Water Resources Research

Storm-driven groundwater flow in a salt marsh



[1] Storms can cause significant groundwater flow in coastal settings, but prior studies of the effects of storms on groundwater flow and transport have largely focused on very large storms and used salinity as a tracer. We have little information about the effects of smaller storms on coastal flow and how storm-induced variability affects key tidal wetlands like salt marshes, which may remain saline throughout a storm. Here we show that even the distant passage of a moderate storm can strongly increase groundwater flow and transport in salt marsh ecosystems and adjacent barrier islands. Groundwater monitoring and radium isotope tracer analyses revealed significant influx of saline creek water into the confined aquifer below the marsh platform, driven by storm surge. This pulse of fluids reached depths exceeding 5 m, and surge-enhanced tides propagated through the aquifer to affect flow in the upland >100 m from the creek bank. Groundwater discharge from the marsh varied significantly prior to the storm, doubling during inundating tides compared to a period of noninundating neap tides. Storm surge then caused groundwater discharge to decline ∼50% compared to similar inundating tides. Ra- and nutrient-poor creek water that entered the confined aquifer below the marsh was quickly enriched in nutrients and carbon, even on 12 h tidal cycles, so that nutrient discharge was likely proportional to groundwater discharge. Storm-related flow could also drive significant contaminant discharge from developed coastlines. The enhanced transport and variability observed here likely affected hundreds of kilometers of the coastline impacted by the storm.

1. Introduction

[2] Coastal groundwater flow exerts key controls on coastal water supply; affects the discharge of nutrients, metals, and carbon to the coastal ocean via submarine groundwater discharge (SGD) [Moore, 2010]; and strongly influences the ecology of coastal wetlands through controls on salinity [Morris, 1995; Gardner and Reeves, 2002]. Large storms dramatically alter hydraulic forcing in coastal settings, causing extensive alterations in groundwater flow and transport, but few studies have investigated the impact of small to moderate storms on coastal groundwater flow.

[3] Most studies of storm-influenced groundwater flow in coastal settings have focused on aquifer salinization caused by very large storms. Anderson [2002] reported significant salinization due to overwash as a major hurricane swamped a barrier island; the aquifer recovery period was more than 3 years [Anderson and Lauer, 2008]. Van Biersel et al. [2007] showed that saline storm surge contaminated freshwater aquifers by entering damaged water supply wells during Hurricanes Katrina and Rita along the U.S. Gulf Coast. Shallow atoll island aquifers required 1–2 years to recover following inundation during the passage of a category 5 tropical cyclone [Terry and Faulkland, 2010]. Smith et al. [2008] used salinity observations to document significant fluid mixing in the upper 2.5 m of sediments below a lagoon during the passage of Tropical Storm Tammy directly over the field site and Hurricane Wilma 180 km south of the site, in 2005. Storm-related changes in groundwater flow are not limited to large storms passing overhead, as demonstrated by Cartwright et al. [2004], who reported the effects of a distant storm that generated high waves (significant wave height 4.5 m), causing the freshwater-saltwater interface in a sandy beach to migrate approximately 5 m inland in the absence of high rainfall or overwash.

[4] These studies demonstrate that storms affect groundwater transport, but their reliance on salinity as a tracer means that they were unable to monitor storm-related groundwater flow in areas where salinity did not change. We argue that studies of the impact of storms on coastal flow systems must go beyond a focus on the freshwater-saltwater interface to include coastal ecosystems like salt marshes, which may remain saline throughout a storm. As discussed in greater detail below, enhanced transport during storms has the potential to affect nutrient cycling and ecological productivity in salt marsh ecosystems. Thus, we focus on storms as a source of variability in salt marshes, which have been overlooked in previous investigations of storms. We also suggest that there are significant impacts from the small to moderate storms that occur more frequently than major hurricanes.

2. Salt Marsh Ecosystems and Groundwater Flow

[5] Salt marshes are highly productive ecosystems that provide key habitat and ecosystem functions, including nutrient and carbon cycling. Groundwater flow plays an important role in regulating nutrient transport and salinity in salt marshes, which in turn strongly affect ecological zonation and productivity [Morris, 1995]. Groundwater discharge from salt marshes also exports nutrients from salt marshes to tidal creeks [Whiting and Childers, 1989], possibly impacting additional estuarine and coastal marine ecosystems.

[6] More than three decades of research has indicated that groundwater flow in salt marsh sediments is dominated by tidal forcing, evapotranspiration (ET) and precipitation, and, in some cases, discharge of fresh groundwater from adjacent forested uplands [Gardner, 1973; Hemond and Fifield, 1982; Dacey and Howes, 1984; Harvey et al., 1987; Nuttle, 1988; Gardner and Reeves, 2002]. Note that uplands in barrier islands are normally sandy, so that rainfall typically infiltrates rather than running off undeveloped uplands.

