Introduction

Sea-level rise is one of the most often cited effects of anthropogenic climate change. Rising sea levels around the world have been documented over the last half century and as global warming continues over the next century, mean sea levels will continue to rise. The Intergovernmental Panel on Climate Change (IPCC) predicts that by 2099, average sea level will rise by 0.18–0.59 m for all future human activity scenarios (Meehl et al. 2007). Rising sea levels are of particular concern to small islands because of their limited ability to store fresh water in the form of surface streams, lakes and groundwater (Pelling and Uitto 2001; White and Falkland 2010).

Previous studies investigating salt-water intrusion of coastal aquifers due to sea-level rise include Tiruneh and Motz (2003) and Werner and Simmons (2009). Masterson (2004) models the effects of sea-level rise on four fresh-water aquifers in Cape Cod, Massachusetts, USA. A rising sea level in the model simulations causes water tables to rise, stream flows to increase, a decrease in the depth of fresh-water/salt-water interfaces, and the thinning of the fresh-water lens in the aquifers. A subsequent study by Masterson and Garabedian (2007) uses a simplified hypothetical model with conditions similar to Cape Cod to show that a 2.65 mm/year sea-level rise between 1929 and 2050 causes a fresh-water lens thickness decrease of 2% away from streams and 22–31% near streams.

The intent of this study is to further investigate the relationship between climate change and salt-water intrusion in a small sandy island aquifer. This study applies two climate-change scenarios, based on the IPCC 2007 report, to a simple variable density finite-difference groundwater flow model of Shelter Island, State of New York (NY), USA. The resulting movement of the fresh-water/salt-water interface is used to estimate changes in the fresh groundwater system. While similar studies (Masterson and Garabedian 2007) have investigated sea-level rise, this study also includes estimated changes in precipitation which have a substantial influence on aquifer conditions. Likewise, although the hydrogeologic conditions of Long Island are similar to Cape Cod (Schubert 1999), the Shelter Island study includes a bottom-restricted aquifer which has a unique effect on aquifer behavior.

Study area

Shelter Island is located at the eastern end of Long Island (Fig. 1). It has an area of 31 km2 and, like many small islands, has very little fresh surface water (Soren 1978). The inhabitants are dependent on the island’s fresh-water aquifer as a source of potable water.

Fig. 1
figure 1

Water-table contour map of Shelter Island. The model is based on transect C–C′ (modified from Schubert 1999)

Full size image

The hydrology of Long Island was first described by Veatch et al. (1906). More extensive hydrology studies were subsequently conducted by Cohen et al. (1968), Franke and McClymonds (1972), and Nemickas et al. (1989). In a hydrogeologic investigation of Shelter Island conducted by Soren (1978), fresh groundwater was found to be restricted to a thin geological layer known as the Upper Glacial aquifer which was susceptible to salt-water intrusion. The Upper Glacial is composed of upper Pleistocene glacial drift deposits that are primarily outwash sand and gravel along with clay, silt, and cobbles less than 15 cm in diameter. The drift is stratified and poorly-to-moderately sorted. A Pleistocene marine unit of gray and green clay underlies the entire island at a depth ranging from 20 to 30 m below sea level. The Pleistocene marine unit marks the lower boundary of the fresh-water aquifer (Soren 1978). Figure 2 shows a vertical section of the proposed hydrogeologic conditions.

Fig. 2
figure 2

Cross section C–C’ from Fig. 1 (modified from Schubert 1999)

Full size image

Although northern sections of the island have elevations over 30 m, the water table rarely exceeds 1.5 m above sea level. Likewise, the few ponds and marshes form primarily in depressions less than 1.5 m above sea level. There are no significant streams on the island. The position of the marine clay layer beneath Shelter Island deforms the island’s fresh-water lens (Fig. 3), preventing upconing, but increasing the lateral movement of salt water within the Upper Glacial aquifer. Thus, the salt-water interface deviates substantially from the Ghyben-Herzberg approximation (Simmons 1986).

