1. Introduction
Tropical cyclones (TCs) threaten coastal communities with flooding, including both freshwater flash floods caused by heavy rain and seawater inundation caused by storm surges. Stormwater management ponds (SMP), including both (wet) retention ponds which hold water year-round and (dry) detention ponds which only hold water for a short time before the water enters streams, are a critical type of infrastructure that protects coastal communities from flood hazards caused by heavy rainfalls and runoff. A study in southwestern Taiwan showed that detention ponds reduce the inundated potential, inundated volume, flood damage, and flood stage of peak flow [
1]. Sahoo and Pekkat [
2] found that detention ponds reduce the effect of urban flood in an urbanizing catchment, reducing flood depth by 46.5% and the inundated area by 43%. Similarly, Chrétien et al. [
3] analyzed retention ponds in a small agricultural catchment of Canada and stated that retention ponds could reduce peak flows by 38%.
Retention ponds fail to function when the rainfall exceeds the pond’s capacity, e.g., with the underprediction of rainfall intensity or unrealistic predictions of post-development flows [
4]. Similarly, during TC-induced coastal compound flooding, the function of retention ponds will fail to retain rainfall runoff water if they are submerged by seawater inundation.
Under climate change scenarios, the projected sea level rise (SLR) and more extreme storms in the future will pose aggravated threats to the proper functioning of SMPs.
Since the last glacial period ended about 15,000 years ago [
5,
6], global temperature rises have caused the thermal expansion of seawater and glacial ice melting, resulting in sea level rise and increased land surface runoff, respectively [
5,
7]. Although studies on exact sea level trends are subject to uncertainty [
8,
9], most coastal areas around the world have experienced and continue to experience sea level rise periods. Depending on the volume of greenhouse gases to be emitted in the coming years, the Representative Concentration Pathway (RCP), which is a greenhouse gas concentration trajectory, is used by IPCC to evaluate climate change scenarios [
10]. According to three different projections in the latest IPCC report, the global average sea level rise in 2100 may be 0.43 m (likely range: 0.29–0.59 m), 0.55 m (likely range: 0.39–0.72 m), or as much as 0.84 m (likely range: 0.61–1.10 m) under RCP2.6, RCP4.5, and RCP8.5, respectively. Griffiths et al. [
11] projected that ninety percent of the coastal areas will experience a sea level rise of more than 0.2 m if global warming exceeds two degrees Celsius above pre-industrial levels. Parris et al. [
12] predicted sea level rise of 0.2 m (lowest), 0.5 m (intermediate-low), 1.2 m (intermediate-high), and 2.0 m (highest) based on different degrees of ocean warming and ice sheet loss.
In addition to SLR, researchers have also reached a consensus on future changes in TC intensity. The heat content stored in the upper ocean provides the energy that fuels the development of TCs. As a result, increased sea surface temperature creates a more favorable environment for TC formation and growth. Emanuel [
13] created an index of the potential destructiveness of TCs, based on total power dissipation integrated over a TC’s entire lifetime. Emanuel used this index to study hurricanes of the past 30 years and found that the index has increased significantly since the mid-1970s [
13]. Webster et al. [
14] examined the intensity of tropical cyclones for more than 35 years in the context of increasing sea surface temperature and found that the number of category 4 and 5 hurricanes increased at different rates in the North Pacific, India, Southwest Pacific, and North Atlantic oceans. According to Mousavi et al. flood elevations caused by catastrophic hurricane storm surges in the Gulf of Mexico will rise by 0.5 m and 1.8 m by the 2030s and 2080s, respectively [
15]. Even though uncertainty exists in hurricane pattern change studies, it is certain that storm surge caused by extreme tropical cyclones will be a significant trigger for coastal flooding in the future.
This study will quantify the risk of SMPs’ loss of function after being submerged under storm surge and seawater inundation during tropical cyclones, and identify how SLR and more extreme storms in future climate change scenarios can exacerbate the situation.