[7] Below the surface of the typical marsh of the southeastern United States is a thick (∼1 m) low-permeability marsh mud or peat layer overlying a confined sand aquifer [Hemond and Fifield, 1982]. Tidally driven flow through this layered system is focused in the confined aquifer below the marsh mud and propagates laterally away from tidal creeks [Harvey et al., 1987; Gardner, 2007]. Active tidal exchange is thought to reach less than 20 m from tidal creeks in the confined aquifer. This exchange is limited by permeability where fine-grained sediments dominate, or by the 12 h tidal period where higher-permeability sediments are present [Harvey et al., 1987]. Fresh groundwater discharging from adjacent uplands has been observed in confined aquifers below salt marshes [Tobias et al., 2001; Gardner and Reeves, 2002; Carter et al., 2008], but inundation by saline creek water at high tide typically maintains saline conditions in the surficial marsh muds. Flow within the marsh mud, then, depends on the balance between hydraulic head in the underlying confined aquifer, recharge at the marsh surface, and ET.

[8] Given this conceptual model, significant areas of salt marsh muds can exist beyond the presumed influence of either tides or freshwater flow in the underlying aquifer. Groundwater flow in these interior zones has been characterized as “stagnant” [Gardner, 1973], with flow on the order of mm d−1 driven only by evapotranspiration [Hemond and Fifield, 1982] and recharge from precipitation or tidal inundation. However, this conceptual model is largely based on studies that focused on calm weather conditions.

[9] Several modeling studies have also investigated flow in tidal wetlands [Hughes et al., 1998; Li et al., 2005; Wilson and Gardner, 2006; Werner and Lockington, 2006; Marani et al., 2006; Gardner, 2007; Cola et al., 2008; Lenkopane et al., 2009], but these have commonly relied on simple sinusoidal tidal signals and most have assumed homogeneous (nonlayered) sediments. Those that considered layered stratigraphy [Hughes et al., 1998; Gardner, 2007] focused on total fluxes across the sediment-water interface rather than flow patterns within the sediments. Prolonged high water levels and high rainfall during storms have the potential to strongly alter the forces driving groundwater flow in salt marshes and barrier islands, but information about variability in groundwater flow within real salt marsh ecosystems is quite limited. Here we show that the passage of a significant but moderate storm 200 km away significantly altered groundwater flow and transport in a salt marsh and adjacent barrier island.

3. Methods

3.1. Field Site

[10] Cabretta Island is a Holocene barrier island on the seaward side of Sapelo Island, Georgia (Figure 1). The average tidal range is 2.5 m, and average annual rainfall is 130 cm. Stratigraphy was determined to depths of 3–4 m by vibracoring along a transect extending from a tidal creek to the ocean, crossing a salt marsh, forested upland, dune field, and beach (Figure 2). Cores in the marsh revealed an organic-rich mud layer overlying a 1–2 m thick confining unit consisting of layers of sand and silt, which were in turn underlain by permeable sand. In the upland adjacent to the marsh (site TT4) surficial sands were separated from deeper sands by a ∼1 m thick confining unit at 3 m depth. Further seaward, sediments were primarily sandy with irregularly distributed mud layers. We focus here on flow and transport on the marsh side of the island, specifically on sites TT1-TT4.

Figure 1.

Cabretta Island, Georgia, and study transect (A–A′). Note the Tropical Storm Fay storm track.

Figure 2.

Stratigraphic cross section showing well locations. Screened depths are indicated by horizontal bars. A continuous sand layer below a confining mud layer in the marsh provides a hydraulic connection between the tidal creek and the upland. Sinusoids indicate average tidal range of 2.5 m; dotted line indicates maximum inundation of the marsh during Tropical Storm Fay based on hydraulic head observations at TT3.

3.2. Monitoring Wells

[11] Nests of groundwater monitoring wells were installed at seven sites along the transect by jetting or hand augering. Wells were constructed of 3.2 cm Schedule 40 PVC and screened below the water table at two or three depths in each nest (Figure 2). Well screens were 15 cm long in shallow (1 m) wells and 30 cm long in deeper wells. Screened intervals were packed with sand, and the holes were backfilled with bentonite at sites where the sediments failed to collapse around the riser. Closely fit (2.5 cm PVC) internal casings were installed in all wells to minimize wellbore storage and limit the volume of water exchanged across the well screen during tidal fluctuations. Casings extended far enough above land surface to prevent overtopping in the marsh. At TT7 the well was sealed with electrical tape to prevent inflow of seawater when the well was overtopped. Wells are designated by a site number followed by the approximate depth of the screened interval, i.e., TT1-5 refers to the 5 m well at site TT1. The site numbers increase with distance from the tidal creek. A tide gauge was not installed in Cabretta Creek until after Tropical Storm Fay had passed, but because the well transect is only 500 m from the southern end of Cabretta Island the water level in the creek is essentially the same as ocean water level.