Fig. 3
figure 3

Diagram of a a typical sandy island aquifer that follows the Ghyben-Herzberg approximation and b Shelter Island’s aquifer where the fresh-water lens is deformed by a clay layer. In both cases, z is the depth of the lens below sea level and h is the height of the water-table above sea level

Full size image

A groundwater budget performed by Schubert (1998) determined that groundwater discharge from Shelter Island to the surrounding Peconic Bay averages 48,700 m3/day. The rate is based on land use and pumping data from 1994 and precipitation data from 1959 to 1994. Water-table mapping indicates that the island’s irregular water table has two mounds that mimic the island’s topography and that groundwater generally flows radially outward to the surrounding shoreline (Fig. 1). In a study by Paulsen et al. (2004), field measurements of submarine groundwater discharge into West Neck Bay ranged from 0.02 to 2.3 m/day depending on location and tide. A borehole 90 m inland from West Neck Bay found the fresh-water/salt-water interface 20 m below sea level, with fine sand from the surface to 30 m below sea level, and marine gray clay below 30 m.

Schubert (1999) created a MODFLOW (Harbaugh et al. 2000) groundwater flow model for Shelter Island simulating a two-dimensional vertical section across Shelter Island in order to determine groundwater distribution, flow paths, and groundwater travel time. Borehole data were used to map the hydrogeologic characteristics of the vertical section (Fig. 2). Schubert (1999) found that almost all recharge left the groundwater system as shoreline underflow with an average flow age of less than 20 years and almost all flow less than 50 years. The small amount of flow that traveled through the Pleistocene marine clay layer and emerged as sub-sea underflow had an average age of about 1,800 years.

Shelter Island’s population was 2,228 in 2000 (US Census Bureau 2001) but summer seasonal population can exceed 10,000 (Simmons 1986). In 1983, by which time the population had stabilized, it was estimated that the total annual pumping for the island was approximately 2,700 m3/day (Simmons 1986). According to Franke and McClymonds (1972), about 85% of public water pumped is returned to groundwater in sections of Long Island without sewers. Only a small portion of Shelter Island has sewers and most water returns relatively close to where it was pumped (Schubert 1998). Assuming an 85% return rate, only 405 m3/day is lost to pumping. The only appreciable non-residential water use on Shelter Island is agricultural withdrawal which is estimated at 130 m3/day (Schubert 1998). Together these pumping rates are approximately 1% of the estimated 48,700 m3/day (Schubert 1998) of submarine groundwater discharge for the island. Thus, the average pumping rate currently has a minor effect on the island’s aquifer.

Because there are no significant streams on Shelter Island, precipitation that is not lost to evapotranspiration becomes recharge to the aquifer (Soren 1978). The climate of Shelter Island is classified as humid subtropical (Kottek et al. 2006) and precipitation is evenly distributed throughout the year (Nemickas and Koszalka 1982). However, measurements of chloride concentration by Simmons (1986) showed that the salt-water zone of diffusion moves seasonally. The water table is highest in early spring and lowest at the end of summer when some near-shore areas have less than 6 m of fresh water below them. Additionally, it was found that monthly variations in precipitation affected water-table fluctuations more than any climatic trends between 1974 and 1983. Miller and Frederick (1969) determined that the mean annual precipitation for Long Island was between 114 and 119 centimeters (cm). The Thornthwaite and Mather (1957) water balance method was used by Bart et al. (1976) to determine that evapotranspiration for the south fork of Long Island from 1930 to 1975 ranged from 41.7 to 62.5 cm with an average of 58.9 cm. According to Nemickas and Koszalka (1982), the average rate of evapotranspiration on Long Island is 50%, and the rate of overland runoff is 1%. Peterson (1987) also used a recharge rate of 50% as a good approximation. However, Steenhius et al. (1985) suggested an alternative method using a recharge rate between 75 and 90% from October 15 to May 15 and a recharge rate of 0% for the warmer summer months when evapotranspiration equals or exceeds precipitation.

Soren (1978) determined that Shelter Island sand has a horizontal hydraulic conductivity between 60 and 80 m/day. According to Smolensky et al. (1989), Long Island sand has a horizontal hydraulic conductivity of 80 m/day with an anisotropy ratio of 10:1. The Pleistocene Smithtown Clay on central Long Island has a vertical hydraulic conductivity of 0.01–0.02 m/day (Misut and Feldman 1996). Effective porosity is estimated to be an average of 30% (Franke and McClymonds 1972; Soren 1978). A study of the North Fork of Long Island by Schubert et al. (2004) used a horizontal hydraulic conductivity of 60 m/day for upper glacial outwash, a horizontal hydraulic conductivity of 25 m/day for upper glacial moraine, an anisotropy ratio of 10:1, and a vertical hydraulic conductivity for confining units of 0.1 m/day.