2. Materials and Methods
2.1. Historical TC-Induced Storm Surge Data
From 1851 to 2018, 243 TCs in 132 years made landfall or caused storm surges on the South Carolina coast. To determine the maximum storm surge heights, all of these storms were simulated using a hydrodynamic model.
Figure 1 depicts the locations where each year’s simulated maximum storm surge occurred.
We used the above modeled historical storm surge height data to estimate the storm surge heights for the South Carolina coast with 5-year, 10-year, 50-year, 100-year, and 1000-year return periods. The storm-surge-height return periods were calculated based on the classical Hazen method [
16]
where
is the return period of a specific storm surge height,
n is the total number of events (132 in this study), and
m is the rank of a specific storm surge height.
m = 1 when the maximum storm surge height is the 132-year high.
The resultant storm surge height return period is shown in
Figure 2. The storm surge heights as a function of return periods fit into h(t) = −0.0394 + 1.38ln(t) − 0.107ln(t)
2, where h is the storm surge height (in m) and t is the return period (in year).
The storm surge heights for the return periods of 5, 10, 50, 100, and 1000 years are 1.9 m, 2.6 m, 3.7 m, 4.0 m, and 4.4 m, respectively. The 1000-year storm surge height has never occurred before and its value is extrapolated from past storm surge records. Note that these storm surge modeling results were based solely on changes in water level caused by TCs. Astronomical tide is another important factor affecting the actual storm surge height. For example, the once-in-100-years (return period = 100 years) storm surge height for the South Carolina coast is 4.0 m, which may coincide with a high tide of 1.5 m—the average high tide for the South Carolina coast—and become 5.5 m, or coincide with a low tide of −1.5 m and become 2.5 m, or coincide with a tide somewhere between the two extremes. Due to the uncertain possibilities of the tidal phase coinciding with storm surge height, the astronomical tide-caused water level changes are not included in the calculation of the storm-surge-height return periods. Instead, the impact of tides is considered as a worst-case scenario in the subsequent inundation modeling and the assessment of the risks of SMPs, in which case it is assumed that the storm surge height with a certain return period coincides with a high tide.
2.2. Inundation
The method for calculating inundation extent as a result of storm surge during a tropical cyclone is illustrated in
Figure 3. While storm surge heights caused by past tropical cyclones were simulated using the coupled hydrodynamic model, inundation areas were calculated based on the geographic information system (GIS) technique using the QGIS platform. It is well established that storm surge height intrudes on land with decay [
17]. The surge decay coefficient (
SDC) is defined as follows:
where
is the storm surge height,
is inundation elevation (the average elevation where the effect of the surge can reach),
is inundation width (the longest distance of inundated area away from coast), and
is the width of the constant surge from the coast.
We assume that the inundation begins at the coastline so that = 0 m. We first run an idealized 2D hydrodynamic model simulation by setting sea level to the storm surge heights (the in Equation (2)) of 1.9 m, 2.6 m, 3.7 m, 4.0 m, and 4.4 m to determine the widths of the inundated area and the end inundation elevation . The widths of the inundated area and the end inundation elevation were obtained manually based on flooding extent maps produced by the hydrodynamic model Delft3D-FLOW.
Delft3D-FLOW is a 2D or 3D simulation program that calculates water’s non-steady flow and transport caused by tidal and meteorological forcing on a rectilinear or a curvilinear fitted grid. The spatial resolution varies from 127 to 1268 m in the x direction and 55 to 557 m in the y direction. The time step is 0.4 s. Such grid spacing and time step allow stable integration in shallow water, e.g., when the water depth is 100 m, the Courant number is dt/(dx/sqrt(gH)) = 0.2, which is much smaller than required for computational stability under CFL conditions. The Holland model [
18] is used to create the wind data based on hurricane track and intensity to drive Delft3D-FLOW.
The SDC values can then be determined based on the above equation. Additionally, later, these SDCs were applied to the inundation area calculation under different SLR scenarios using the geographic information system technique on the QGIS platform, which is more computationally efficient than numerically modeling each individual scenario. The QGIS Geospatial Data Abstraction Library (GDAL) was used to process the inundation maps. The steps for determining whether a giver inland location is inundated are as follows:
Calculate the raster distance (RD) from the coastline to the inland location using the Proximity function. Note RD = 0 m at the coastline.