[12] Data loggers were installed in the wells to record pressure and temperature at 20 minute intervals, but TT5 was not instrumented during August 2008, and the shallow well and data logger at TT7 were lost to beach erosion during the storm. Wells at TT3 and TT6 were instrumented with loggers that also recorded salinity. The salinity records from TT3 revealed either fouling (TT3-5) or, despite the internal casings, significant (5 parts per thousand (ppt)) artifacts related to tidal fluctuations, so we present salinities obtained from geochemical sampling instead. Salinities at TT6 did not vary by more than 2 ppt during the storm.

3.3. Radium Isotopes

[13] Naturally occurring radium isotopes have been used as marine tracers for decades; here, we review them briefly for the hydrologic community. As summarized by Moore [2010], Ra isotopes provide fundamental information on the interaction of sediments, groundwater, and estuarine waters. Within the decay series of uranium and thorium are four radium isotopes. Each derives from decay of a thorium parent, which is tightly bound to particles. Because radium daughters are mobilized in saline waters, sediments provide a continuous source of Ra isotopes to salty aquifers at rates set by the decay constants of the Ra daughters. The Th isotope activities in the sediments and the distribution coefficient of radium between the sediments and water determine the potential input of each Ra isotope to the water. With half-lives ranging from 3.66 days (224Ra) to 11.4 days (223Ra) to 5.7 years (228Ra) to 1600 years (226Ra), this quartet of isotopes provides powerful constraints on flow and mixing within coastal aquifers.

[14] The activity of each Ra isotope in saline groundwater reflects a balance between generation of the isotope, removal by radioactive decay, and groundwater transport, which may include import of Ra-poor surface water. Ra activities in surface water are lower than in groundwater because Ra generation in surface water is negligible. Variations in Ra activity in saline pore water indicate variations in transport, because generation and radioactive decay occur at fixed rates.

[15] The wells were sampled for Ra on 19 and 23 August and 8 September 2008, with the exception of TT4 and TT5, which were not sampled because of their low (<10 ppt) salinities. The wells were purged prior to sampling by pumping out two well volumes of water. Four L samples from each well and 26 L samples from the tidal creek (∼0.25 m depth) and the adjacent coastal ocean were collected for Ra analysis. Ra was concentrated onto a column of Mn fiber and analyzed according to the procedure of Moore et al. [2006].

3.4. Nutrients

[16] The wells were sampled for nutrients on 23 August and 8 September by purging two well volumes using a battery-powered peristaltic pump before collecting and processing fluids. Samples from the tidal creek (∼0.25 m depth) and the adjacent coastal ocean were collected as well. After purging, an initial 1 L sample was collected and its temperature, pH, and dissolved O2 concentration were determined using an Orion® high-precision temperature/pH meter and a galvanic oxygen (O2) sensor (model 842). These measurements were made immediately following sample collection to minimize potential changes following removal of the sample from the aquifer. Additional samples were then collected, processed, and stored for subsequent laboratory analysis. A 5 mL sample was injected into a He-purged, crimp-sealed headspace vial and acidified (with HCl) to determine the CO2 concentration. Another sample was collected into a prerinsed syringe and filtered (0.2 equation imagem), subsampled, and fixed according to sample type, then stored for the analyses described below. Five mL ammonium (NH4+) samples were fixed immediately with phenol in the field and stored at 4°C prior to analysis. A 20 mL sample was acidified with nitric acid for subsequent analysis of dissolved phosphate. A 30 mL sample for the determination of dissolved organic matter (carbon and nitrogen) and dissolved oxidized inorganic nitrogen (= NO3 + NO2) was stored frozen. All sample vials used in this work were acid-washed, rinsed with ultra pure water, and combusted at 500°C prior to use. Analyses were conducted according to the protocols described by Weston et al. [2006].

4. Results

[17] Tropical Storm Fay moved slowly northward along the east coast of Florida for 2 days before reaching its nearest approach, 200 km to the south of Cabretta Island (Figure 1). Rainfall, tides, and waves peaked on Cabretta Island on 22 August 2008, and the significant wave height at Gray's Reef Buoy, 30 km offshore, exceeded 3 m. Sapelo Island received 7.5–8.0 cm of precipitation on 21–22 August (Figure 3), which is equivalent to the cumulative average rainfall for the entire month. Observed and predicted tide data from the two nearest NOAA tide gauges were retrieved from http://tidesandcurrents.noaa.gov. The maximum storm surge recorded at St. Simons Island, Georgia, 35 km to the south (the nearest benchmarked tide gauge), was 1.2 m (Figure 4a). At Fort Pulaski, 85 km to the north, the maximum surge was 85 cm. Observed tides exceeded predicted tides for nearly 8 days, building slowly for 5 days (17–22 August) and then receding over a period of 3 days (22–25 August; Figure 4a). Note that the storm struck as tidal amplitudes were decreasing following spring tides.