Model description

A variable-density finite-difference groundwater flow model was created for Shelter Island using the US Geological Survey (USGS) program SEAWAT (Guo and Langevin 2002; Langevin and Guo 2006) which was derived from MODFLOW (McDonald and Harbaugh 1988; Harbaugh et al. 2000) and MT3DMS (Zheng and Wang 1998), a solute transport application.

The model consisted of a cross-sectional area 37 m (120 ft) deep and 4,877 m (16,000 ft) wide cutting east–west across the island (Fig. 4). The model grid consisted of 16 vertical layers, each 2.3 m (7.5 ft) thick, starting from 1.5 m above mean sea level to 35.3 m below sea level to the top of the Pleistocene marine clay layer below which groundwater flow was believed to be negligible (Schubert 1999). The horizontal grid spacing ranged from 122 m (400 ft) off shore and inland to 15 m (50 ft) near the fresh-water/salt-water interface (Fig. 4).

Fig. 4
figure 4

SEAWAT model created for cross-section C–C'. Salt-water concentrations are 34.9 kg/m3 and the marine clay unit is inactive

Full size image

Model boundary conditions were based on the findings of Schubert (1999). The Pleistocene marine clay layer was set as a no-flow boundary to represent the bottom of the model. The salt water surrounding Shelter Island provided a constant head source or sink for any salt-water flow caused by movement of the salt-water/fresh-water interface. The water-table surface acted as the recharge boundary. The fresh-water/salt-water interface acts as a nearly impermeable boundary that moves under prevailing hydrologic conditions and discharge occurs near shore as submarine groundwater seepage (Schubert 1998). The edges of the fresh-water lens were initially delineated by the salt concentration in each cell. Fresh-water cells had a salt concentration of less than 0.25 kg/m3 (Simmons 1986) and saline groundwater cells were set to 34.9 kg/m3 (Masterson 2004). The difference between mean sea level and the National Geodetic Vertical Datum (NGVD) of 1929 was determined to be 0.12 m based on readings obtained from local tidal measurement stations (Schubert 1999).

Four hydrogeologic units were represented in the model (Fig. 4): (1) a poorly sorted mixture of clay, silt, sand, and gravel referred to as till, (2) upper glacial aquifer moraine and outwash, (3) a sand clay unit, and (4) a Pleistocene marine clay unit. Initial parameters were based on findings by Schubert (1999) and calibrated results are shown in Table 1.

Table 1 Parameters used for each geologic unit in the model. Hydraulic conductivities varied slightly at borders between hydrogeologic units
Full size table

In order to maintain model simplicity, several assumptions were made. First, it was assumed that diffusion played a minor role in the long-term movement of the fresh-water/salt-water interface. Molecular diffusion is proportional to the concentration gradient and the velocity of the diffusive particles which, in turn, is primarily dependent on temperature, particle size, and fluid viscosity (Bird et al. 1960). Since none of these characteristics, with the minor exception of temperature, should change within the study period, diffusion was ignored. Similarly, transverse mechanical dispersion was not used because the model was two-dimensional and solute concentrations were assumed to be uniform perpendicular to the cross-section of the model.

Most importantly, because this study was concerned with the long-term movement of the fresh-water/salt-water interface, the daily and seasonal movement of the interface and the additional mixing it causes was not considered. Simmons (1986) observed that the interface moves with the seasons and varying recharge. So it is important to note that the simulated movements of the fresh-water/salt-water interface were long-term averages that marked the new baseline from which regular season variability fluctuated. Accounting for both annual and seasonal variation in recharge was beyond the scope of the long-term study, especially considering that future climate change will have uncertain effects on extreme weather events and cycles. Thus, the 50% average recharge rate approximation (Nemickas and Koszalka 1982) was used for all simulations.

Calibration

Wells shown in Fig. 2 were used to calibrate the model’s hydraulic conductivities. Calibration was performed for March 1995 during which water levels were approximately 24% below long-term averages due to low recharge in 1994, which was 26% below average. Therefore, the model was calibrated using March 1995 well-head levels and a recharge of 50% of 1994 precipitation. Initial head conditions were set at 1.5 m for all variable-head cells. The transient SEAWAT model reached full steady-state equilibrium conditions within 20,000 days (approximately 55 years) which agreed with the flow path estimates of Schubert (1999). Compared to well-head observations, the calibrated model had a mean absolute error of 0.07 m.