Determine the water depth (WD) based on , SDC, RD, and DEM elevation (DEM) using the formula: .
WD > 0 indicates inundation. The Raster calculator function in QGIS is used in this step.
The storm surge height, inundation width, the inundation elevation, and surge decay coefficient for storm surges with different return periods are listed in
Table 1.
2.3. Experiments—Impact of SLR to Stormwater Ponds
For each return period of storm surge height, the inundation is calculated under the following scenarios: (1) no tide and no SLR; (2) with tide and no SLR; (3) with tide and SLR = 0.5 m; (4) with tide and SLR = 1.0 m; (5) with tide and SLR = 1.5 m; and (6) with tide and SLR = 2.0 m. These SLR scenarios include the low-, intermediate–low-, intermediate–high-, and high-risk scenarios proposed by Parris [
12]. A total of 36 calculations (6 TC storm surge return periods each under 6 different SLR scenarios) were carried out. In each calculation, the number of submerged SMPs was counted. The impact of SLR scenarios on the number of SMPs affected during non-storm (return period of zero years) and storm conditions (return periods of 5, 10, 50, 100, and 1000 years) were assessed. The high-risk “hotspots” were identified where mitigation and adaptation efforts are required under future climate SLR scenarios.
Furthermore, the South Carolina coast was divided into three domains based on topography and the three SMP clusters centered in these three regions: Long Bay, Charleston, and Beaufort. In addition to the assessment of the entire South Carolina coast as a single domain, the impact of SLR on SMPs was also evaluated in detail for each of these three domains individually. Note that some SMPs are outside of these three regions.
2.4. Storm Water Pond and Topography Data
The South Carolina storm management pond survey [
19] based on aerial images, which identified approximately 21,594 SMPs, was used in this study for the SMP locations (
Figure 4). The aerial images used to identify SMPs are provided by the National Agriculture Imagery Program (NAIP). The spatial resolution of these aerial images is about 1 m. The SMPs are concentrated in three high-density clusters in the Long Bay, Charleston, and Beaufort areas. These topography data are an important input for producing the inundation area. The topography data used in this study were obtained from the United States Geological Survey 3D Elevation Program (3DEP), with a spatial resolution of 30 m.
5. Conclusions
Tropical cyclone-induced storm surges cause inundation that could threaten the proper function of stormwater management ponds in the coastal South Carolina area, and the risks could be aggravated by future sea level rise.
A framework is developed to assess the impact of TC-induced storm surges and inundation on stormwater management ponds. A hydrodynamic model simulates storm surges caused by historical hurricanes that struck the coast of South Carolina between 1851 and 2018. Storm surge heights with various return periods are established. Inundation caused by tides and storm surges could threaten storm management ponds in coastal regions. The rise in sea level could exacerbate these risks. Under various SLR scenarios, the inundations caused by storm surge heights with return periods of 0 (non-storm high tides), 5, 10, 50, 100, and 1000 years are simulated, and the SMPs at risk of being affected by these events are counted. For the entire South Carolina region, the number of SMPs at risk of being inundated by tides and storm surges would increase almost linearly with SLR by around 10 SMPs for every inch of SLR, for TC storm surges with all return periods.
Three coastal regions—Long Bay, Charleston, and Beaufort—are analyzed in detail to identify the communities whose SMPs are at high risk in the face of future SLR scenarios. Hilton Head Islands, Charleston Harbor, Kiawah Island, George Town City, and Pawleys Island are identified as these high-risk hotspot areas. If submerged by storm surge and seawater inundation, current SMPs installed in South Carolina could be incapable of functioning to mitigate the increased surface runoff and flash flooding caused by extreme rainfalls. Therefore, the findings of this study provide necessary information regarding the locations where current SMPs must be re-designed and where more SMPs must be developed. The modeling and analysis system developed for this study can be applied to assess the impact of SLR and other climate change scenarios on SMP facilities in other regions.