Figure 3.

Precipitation for the month of August 2008 at the Marsh Landing station on Sapelo Island, 4.5 km from the well transect. Daily totals are based on 15 min data obtained from the National Estuarine Research Reserve System Centralized Data Management Office (http://cdmo.baruch.sc.edu).

Figure 4.

Tide and hydraulic head data from Tropical Storm Fay. All data are relative to mean sea level (MSL). (a) Tidal record from the NOAA tide gauge at St. Simons Island, Georgia. Storm surge (observed minus predicted tide) increased gradually for 5 days (17–22 August) and then receded over a period of ∼2.5 days (22–25 August). (b) Hydraulic head at TT1, near the creek bank. (c) Hydraulic head at TT3 (Salicornia zone). (d) Hydraulic head at TT4 (upland). Hydraulic head in the confined aquifer at TT4-4 fell rapidly following the peak of the storm and stabilized around 26 August about 10 cm above prestorm averages, suggesting a rapid response to the retreat of storm surge in the confined aquifer and slower discharge from the unconfined upland aquifer.

[18] The hydraulic head records from the period surrounding Tropical Storm Fay are shown in Figure 4 for six representative wells that describe flow on the marsh side of the island. The records from TT2 are intermediate between TT1 and TT3, and the 5 m wells at TT1 and TT3 differ from the 2.5 m wells by no more than 4 cm over the period shown. In these records, inundation by the tide is indicated by sharp peaks in hydraulic head, which reflect the weight of the overlying tide water. These peaks are most visible in the 1 m wells at TT1 and TT3. The marsh was not inundated during the neap tides of 10–12 August, and TT4 was never inundated. During periods when TT3 was inundated, the hydraulic head at TT3-1 fell rapidly after the tidal peak until the site was exposed, then remained constant at the elevation of land surface for some time. This indicates very slow drainage through the marsh mud at this site. The height of the peak in hydraulic head during inundation in the shallow wells is a good indication of the actual height of the tide, particularly during periods when the marsh was fully inundated twice a day. We compared the St. Simons tide gauge data to the peak hydraulic head data from TT1 during inundation to confirm that St. Simons was a reasonable proxy for Cabretta Creek water levels. We found that, adjusting for a lag, the water level in Cabretta Creek is highly correlated to the water level at the St. Simons tide gauge, with differences in amplitude of 3 cm or less except during the first high tide on 22 August, during the peak of the storm, when the difference approached 8 cm. The tides during which the amplitude differed did not correlate with storm surge.

[19] The hydraulic head records from the confined aquifer predictably show significant damping of the tidal signal with increasing distance from the creek. The large tidal fluctuations observed at TT1-2.5 demonstrate that the aquifer is well connected to the tidal creek (Figure 4b). Tidal fluctuations are much smaller 60 m inland at TT3-2.5 (Figure 4c). Tidal fluctuations of essentially the same magnitude, although lacking the peaks associated with tidal loading, also appear in the confined aquifer below the upland (Figure 4d). Tidal fluctuations do not affect the overlying unconfined aquifer at TT4.

[20] The hydraulic head records also indicate differences in groundwater flow over the spring-neap tidal cycle. When the marsh was not inundated (10–12 August) the groundwater flow direction in the marsh mud near the creek oscillated with the tide, i.e., upward during high tide and downward during low tide (Figure 4b). At TT3, ET during a series of noninundating neap tides allowed the head in the mud to fall below the head in the underlying confined aquifer (Figure 4c). The resulting upward flow, combined with continued ET, explain the high salinities of the Salicornia zone. When the tidal amplitude increased, the marsh became inundated at high tide, and groundwater flowed downward through the muds (Figures 4b and 4c). Horizontal gradients between the nests, which are particularly evident when comparing hydraulic heads at TT1-2.5 with any other well screened in the confined aquifer, indicate lateral discharge toward the creek at low tide throughout the spring-neap cycle.

[21] Groundwater flow during the passage of the storm differed significantly from both the neap and spring tide periods that preceded it. Hydraulic head increased significantly throughout the marsh and upland in response to high tides and rainfall during Tropical Storm Fay; these increases were independent of the tidal stage (Figure 4). This rise indicates an increase in groundwater storage throughout the system but particularly in the unconfined upland, where the water table rose roughly 50 cm over 2 days (21–23 August; Figure 4d). The marsh could not accommodate as large a volume of additional water as the upland, because the water table in the marsh was already near the land surface. Note that 8 cm of rain fell during the storm, whereas the upland water table rose 50 cm. For 8 cm of rain to cause a 50 cm rise in the water table, the unconsolidated sands of the upland would have to have a specific yield of only 0.16, which is possible for this type of sediment. A specific yield of 0.25 would explain only 32 cm of the rise. Therefore, at least some of the rise in the upland water table could be attributed to the influx of water into the island from the marsh because of storm surge. It is not possible to discern from the current data exactly how much of the additional rise at TT4 can be attributed to flow from the marsh or from the beach, but hydraulic heads at TT3 and TT6 (not shown) exceeded the head at TT4, so flow from both directions could have contributed to the rise. Note, however, that TT4 was only ∼30 m away from the high-tide line on the landward side of the island during maximum inundation from Tropical Storm Fay (Figure 2). The high-tide line on the seaward side of the island was never closer than 50 m to TT4.