Validation

Validation of the calibrated model was performed using additional well-head data obtained from the Suffolk County Department of Health which regularly monitors two of the wells used: well 51172 and well 51179 (Fig. 2). Upon review of the well-head data and precipitation records (Misut et al. 2003), it was apparent that years 1994 and 1995, when the model was calibrated, were below the long-term precipitation average. However, 1996 was an above average year with a total precipitation of 143 cm. To test whether the model would correctly respond to a substantial perturbation of recharge, a simulation was run with 1994 (86 cm of precipitation) as the initial recharge followed by 1996 (143 cm of precipitation), a higher recharge year. The 1996 results were then compared to the two well-head observations from March of 1997 in order to consistently use readings from the same time of year with a 1-year delay. Compared to the calibration, the validation simulation absolute error increased from 0.08 to 0.20 m for well 51172. For well 51179, the simulation absolute error increased from 0.05 to 0.21 m. Given that the seasonal well-head data in 1997 ranged by 0.28 m for well 51172 and 0.68 m for well 51179, the results were reasonable.

Climate change simulations

The IPCC sea-level rise predictions to the year 2099 range from 0.18 to 0.59 m for all future human activity scenarios (Meehl et al. 2007). The effective precipitation predictions over the same time period range from –2 to 15% for the median, most likely human activity scenario in eastern North America (Christensen et al. 2007). Actual precipitation is expected to be higher but is offset by an increase in evaporation caused by higher temperatures.

  1. Scenario 1.

    Because the model calibration used recharge from a below average precipitation year, a current baseline run was conducted using a long-term average annual precipitation of 112 cm (Miller and Frederick 1969) which yielded a recharge rate of 1.53 mm/day using a 50% recharge assumption. All other parameters were unchanged.

  2. Scenario 2.

    The second simulation was a scenario where the effects of climate change were mild with respect to groundwater resources; precipitation increased 15% and sea level rose only 0.18 m (0.6 ft). This represented the maximum predicted increase in precipitation coupled with the minimum predicted sea-level rise of the 2007 IPCC report. The constant head cells around the island were increased from 0.12 to 0.3 m and the recharge rate was increased from 1.53 to 1.76 mm/day. All other parameters were unchanged.

  3. Scenario 3.

    The third simulation was a scenario where the effects of climate change were severe with respect to groundwater resources; precipitation decreased 2% and sea level rose 0.61 m (2 ft). This represented the approximate maximum predicted decrease in effective precipitation coupled with the maximum predicted sea-level rise of the 2007 IPCC report. The constant head cells around the island were increased from 0.12 to 0.73 m and the recharge rate was decreased from 1.53 to 1.50 m/day. All other parameters were unchanged.

Results

The hydraulic head levels for the calibrated model are shown in Fig. 5, the salt-water/fresh-water interface in Fig. 6, the velocity vector field in Fig. 7, and the simulated flow path lines in Fig. 8. The model flow results (Figs. 7 and 8) are in general agreement with previous work by Paulsen et al. (2004) that found groundwater seepage fronts below sea level near the shoreline, often visible at low tide, were responsible for most of the discharge of the aquifer. At the area of maximum submarine groundwater discharge (SGD; Fig. 8), the model calculated a discharge of 0.015 m/day into West Neck Bay. The modeled SGD was at the low end of the discharge range of the Paulsen et al. (2004) findings of 0.02–2.3 m/day. This is to be expected given that the model was calibrated during a period of low recharge. However, it should also be noted that, because submarine groundwater discharge is highly dependent on recharge, comparing discharge in different years or seasons is difficult.

Fig. 5
figure 5

Calibrated model hydraulic head levels

Full size image
Fig. 6
figure 6

Calibrated model salt concentration from fresh water to seawater

Full size image
Fig. 7
figure 7

Relative velocity vector field for the calibrated model. The arrow size in each cell is proportional to the flow velocity and points in the direction of flow

Full size image
Fig. 8
figure 8

Flow path lines for the calibrated model. Area of maximum submarine groundwater discharge (SGD) in West Neck Bay which occurs 800 m from the west end of the model grid

