Charleston is located on the South Carolina coast and sits on a peninsula with the adjacent Ashley River to the west and Cooper River to the east (Figure 1). Charleston Harbor is south of the peninsula and acts as an estuarine buffer between the rivers and the Atlantic Ocean (Purcell et al., 2020). The Charleston peninsula had undergone extensive urban development, which involved filling or draining marshes and other natural areas (Butler, 2020), increasing the volume of stormwater runoff draining into the coastal waterways (Blair et al., 2014a).

Floodwater was sampled within the Charleston peninsula at four sites: (i) Hagood Ave. at Fishburne St. (HF) (32°47′30.0′′N 79°57′22.9′′ W), (ii) Coming St. at Calhoun St. (CC) (32°47′6.0′′N 79°56′22.6′′ W), (iii) East Bay St. at HWY 17 (EB17) (32°48′01.9′′N 79°56′10.9′′ W), and (iv) Magnolia Cemetery (MC) (32°49′3.3′′N, 79°56′37.0′′ W). The MC site is a tidal creek next to a historic cemetery and was designated as the control site as this site did not receive stormwater runoff and was only impacted by tidal flow from the Cooper River (Figure 1). These sites were chosen based on the updated flood hazard data from the US Federal Emergency Management Agency (FEMA) Flood Hazard and Risk Data Viewer (FEMA, 2020). The viewer provided a flood zone map of where potential areas in Charleston would be inundated by a flood event (FEMA, 2020). The MC, HF, and EB-17 sites experience sunny-day tidal flooding where the flood tide height is at least 2.1 m MLLW (Coz et al., 2021). Our observations during sampling confirmed commonly flooded sites during stormwater and tidal flooding.

Floodwater sampling was conducted from December 2021 to December 2022 as outlined: (i) the MC site from March 2022 to September 2022, (ii) the HF site from December 2021 to July 2022, (iii) the EB17 site during June 2022, July 2022, and December 2022, and (iv) the CC site from June 2022 and December 2022. At all sites, floodwater samples were collected near stormwater drains or culverts. The CC site was only sampled during rainfall events due to the site being farther inland and not impacted by tidal flooding. The HF and EB17 sites were sampled during both rainfall and high tide flooding events due to the proximity of the Ashley River and Cooper River, respectively.

When at least 1 cm depth of water was flowing into the drain, a floodwater sample was collected with either water flowing into the mouth of a sampling container or completely submerging the container while tightening the cap. Water samples for trace metal and coliform analyses were collected using 50 mL sterile centrifuge tubes and stored in a cooler for transport. Bright vests were worn during sample collection for visibility and personal protection and sample contamination was minimized. Measures were taken to prevent bed sediment collection and minimize turbid conditions by gently sampling water from the stormwater drain. This method is based on the United States Environmental Protection Agency (USEPA) standard method for industrial stormwater sampling (USEPA, 2021).

Tidal height and precipitation data were obtained from the NOAA online portals (Center for Operational Oceanographic Products and Services and NOAA, 2023; National Weather Service and NOAA, 2024). Tide data from the tide gauge station closest to the sampling site was used to record the most representative tide height.

Watershed delineations were created to identify the boundaries surrounding the floodwater sites where stormwater runoff would flow toward lower elevations. ArcGIS Pro (version 3.1, ESRI, Redlands, CA) and the built-in Hydrology Tools were used to perform watershed delineations. A 1-m resolution digital elevation model (DEM) created using the available Light Detection and Ranging (LiDAR) data was used to determine the elevation contours of the peninsula. The errors in the DEM were corrected by using the Fill tool in the Hydrology Toolset to allow for a better representation of the flow direction (Arifjanov et al., 2022). Streamlines were generated after the filling of the sinks to determine where the stormwater runoff would flow and if the water would flow toward the floodwater sites by using the Flow Direction tool. Finally, a watershed boundary for all the streamlines was created for each sampling site by converting the raster to a polygon (Villines et al., 2015; Goji Tumba and Amusuk, 2017). The polygons at each site were all spatially joined to make all the watersheds into one layer. The impervious surface layer was obtained from the United States Geological Survey (USGS) National Land Cover Database (NLCD) (USGS, 2021) which provides impervious surface percentages for the Charleston peninsula in 2019. This layer was joined with the watershed boundaries layer.