[22] Hydraulic gradients and groundwater flow directions changed markedly during the storm (Figure 4). Vertical hydraulic gradients that oscillated (TT1) or indicated upward flow (TT3) during the noninundating neap tides were replaced during the storm with an extended period of downward flow (recharge), although the vertical gradient in hydraulic head was reduced during the storm compared to normal spring tides. The most striking storm-related change occurred at TT4, where vertical gradients began to oscillate with the tide during the peak in storm surge. Tides commonly exceed the elevation of the water table in the upland, but the tidal signal is normally damped so much by the time it reaches the upland that downward flow across the confining unit persists at TT4 even at high tide. The long period of storm surge allowed the signal to propagate more than 100 m from the creek bank, indicating a unique role of storms and storm surge in marsh hydrology. Hydraulic gradients quickly returned to prestorm orientations once the storm surge receded, although the water table in the upland remained elevated at TT4 through the 8 September 2008, sampling.

[23] Although slight freshening was observed at all sampled wells on the marsh side of the island during the storm, salinities changed less than 4 ppt during the storm (Table 1). Salinity was not recorded at TT4 prior to the storm but increased only 1 ppt between 23 August and 8 September, from 4 to 5 ppt at TT4-2 and from 9 to 10 ppt at TT4-4.

Table 1. Radium Isotope Activitiesa
SampleDepthDateSalinity226Ra (dpm/L)228Ra (dpm/L)ex224Rab (dpm/L)223Ra (dpm/L)
  • a

    LT, low tide; nd, not determined; dpm, disintegrations per minute.

  • b

    ex224Ra = 224Ra − 228Th.

TT12.5 m19 Aug 200834.21.707.5312.60.24
 2.5 m23 Aug 200833.50.652.352.80.11
 2.5 m8 Sep 0833.41.607.7111.20.52
 5 m19 Aug 200838.33.242.366.90.29
 5 m23 Aug 200832.13.242.816.50.42
 5 m8 Sep 2008nd3.332.988.30.46
TT22.5 m19 Aug 200836.01.757.8815.60.82
 2.5 m23 Aug 200833.91.225.907.80.17
 2.5 m8 Sep 2008nd1.868.6114.70.70
 6 m19 Aug 200837.33.388.6916.50.58
 6 m23 Aug 200834.82.104.607.10.40
 6 m8 Sep 200836.83.608.7813.81.01
TT32.5 m19 Aug 200833.61.296.0612.00.50
 2.5 m23 Aug 200831.91.105.00ndnd
 2.5 m8 Sep 200834.71.416.349.60.34
 5 m19 Aug 200838.62.806.3910.70.91
 5 m23 Aug 200834.72.204.607.80.41
 5 m8 Sep 200838.33.928.1311.60.62
Cab CreekCab Creek19 Aug 200835.70.521.390.80.03
 Cab Creek LT23 Aug 200832.90.531.191.80.08
 Cab Creek LT8 Sep 200835.70.801.992.70.13
AtlanticAtlantic LT23 Aug 200834.00.891.280.30.05
 Atlantic LT8 Sep 2008nd0.350.870.90.05

[24] Radium isotope tracers provide additional information about groundwater flow during this period. For simplicity we focus on 224Ra and 228Ra activities here (Figure 5), but all Ra isotopes behaved similarly (Table 1). Note that the longer-lived isotopes, 228Ra and 226Ra, act as conservative tracers in this system because of their long half-lives (5.7 and 1600 years, respectively) and correspondingly slow regeneration time. Prior to the storm (19 August), Ra activity in the marsh groundwater greatly exceeded the activity in the creek for all four isotopes, with the exception of TT1-5, which is screened in the thick sand layer adjacent to the tidal creek and thus is likely to be regularly flushed by tidal fluctuations. Samples collected immediately following the storm (23 August) revealed significant decreases in 228Ra activity in all sampled wells, particularly at 5–6 m depth in the marsh (Figure 5), with declines of 30–55% at TT2. Ra activities returned to prestorm levels by 8 September, far too soon for the long-lived isotopes to regenerate.

Figure 5.

Radium activities from the marsh wells before (19 August), just after (23 August), and 2 weeks after the storm (8 September). (a) The 224Ra activity in the shallow (2.5 m) wells. (b) The 224Ra activity in the deep (5–6 m) wells. (c) The 228Ra activities in the 2.5 m wells. (d) The 228Ra in the 5–6 m wells. Ra activities declined significantly during the storm. N.D. indicates no data.