Full size image

The results of the climate-change simulations for the West Neck Bay side of the model (Fig. 1) are shown in Fig. 9. In order to investigate the effects on potable water resources, the inner edge of the fresh-water/salt-water interface was defined as 0.25 kg/m3 which is 50% of the US National Secondary Drinking Water Standards’ suggested maximum total dissolved solids concentration of 500 mg/L (NSDWS 2002). Compared to the current long-term average fresh-water/salt-water interface position, scenario 2 created an interface that moved further seaward by an average of 23 m and a maximum of 60 m near the bottom of the interface. Conversely, scenario 3 created an interface that moved landward by an average of 16 m and a maximum of 37 m. The interface at the Coecles Inlet side of the model (Fig. 1) showed similar results. Movement of the interface was measured as a horizontal shift from the long-term average within the model grid and not as a displacement perpendicular to the interface. The outer edge of the fresh-water/salt-water interface (Fig. 9) was defined as 32 kg/m3. Figure 9 shows that the zone of diffusion shifted with changes in sea level and recharge but without appreciable overall variation in its average thickness. For scenario 2, the water table rose by an average of 0.27 m and maximum of 0.34 m. For scenario 3, the water table rose by an average of 0.59 m and maximum of 0.62 m.

Fig. 9
figure 9

Position of the fresh-water/salt-water interface in West Neck Bay. The seawater interface is defined as a salt concentration of 32.0 kg/m3 and the potable water interface is defined as a salt concentration 0.25 kg/m3. MSL mean sea level for 2007

Full size image

For the cross-section of Shelter Island used in this study, the original volume of fresh water, assuming a 1-m-thick vertical slice, was approximately 78,000 m3. For scenario 2, mild effects, the fresh-water volume of the cross-section increased by approximately 2,400 m3. This 3% increase in fresh-water volume was caused by a water-table rise which contributed 1,000 m3 and the outward movement of the fresh-water/salt-water interface which contributed 1,400 m3. For scenario 3, severe effects, the fresh-water volume of the cross-section still increased by 1,100 m3. This 1% increase in fresh-water volume was caused by the water-table rise which contributed 2,000 m3 and the inward movement of the fresh-water/salt-water interface which decreased the fresh-water volume by 900 m3.

Discussion

The predicted movement of the fresh-water/salt-water interface agreed with the conceptual model of Simmons (1986). The marine clay unit restricted the movement of the bottom of the fresh-water lens and only the sides moved landward or seaward in response to hydrological conditions. This restricted bottom appeared to be very important to Shelter Island’s response to climate change. Fetter (1972) created an analytical solution for an unrestricted fresh-water lens underlying an oceanic island (Fig. 3). The cross-sectional water-table profile for an infinite-strip island is described as:

$$ {h^2} = w{{\left[ {{a^2} - {{\left( {a - x} \right)}^2}} \right]} \mathord{\left/{\vphantom {{\left[ {{a^2} - {{\left( {a - x} \right)}^2}} \right]} {K\left( {1 + G} \right)}}} \right.} {K\left( {1 + G} \right)}} $$

where w is the rate of recharge, 2a is the width of the island, h is the water table head, x is the distance from the shoreline, K is the average hydraulic conductivity, and G is the dimensionless ratio of fresh-water and salt-water density: ρ f /(ρ s −ρ f ). According to the Fetter (1972) solution, head levels in an unrestricted island aquifer should change proportional to the square root of any change in recharge. Because the fresh-water lens volume should change approximately linearly to head levels, any changes in lens volume should also be proportional to the square root of any change in recharge.

If the Shelter Island aquifer behaved as an idealized unrestricted aquifer, sea-level rise would have negligible impact on lens volume as long as the water table did not exceed ground level and shoreline inundation was minimal. A rising sea level would cause the fresh groundwater lens to float higher while maintaining the lens shape dictated by the Ghyben-Herzberg relationship. Meanwhile, a 15% recharge increase would increase heads, relative to sea-level rise, and lens volume by the square root of the 15% increase or 7.2%. Likewise, a 2% recharge decrease would decrease heads, relative to sea-level rise, and lens volume by 1.0%. However, in the scenario with 0.18 m sea-level rise and 15% increase in recharge, the heads increased 8.5%, but the volume only increased by 3%. In the scenario with 0.61 m sea-level rise and 2% decrease in recharge, the heads decreased 1.9% relative to sea level, but the volume still increased by 1%. In each scenario, the head levels and water table responded as expected, but the change in lens volume did not. These discrepancies are best explained by the influence of the marine clay unit that restricts the bottom of the Shelter Island aquifer.