The Manta+20 Eureka™ Water Probe (Eureka™ Water Probes, Austin, TX) was used to collect water quality parameters, including water temperature (°C), pH, conductivity (μS/cm), and dissolved oxygen (mg/L). After all the readings for each parameter remained stable, measurements were recorded, and the probe was rinsed with deionized water (18 MΩ.cm) before and after measurements to prevent cross-contamination. Conductivity measurements obtained from the Water Quality Data Portal for Ashley River and Cooper River were averaged from 1981 to 2021 (National Water Quality Monitoring Council, 2023). The mean conductivity values for the Ashley River and Cooper River are 31,745 μS/cm and 30,867 μS/cm, respectively. The conductivities from the sampling sites that had tidal influences were compared to the corresponding river conductivity value to identify if the floodwater was brackish.

Bacterial analysis began within 4 h of collection to enumerate coliforms and ampicillin-resistant coliforms. All bacteria culture plates were filled with 1 mL sample volume. This was either achieved with sterile water, ampicillin solution, or a mix of site water with sterile water or ampicillin solution depending on the site. For the MC site, the Compact Dry EC plates (Hardy Diagnostics, Santa Maria, CA) were rehydrated with either ampicillin (32 mg/L) or sterile water for the membrane filtration using a sterile 0.45 μm filter of floodwaters since this site had lower bacterial concentrations (APHA et al., 2018). The Compact Dry EC plates contain chromogenic enzyme substrates, so E. coli will form blue colonies whereas other coliforms will turn pink-purple (Kodaka et al., 2006). For the other three sampling sites (HF, EB17, and CC), floodwater samples were diluted with either sterilized water or a solution of ampicillin directly applied to the plate. Plates were incubated at 35°C ± 2°C, and after 18–24 h, the colony-forming units (CFUs) were enumerated. One CFU equals one bacterium, therefore, a high number of bacteria in the floodwaters would result in high CFUs. E. coli and coliforms were enumerated to yield total coliform (TC) count and reported in CFU/100 mL. If the CFUs counted on the plates prepared with a sample were >150 CFUs, that sample was coded as Too Numerous To Count or TNTC (SC DHEC, 2019). If a 150-CFU threshold was reached for a sample, a value of 150 CFU/100 mL was reported as a conservative estimate of TC coliform concentration.

Total concentrations of dissolved trace metals (Al, Sb, As, Ba, Be, Cd, Cr, Co, Cu, Pb, Mn, Mo, Ni, Se, Ag, Tl, Th, U, V, and Zn) in the water samples were determined using an inductively coupled plasma mass spectrometer (ICP-MS) (iCAP RQ ICP-MS, Thermo Fisher Scientific, Waltham, MA) based on the revised USEPA Method 200.2 Revision 2.8 (Martin et al., 1994). The water samples were filtered using 0.45 μm polyethersulfone (PES) syringe filters and acidified to 2% v/v using nitric acid. A multi-element standard mix (High Purity Standards, North Charleston, SC) was diluted to calibrate the inductively coupled plasma mass spectrometer (ICP-MS) consisting of serial dilutions from 10⁻⁴ to 10 mg/L. Check standards and blanks were incorporated for each set of water samples.

The raw data were organized, and outliers were removed for the average colony concentrations for E. coli, coliforms, and TC by using the interquartile range and calculating the inner and outer fences to identify the boundaries of the data (Schwertman and de Silva, 2007). After removing the outliers (63% of total samples removed), principal component analysis (PCA) was performed on the data using R software. The normalized data from all the experimental variables were transformed into principal components (PCs), which are linear projections that are arranged by the size of the eigenvalue. This analysis reduced the dimensionality and identified any relationships between multiple variables (Demšar et al., 2013). Additional R packages, including factoextra, were used to generate biplots (Kassambara and Mundt, 2020), ggplot2 was used to make plots aesthetically appealing (Wickham, 2016), and psych was used to calculate the cumulative variance (Revelle, 2023).