[25] Additional Ra samples were collected over a full tidal cycle in July 2009 to confirm that the large variations in Ra activities observed during the storm could not have been caused by normal tidal variations. We found no variations comparable to those observed during Tropical Storm Fay except at TT2-2.5 and at TT3 (Figure 6), and those variations did not follow an obvious tidal pattern. This suggests greater variability in Ra activities in those regions, which would mean that movement of relatively small volumes of groundwater, driven by either tidal forcing or sampling, could explain the observed variations. For example, TT2-2.5 is screened close to the confining muds, and higher-Ra fluids from the mud may have been drawn into the well by tidally influenced flow or by sampling. Regardless of the cause, this means that we can not be certain that storm-driven flow caused the Ra variations observed 19–23 August at TT2-2.5 and TT3. The dilution of Ra at TT1-2.5 and TT2-5 during Tropical Storm Fay significantly exceeded variations observed in July, which indicates that Tropical Storm Fay affected a substantially greater extent of the marsh than is reached by standard tides.

Figure 6.

The 224Ra activities over a full tidal period at nests (a) TT1, (b) TT2, and (c) TT3. LT, low tide; HT, high tide.

[26] Nutrient and dissolved carbon samples were collected from the wells on 23 August, when many had low 228Ra, and again on 8 September after the 228Ra had returned to prestorm conditions. These samples revealed high concentrations of dissolved inorganic nitrogen (DIN) (primarily NH4+), phosphorus, and carbon, and of dissolved organic nitrogen and carbon relative to Cabretta Creek and nearby surface ocean water (Table 2). These species also behaved differently from the Ra isotopes. For example, concentrations of DIN did not follow the severe 228Ra depletions noted on 23 August (Figure 7), but they clearly suggest a landward shift in the pattern of DIN concentrations at the peak of storm surge.

Figure 7.

Dissolved inorganic nitrogen concentrations for (a) 23 August (immediately following the storm) and (b) 8 September 2008.

Table 2. Groundwater Concentration of Organic and Inorganic Nitrogen, Inorganic Phosphorus, Organic and Inorganic Carbon, and 228Raa
DateSampleNOx (μmol)NH4 (μmol)DIN (μmol)PO4 (μmol)DIC (mmol)DON (μmol)DOC (μmol)228Rab (dpm/L)
  • a

    DIN, dissolved inorganic nitrogen; DIC, dissolved inorganic carbon; DON, dissolved organic nitrogen; bdl, below detection limit.

  • b

    Here nd indicates not determined, meaning sample size was insufficient (1 m wells) or sample was fresh (TT-4).

23 Aug 2008Cabretta Creek1.4731.733.22.731.4315.1347.91.19
8 Sep 2008Cabretta Creek0.6624.325.06.862.1322.2545.01.99
23 Aug 2008Ocean2.435.37.81.491.2610.6269.91.28
8 Sep 2008Ocean0.
23 Aug 2008TT1-10.1852.852.926.762.2218.7491.6nd
8 Sep 2008TT1-10.7455.856.62.342.2230.2650.1nd
23 Aug 2008TT1-2.50.0036.
8 Sep 2008TT1-2.52.0143.545.546.752.5013.5395.37.71
23 Aug 2008TT1-50.55162.4162.987.874.8336.3848.62.81
8 Sep 2008TT1-50.58196.0196.6111.154.9223.9952.42.98
23 Aug 2008TT2-10.0071.971.9103.563.3825.4663.9nd
8 Sep 2008TT2-10.00107.9107.98.852.9711.2839.9nd
23 Aug 2008TT2-2.50.0040.340.355.442.4713.8371.95.90
8 Sep 2008TT2-2.50.00293.6293.648.842.81bdl385.68.61
23 Aug 2008TT2-60.00349.2349.277.414.42bdl985.54.60
8 Sep 2008TT2-60.42402.3402.776.883.909.01133.38.78
23 Aug 2008TT3-10.51192.8193.327.532.9536.51088.4nd
8 Sep 2008TT3-10.15235.2235.38.263.1345.8969.2nd
23 Aug 2008TT3-2.50.00708.3708.3135.795.3317.21559.95.00
8 Sep 2008TT3-2.50.00825.6825.6132.665.34bdl1640.66.34
23 Aug 2008TT3-50.00464.5464.553.872.14bdl1053.04.60
8 Sep 2008TT3-50.69478.3479.054.025.445.41346.48.13
23 Aug 2008TT4-20.00175.6175.665.383.0227.01056.7nd
8 Sep 2008TT4-20.27222.3222.652.722.6833.41758.5nd
23 Aug 2008TT4-40.13343.8343.9185.065.9992.42495.7nd
8 Sep 2008TT4-40.00362.0362.0183.836.5574.03585.9nd

