table 4) and selected GPR reflections of GWS ( Figure 6) are used to contour GWS trends (summer/fall) in the Ni-les'tun Unit floodplain ( Figure 13A). Shallow groundwater is assumed to jgg.ccsenet.org Journal of Geography and Geology Vol. 12, No. 1; flow perpendicular to the GWS contours, from Fahys Creek marsh (~3.0 m elevation) to the Coquille Estuary channel (~1.0 m elevation or MSL). The corresponding GWS gradients (0.001-0.002) are oriented northwest to southeast on the western side of the Ni-les'tun Unit. However, piezometer well conductivities indicate a reversed direction of groundwater flow on the southeastern side of the floodplain. Summer well water conductivities decreased from 16,000 μS/cm at the channel levee site P6 to 4,130 μS cm -1 at the landward site P8 (Table 3). The well water conductivities remained relatively high during the following winter (P6, 18,783 μS cm -1 and P8, 3,763 μS cm -1 ), demonstrating continued subsurface salinity intrusion during conditions of abundant surface water supply. Surface water conductivities at piezometer sites P6 (1,210 μS cm -1 ) and P8 (280μS cm -1 ) were substantially lower than their corresponding subsurface counterparts, reflecting a surface discharge of freshwater across the floodplain surface during wet winter months. The relations shown in Figure 13 suggest two asymmetries of general groundwater flows in the Ni-les'tun Unit floodplain. A horizontal asymmetry exists between the western (freshwater dominant) and eastern (saline water dominant) sides of the Ni-les'tun Unit ( Figure 13A). A vertical asymmetry is interpreted to exist between near surface flow (freshwater discharge) to the Coquille Estuary channel, and deeper subsurface flow (salinity intrusion) landward, under the Ni-les'tun Unit. These two asymmetries in the shallow groundwater flows are represented by a four-part flow model ( Figure 13B), including both the horizontal and vertical flow asymmetries. Seasonal variations in groundwater recharge could add to the spatial complexities of groundwater influences on rehabilitated channels in the Ni-les'tun Unit, as discussed below. jgg.ccsenet.org Journal of Geography and Geology Vol. 12, No. 1; 2020

Estimated Seasonal Recharge Volumes
Estimated values of Ni-les'tun shallow aquifer recharge are based on 1) averaged GWS heights above mean sea level (1.00 m NAVD88) from monitored piezometers (2005)(2006) (Table 4), 2) average porosity (41 volume %) (Table 6), and 3) estimated Ni-les'tun floodplain surface area (200 hectares). The summer recharge volume (0.10x10 6 m 3 ) (Table 7) is supplied by 1) groundwater and Fahys Creek base flow from the ancestral valley hillslopes, sourced from the large dune sheet aquifers ( Figure 2) and 2) from subsurface saltwater intrusion from the Coquille Estuary (Figures 12 and 13). Such complex relations between freshwater supply from dune aquifers and subsurface saltwater supply from a marine source have been reported for other coastal systems (Acworth et al., 1998). In the Ni-les'tun floodplain the fall recharge volume (0.48x10 6 m 3 ) represents additional water supply, i.e., above base flow, from direct precipitation on the marsh and from hillslope surface runoff and/or interflow. Fall precipitation on the Ni-les'tun marsh (~40 cm in the study region) (NOAA, 2019a) could account for nearly 2x of the net fall recharge. The winter recharge volume (0.38x10 6 m 3 ) represents additional direct precipitation on the marsh, and hillslope runoff, and Coquille River overbank flooding. For example, the Coquille River gauge height at Coquille, Oregon ( Figure 2)  However, direct precipitation on the Ni-les'tun marsh (~90 cm in the study region) during the winter period could account for >4x the winter net recharge (above the fall recharge). Even so, sustained elevated conductivities of piezometer groundwater 3,763-18,783 μS cm -1 in the southeastern part of the Ni-les'tun Unit were measured during the winter period (Table 3). Therefore, much of the surface water supply to the Ni-les'tun marsh, from direct precipitation and/or Coquille River/Estuary overbank flooding, did not infiltrate into the shallow groundwater aquifer due to likely subsurface saturation, but rather it drained from the floodplain surface to the Coquille Estuary channel by surface/drainage channel discharge during winter months ( Figure 13B). Notes: Average groundwater surface elevations (GWS elev. m) are from seasonally monitored piezometers (2005)(2006) (Table 4)

Water Quality Testing in The Constructed Channel Network