As sea level rises, the water table rises concurrently and the effective volume of the fresh-water lens increases because less of the standard lens shape of an island aquifer is obstructed by the marine clay unit (Fig. 3). This explains why the least favorable climate-change scenario, despite a decrease in recharge, still resulted in an increase in lens volume. It appears that sea-level rise is less detrimental to Shelter Island compared to typical sandy islands in the sense that it could actually increase fresh-water resources as long as shoreline inundation is minimal.

In the case of the most favorable scenario where recharge increased 15%, the lens volume did not increase by the expected 7.2% because the restrictive clay unit prevents the bottom of the lens from expanding along with the sides. The result was a mere 3% increase in volume. In general, any increases in recharge will have a smaller benefit than normally expected because of the marine clay layer’s limiting effect on the storage capacity of the aquifer.

The Masterson and Garabedian (2007) study found a slight to moderate decrease in fresh-water lens thickness due primarily to stream loss, but Shelter Island has no notable streams that intersect the water table. An analytical study, by Werner and Simmons (2009), also found that constant head aquifers, approximated by aquifers with substantial intersecting streams, experience more severe salt-water intrusion than aquifers with a constant flux. Thus, the results of this study are more in line with the Theiler and Hammar-Klose (2000) assessment of coastal aquifers that are relatively immune to salt-water intrusion caused by sea-level rise.

The estimates of changes in fresh-water lens volume have additional uncertainty because the model did not consider surface topology and the landward movement of the shoreline that would occur as rising sea level inundates the coast. As previously noted, low lying areas that are likely to be inundated or intersect with the rising water table will experience more fresh-water loss. For the cross-section modeled in this study (Fig. 2), assuming a stationary shoreline simplified modeling while introducing an acceptably small error. According to the USGS Topographic Quadrangle Map for Greenport, NY, the 2-m elevation contour is within 10 m of the shore for the study cross section and within 50 m of the shore for almost the entire island. Because most of Shelter Island is at least 5 m above NGVD, the results of the study can be generally applied to the rest of Shelter Island. Thus, it is expected that the fresh-water lens under Shelter Island will remain largely intact over the next century.

The IPCC Fourth Assessment Report sea-level-rise predictions may be conservative. Horton et al. (2008) have used a semi-empirical method to estimate a new range of likely sea-level-rise estimates for 2100 that range from 0.47 to 1.00 m. Meanwhile, Pfeffer et al. (2008) have estimated a reasonable worst-case sea-level rise that includes rapid deglaciation mechanisms not accounted for by the IPCC report. The estimate ranges from a 0.8–2.0 m mean sea-level rise by 2100. However, even with a sea-level rise more than three times higher than the IPCC estimates, given the elevation of Shelter Island, only a small portion of the island should experience inundation and salt-water intrusion.

Conclusion

In order to quantify the effects of climate change on a small sandy island, a variable-density transient-groundwater-flow model was created for Shelter Island, NY, USA. The 2007 IPCC report predictions for changes in precipitation and sea-level rise over the next century were used to create two future climate scenarios. A scenario most favorable to groundwater retention consisting of a predicted precipitation increase of 15% and a sea-level rise of 0.18 m was compared to the current long-term average. This resulted in a seaward movement of the fresh-water/salt-water interface by an average of 23 m and a maximum of 60 m. The water table rose by an average of 0.27 m. A scenario least favorable to groundwater retention, consisting of a predicted precipitation decrease of 2% and a sea-level rise of 0.61 m, was compared to the current long-term average. This resulted in a landward movement of the interface by an average of 16 m and a maximum of 37 m. The water table rose by an average of 0.59 m. The estimated change in fresh-water resources ranged from an increase of 1 to 3%. The discrepancy in expected changes in fresh-water lens volume was best explained by the restrictive marine clay unit which deforms the aquifer underlying Shelter Island. Although it would be reasonable to assume that the severe conditions would decrease the available fresh water because the fresh-water/salt-water interface moves landward, the loss caused by this movement is more than compensated for by the increased lens thickness resulting from a rising water table. While a relatively flat low-lying island with a water table already close to the topographical surface would lose fresh-water resources due to climate change, an island with high bluffs, like Shelter Island, can survive climate change with its fresh-water aquifer relatively intact. This relatively small change in fresh-water resources over the next century suggests that the primary challenge of climate change on Shelter Island will not be potable-water retention, but rather dealing with other factors such as increased storm surge damage.