5. Discussion

[27] The decreases in Ra activities during the storm in the confined aquifer below the marsh require explanation. Enhanced recharge through the surficial mud layer is unlikely to have flushed Ra from the confined aquifer below, because, in addition to the low permeability of the mud, hydraulic head gradients across the mud layer declined prior to sampling on 23 August, rather than increasing (Figure 4). Ra could not have been flushed from the confined aquifer by discharge from the upland because that would presumably have caused significant reductions in Ra activities at TT3, which were not observed. Moreover, if either of these mechanisms flushed the Ra from the system, there is no simple explanation for the rapid return of the long-lived isotopes to their prestorm levels. Instead, the combination of hydraulic and Ra data indicate that storm surge allowed a significant volume of Ra-poor creek water to enter the confined aquifer via the creek bank. These fluids were then discharged following the storm, allowing existing Ra-enriched groundwater to return to the aquifer.

[28] The 228Ra isotope data were used to estimate the extent of mixing during normal 12 h tides and during tides enhanced by storm surge. Low Ra isotope activities at TT1-5 (Figures 5b and 5d) and its proximity to the creek suggest that it is regularly flushed by the tide. The volume of flushing during normal tides was estimated for this well assuming two-end-member mixing between the creek, ∼1.3 disintegrations per minute (dpm)/L, and groundwater from TT1-2.5, 7.5 dpm/L (Table 1). This yields a flushing efficiency of ∼80% at TT1-5 for normal 12 h tides. Flushing associated with the storm surge was estimated by comparing Ra activities in each well on 19 and 23 August, again assuming a creek end-member of 1.3 dpm/L. Flushing reached more than 25 m from the creek bank, with flushing of 83% at TT1-2.5 and 55% at TT2-6.

[29] It is clear that groundwater flow and transport changed during the storm, but it is important to determine how much these changes increased the overall groundwater flux through the sediments. This flux is most easily estimated by monitoring groundwater discharge from the creek bank. According to Darcy's law, this discharge is proportional to the hydraulic gradient between the tide and the confined aquifer at TT1, so we used the data presented in Figure 4 to calculate relative fluxes at 20 min intervals. We calculated the total normalized flux per tidal cycle by summing the 20 min relative fluxes over each tidal cycle, then normalizing the results relative to the higher of the inundating tides on 16 and 17 August, when storm surge was minimal. Groundwater discharge from the creek bank was significantly lower during noninundating neap tides (10–12 August) than during inundating tides (Figure 8a), consistent with the modeling results of Wilson and Gardner [2006], who showed that inundating tides generate larger fluxes than noninundating tides. Discharge increased by a factor of 2 following the inundating tide on 12 August.

Figure 8.

Groundwater flux and variations in the tidal signal. (a) Normalized groundwater discharge from the creek bank. (b) Observed tidal range (difference between high tide and low tide). (c) Time-averaged storm surge, calculated using a 12 h running average.

[30] Discharge declined significantly during the storm. We considered the possibility that tidal amplitude could explain the decline, but the tidal amplitudes during the storm were higher than during the prior neap tide period (Figure 8b). Moreover, the factor of 2 difference in discharge between neap and spring tides 10–15 August was controlled by inundation rather than tidal amplitude. The decline in groundwater discharge from the creek bank mirrors the rise in storm surge (Figure 8c). Groundwater discharge declined as storm surge drove water into the creek bank and returned to normal values as the storm surge retreated. Groundwater discharge was likely slightly elevated following the storm while the additional water stored during the storm exited, but the increase is small compared to the spring-neap variations that occurred 10–15 August. This result is consistent with the findings of Gu et al. [2008], who found that high-flood stage in a stream could decrease groundwater discharge to the stream. At Cabretta Island, the rise in mean water level reduced the net migration of existing Ra- and nutrient-enriched pore fluids toward the creek prior to the 23 August sampling, while also allowing the entry of new creek water. Predicted tides (http://tidesandcurrents.noaa.gov) for the second half of August would have inundated the site at least once a day, so groundwater discharge would likely have remained relatively constant throughout this period, if the storm had not struck.

[31] Nutrient and dissolved carbon concentrations observed on 23 August did not decrease to the same extent as radium concentrations, suggesting that the pulse of Cabretta creek water that entered the aquifer stimulated rates of microbial remineralization, thus maintaining nutrient and carbon concentrations at high levels relative to the creek. TT1-5 also maintained high DIN concentrations relative to the creek despite being regularly flushed by 12 h tides, which further supports the idea that increased groundwater flushing stimulates remineralization of organic matter. Thus, nutrient discharge likely decreased with groundwater discharge during the storm, but the longer-term impact of the storm on the marsh and on nutrient cycling is not yet understood.