Following the initial GPR profiling in 2003 it was recommended that rehabilitation plans for the Ni-les'tun Unit include a network of well-drained branched (dendritic) tidal channels in the central marsh area. Constructed channels were not recommended for the peripheral northern marsh areas where auger-gouge coring indicated reducing groundwater conditions. Such groundwater sources could be low in pH and dissolved oxygen, as was later confirmed in piezometer well water testing. Inter-channel depressions or window ponds were not recommended for the southeastern marsh areas where GPR profiling indicated subsurface salinity intrusion. Saline groundwater conditions in those areas were later confirmed by conductivity measurements of piezometer water samples. Patches of saline tolerant vegetation (Salicornia) in small depressions indicated saline capillary rise during periods of summer evaporation. Such saline conditions could lead to saline toxicity in the broader marshgrassland top soils and/or shallow window ponds (Beyen and Meire, 2003). In consultation with wetland hydrologists from Ducks Unlimited (USFWS, 2019b), the Bandon Marsh National Wildlife Refuge staff designed a network of reconstructed channels in the Ni-les'tun Unit ( Figure 14). The channels were connected to Coquille Estuary tidal flow through several ungated tidal creek entrances ( Figure 15). Rehabilitation of the drainage/tidal creek channels in the Ni-les'tun Unit (2009)(2010)(2011) provided an opportunity to test initial water-qualities in the rehabilitated channel network, on a seasonal basis (Beard, 2013). Channel water sampling sites were positioned 1) near previously installed piezometer sites, 2) in marginal areas of expected groundwater input, and 3) at the Coquille Estuary channel bank and in the Fahys Creek distributary channel. Channel bank elevations ranged 1.4-3.2 m and averaged 2.3 m NAVD88 (Table 8). Channel depths ranged 1.5-2.5 m depth below bank surfaces, placing channel thalwegs within intertidal flow (0-2 m elevation or 0±1 m mean tidal level), following channel network connection to the Coquille Estuary.   Figure 15. Channel water quality testing sites in the Ni-les'tun Unit Channel water-quality testing sites were occupied in the summer (2011) and winter (2012) following completion of channel connections, removal of Coquille River bank dikes, and removal of storm/tide gates from tidal creek entrances in the Ni-les'tun Unit floodplain (Beard, 2013). Water quality parameters were monitored in previously installed piezometers in 2005 and 2006 (Table 3).
Following completion of the extended channel network, and removal of storm/tide gates at channel entrances in the Ni-les'tun Unit in 2011 (Figure 15), the newly constructed channels were tested for initial water-quality conditions in summer 2011 and winter 2012 (Beard, 2013). Channel water-quality analysis was done at near surface levels (0.3 m water depth) during flood slack water tidal stages to screen for potential low water-quality values. Channel water pH and dissolved oxygen values could decrease in channel bottoms, relative to the near surface sampling levels used in this study, due to potential groundwater exchange and/or thermohaline stratification. The near surface water levels analyzed in these preliminary surveys were thought to represent relatively well-mixed units within the channel network reaches. Analysis at different channel depths and at different tidal stages for channel water-quality testing is warranted for future extended surveys.
The channel summer water-qualities varied substantially between sites in the Ni-les'tun Unit, with relatively low dissolved oxygen (≤ 5.5 mg L -1 ) at sites C1, C1Bs, C6, C6.2, and C8, and high salinity at sites C2, C5, C6, and 10A (Table 9). Strong channel flow mixing, or flushing, could alleviate some of the summer water-quality issues, but poorly mixed channel bottoms or isolated window ponds could intercept low quality groundwater conditions during summer periods of low surface-water recharge. Channel winter water-quality was dramatically improved overall, with only two channel sites (C1B and C2) falling below 5.0 mg L -1 dissolved oxygen (Table 10). In summary, most winter channel waters were of sufficiently high-water-quality to support over-wintering juvenile salmonids (Geist et al., 2006), but some summer channel-waters could pose problems to wildlife, depending on the species' water-quality sensitivities.
Three interior channel sample sites (C2, C5, and C6) demonstrated significantly higher conductivities (3,350-5,730 μS cm -1 ) than the nearby Coquille River/Estuary sample site C3 (261 μS cm -1 ) during the same winter sampling period (Table 10). Twice-daily tidal flow exchanges with the Coquille Estuary during winter periods of very-low estuary water conductivity, as represented by site C3, do not account for the brackish water compositions in the interior channel sites C2, C5 and C6. The channel waters in those reaches reflect significant inputs from saline groundwater, as represented by piezometer sites P6 (18,780 μS cm -1 ) and P7 (3,840 μS cm -1 ) ( Table 3 and Figure  15). In the western part of the Ni-les'tun Unit, a low value of summer conductivity in channel site C9 (1,230 μS cm -1 ) falls well below conductivities in channel sites located down-channel (C10a at 33,510 μS cm -1 ) and upchannel (C1B at 4,060 μS cm -1 ) of the intermediate site C9 (Table 9). The anomalous low conductivity in site C9 during the summer period is attributed to the influence of fresh groundwater supply to the C9 channel reach from the northwest side of the Ni-les'tun Unit (Figure 13), as indicated by piezometer site P3 summer conductivity (<10 μS cm -1 ). In summary, the comparisons of water conductivities, as conservative tracers, between channel and nearby piezometer sites confirmed the input of local groundwater supply to the rehabilitated channels in the Niles'tun Unit, during both summer and winter floodplain/aquifer conditions.  Notes. The 2012 sampling period occurred in February. See Table 9 for sampling details.
The performance of the completed Ni-les'tun Unit channel project has been monitored for recruited wildlife since the project completion in 2012 (Silver et al., 2015;Brown et al., 2016). Like other tide-gate removal projects in the region (Souder et al., 2018), the recovery of the rehabilitated Ni-les'tun Unit wetlands has been successful in recruiting desired wildlife species. However, some water-quality issues arose during post-construction summer periods, leading to increased mosquito breeding in the rehabilitated wetlands. Modifications to the constructed channel network, including improved flushing of sluggish tidal creeks and better drainage of isolated inter-channel depressions, helped to alleviate the mosquito breeding problem (USFWS, 2015). Monitoring of the Ni-les'tun Unit floodplain recovery, constructed channel performance, and wildlife recruitment is ongoing. Monitoring results can be obtained from the Bandon Marsh National Wildlife Refuge, Bandon, Oregon.

Potential Climate Change Impacts
Understandings of the hydrogeomorphology of the rehabilitated Ni-les'tun Unit are necessary to predict 'mostlikely' responses of the floodplain wetlands to future climate change. The two expected drivers of future climate change in the Ni-les'tun Unit are prolonged summer droughts and sea level rise. Potential future warming and drying in the southwestern Oregon (Dalton et al., 2017) could cause longer summer drought conditions. Decreasing freshwater recharge and increasing evaporation could lead to 1) greater salinities in central marsh top soils, 2) lower GWS depths and corresponding channel water depths in peripheral supratidal channels, and 3) increasing thermohaline stratification of inter-tidal channels/ponds. Increases in topsoil salinity would be most problematic in the southeastern parts of the Ni-les'tun Unit if summer drought conditions were to increase with global warming. Decreases in bottom water-quality in the northern peripheral channels of the Ni-les'tun Unit could occur if channel water recharge becomes increasingly dependent on late-summer hillslope discharge of groundwater with low dissolved oxygen and/or low pH (Table 3).
Predicted sea level rise (possibly ~1 m within another 100 yr) (Kopp et al., 2014;Horton et al., 2014;Deconto and Pollard, 2016;Clark et al., 2016;Hansen et al., 2016) could have very different impacts on the rehabilitated Niles'tun wetlands. Rising tidal levels of 0.5-1.0 m and extended seasonal salinity in the Coquille Estuary could be expected to 1) initially increase the landward extent of the salt marsh in the Ni-les'tun Unit (Frenkel et al., 1981) and 2) increase the surface area of marsh tidal flooding in the Ni-les'tun Unit (Figure 8). The conversion of freshwater grassland to salt marsh could proceed as much by rising subsurface salinity intrusion as by enhanced tidal creek flow ( Figure 13B). Depending on the rate of submergence relative to sedimentation from Coquille River/Estuary overbank flooding, the interior flats could remain as upper-intertidal salt marsh or could convert to a lower-intertidal mud flat. The average net sedimentation in the Ni-les'tun Unit, above the 1 st buried marsh at AD1700 is 0.67 m (n=13) ( Table 1) for the last ~300 years, including the last ~100 years of logging in the Coquille River drainage basin. The Ni-les'tun Unit floodplain, generally ~2.25±0.25 m NADV88, or 1.25±0.25 m MTL (at present), could remain sparsely vegetated with ~1.0 m of relative sea level rise and 0.3 m sedimentation during the next century. However, with greater submergence rates and/or lower sedimentation rates, during a similar time interval, most of the Ni-les'tun Unit would convert to lower-intertidal mud flat. Under such a scenario, the higherelevations (>3.0 m NAVD88) and focused hillslope freshwater discharges to the Fahys Creek marsh (Figures 3  and 12A) could provide a small refugia for supratidal marsh habitat in the lowermost reaches of the Coquille Estuary.

Conclusions
The objectives of this study are to establish whether and/or how groundwater conditions could influence rehabilitated wetland habitats in the fluvial-tidal wetlands of the Ni-les'tun Unit floodplain. The shallow GWS elevations and hydrogeochemical compositions of the shallow groundwater flows were found to locally influence water elevations and water-quality in rehabilitated drainage/tidal creek channels during summer periods of little to no surface water recharge. Specifically, during dry summer periods, sustained groundwater flow from hillslope/creek drainages recharged supratidal channels in the northwestern part of the Ni-les'tun Unit floodplain. Smaller groundwater flows from subsurface intrusion of estuary channel water in the southeastern part of the floodplain probably kept intertidal channel bottoms from 'dying-out' during summer minimum ebb tidal levels. However, groundwater-quality compromised some channel reaches that depended on groundwater recharge during summer conditions of no surface water supply. For example, peripheral channels along the northern floodplain margins were depressed in pH and dissolved oxygen due to low-quality groundwater supply from upland dunesand aquifers. Though not evaluated in this study, subsurface salinity intrusion and capillary rise, during dry summer conditions, could increase salinity toxicity in isolated inter-channel depressions and/or window ponds in the southeastern part of the floodplain. Perhaps the most important result from this study is the recognition of the complex shallow-groundwater dynamics that underlie the Ni-les'tun Unit floodplain. Understandings of such shallow groundwater flow complexities in other fluvial-tidal floodplains could help designers to optimize for desired outcomes in future projects of floodplain wetland restoration.