[32] Our results suggest that Ra can be a very useful tracer in coastal groundwater systems, but our understanding of variations in Ra activities in groundwater remains incomplete. As previously indicated, variations in Ra activity at TT2-2.5 during a single day suggest that Ra is spatially variable over reasonably small distances in the subsurface. Longer-term variations also require additional study. For example, Ra activities in the deep (5 m) and shallow (2.5 m) wells were very similar in August–September 2008, but Ra activities 2.5 m wells were lower in July 2009. It is worth noting that the 13 July 2009 sampling was preceded by 8 cm of rain 6–10 July, 2009, accompanied by storm surge of 50 cm (estimated from NOAA Fort Pulaski, Georgia, and Fernandina Beach, Florida, stations; http://tidesandcurrents.noaa.gov), but a full analysis of the similarities and differences between that storm and Tropical Storm Fay is beyond the scope of the current paper.

[33] The findings presented here have significant implications for coastal development. The stagnant zone identified in midmarsh areas during calm conditions is clearly subject to enhanced flow during storms, and the degree of tidally influenced groundwater flow below the upland 100 m away from the creek bank indicates a strong connection below the marsh between developed, potentially contaminated uplands and tidal creeks. This is particularly relevant because poorly performing septic systems, rather than storm water runoff, may contribute the major portion of fecal coliform contamination in coastal surface waters [Cahoon et al., 2006]. The general suitability of barrier island sands for septic systems is limited, causing several authors to suggest that permitting policies and practices may not be sufficient in such systems [Cahoon et al., 2006]. A survey of coastal South Carolina and Georgia regulations reveals that required setbacks for septic systems adjacent to marshes are commonly 15 to 30 m. Our observations indicate that septic fields and underground storage tanks on barrier islands could experience significant flushing during storms, even if located 50 m from the marsh.

[34] We note that marshes that are not connected to uplands may behave very differently from the fringing marsh described here. For example, once a marsh island is completely submerged and saturated at high tide, very little additional water can enter. In the marsh at Cabretta Island, the marsh was fully inundated and saturated at high tide during Tropical Storm Fay, but water could continue to enter the creek bank because additional storage was still available in the upland, where the water table had not yet risen to the elevation of the tide. Marshes that become isolated from uplands by trenching or canal construction may lose this mechanism for periodic flushing of the marsh interior. Such engineering measures also leave potentially contaminated uplands with an even closer hydraulic connection to the new canal or trench.

6. Conclusions

[35] Moderate storms can drive significant groundwater flow and transport in coastal settings. Following the passage of a tropical storm 200 km to the south of a Georgia salt marsh, the influence of Ra-poor creek water was observed more than 25 m from the creek bank, at depths exceeding 5 m. This flow is much larger in scale than typical tidal flow and suggests that groundwater-related variability in the marsh interior is much more extensive than previously thought. The transition from noninundating neap tides to inundating spring tides prior to the storm increased groundwater discharge to the creek by a factor of 2. Storm surge then caused groundwater discharge to fall to levels similar to those of neap tides during the storm. Microbial activity in marsh muds appeared to be stimulated by the storm-induced flushing of the aquifer with oxygenated water, because creek water that entered the marsh was rapidly enriched in nutrients and carbon.

[36] Confined aquifers below typical salt marshes in the southeastern United States are capable of transmitting tidal signals much farther than unconfined beach aquifers. Although surface flooding is typically the major concern in developed settings, storm-driven groundwater flow has the potential to impact septic systems and underground storage tanks even in areas that are not inundated. Salt marshes that are not connected to uplands, whether naturally or through anthropogenic alterations, likely lack extensive storm-related variability in flow, because groundwater flow is minimal once they are inundated and saturated at high tide.

[37] Radium isotopes were again shown to be very useful tracers of groundwater flow in coastal settings, revealing significant fluid flow where simple tracers like salinity showed very little variation. Radium activities vary spatially and temporally in the subsurface, however, and have the potential to be affected by groundwater withdrawal when sampling.

[38] Finally, we note that the enhanced rainfall and surge associated with storms can extend hundreds of kilometers from the center of the storm. The variability observed on Cabretta Island was likely reproduced over at least a 300 km portion of the coastline during the passage of the storm.


[39] We thank D. Saucedo and J. Garbisch for assistance on Sapelo Island; K. Hunter and R. Styles for laboratory analyses; W. Sheldon for assistance with tidal data and predictions; and K. Hunter, B. Long, A. Mehrzad, and A. Hougham for their assistance in the field. This work benefited from discussions with J. Morris. We thank three anonymous reviewers and an associate editor, whose comments contributed to significant improvements in the manuscript. This material is based upon work supported by the National Science Foundation under grants 0711301 and 0711215 and by the Georgia Coastal Ecosystems Long Term Ecological Research program (OCE 06-20959). Any opinions, findings, and conclusions or recommendations expressed in this material are those of the author(s) and do not necessarily reflect the views of the National Science Foundation.