The four sampling sites were located on different parts of an urban peninsula, which had varying percentages of impervious surfaces. The MC site had the least impervious coverage on the peninsula, ranging between 1% and 19%. The CC site had the highest impervious coverage, ranging from 80% to 100%. The HF and EB17 sites had similar impervious coverages, between 50% and 79%.

The HF site had the smallest watershed boundary, which was only 3.23 hectares. The EB17 site, on the other hand, had the largest watershed boundary of 55.04 hectares. Both these sites were affected by compound flooding. The MC site had the second-largest watershed boundary of 19.02 hectares and was mainly affected by tidal flooding. Finally, the CC site watershed boundary was 10.52 hectares and experienced only rainfall flooding (Figure 2).

Chemical characterization was performed independently for each site due to their unique geography and flooding influences. The medians, first and third quartiles, were used because these water quality data are not normally distributed. The conductivity data indicated that the MC site had brackish water in the floodwater with a median conductivity of 27,130 μS/cm. The median tide height at this site was 1.32 m MLLW, and no rainfall was recorded at this site during the sampling events. The median TC and AmpRC colony concentrations were 4,259 CFU/100 mL and 764 CFU/100 mL, respectively, or AmpRC counts were 17.93% of TC counts for this site. Of trace metals, only As was detected at this site with a median value of 4 μg/L (Table 1).

The CC site was only influenced by stormwater runoff, and the stormwater had a median conductivity of 100 μS/cm and a median rainfall of 1.82 cm during the sampling events at this site. The median TC and AmpRC colony concentrations were 33,850 CFU/100 mL and 259 CFU/100 mL, respectively, or the AmpRC counts were 0.76% of TC counts at this site. Both As and Cr were detected in the floodwater at this site with median concentrations of 3 μg/L and 5 μg/L, respectively (Table 1).

The median conductivity of 6,340 μS/cm confirmed that HF site was impacted by compound flooding (typical conductivity maximum values for freshwater and saltwater are 700 μS/cm and 25,000 μS/cm, respectively). The median rainfall amount and tide height at this site during sampling events were 1.75 cm and 1.32 m MLLW, respectively. The median TC and AmpRC colony concentrations were 12,300 CFU/100 mL and 6,334 CFU/100 mL, respectively or the AmpRC counts were 51.49% of TC counts at this site. Only As was detected at this site with a median concentration of 3 μg/L (Table 1).

The EB17 site was also impacted by compound flooding like the HF site, and the floodwater had a median conductivity of 976 μS/cm, which is above the threshold of freshwater conductivity of 700 μS/cm. The median rainfall amount and tide height at this site during sampling events were 0.97 cm and 1.32 m MLLW, respectively. The median TC and AmpRC colony concentrations were 49,500 CFU/100 mL and 9,334 CFU/100 mL, respectively or the AmpRC counts were 18.86% of TC counts at this site. Trace metals, As and Cr, were detected in the floodwater with median concentrations of 4 μg/L and 8 μg/L, respectively (Table 1).

Principal component analysis (PCA) was used to correlate the physical, chemical, and microbial contaminant data from all the sites. Eight principal components (PCs) comprising eight variables (contaminant data), including TC, AmpRC, tide height, rainfall, conductivity, As, Cr, and Zn, were calculated. The proportion of variances for each PC were 33%, 21%, 14%, 11%, 9%, 7%, 3%, and 2%, respectively. The first two PCs (PC1 and PC2) were used for the biplots as they accounted for the two highest variances (33.1% and 20.6%, respectively). Rainfall, TC, and AmpRC were strongly and positively correlated with each other. The tide height and conductivity are strongly correlated to each other and negatively correlated with rainfall and TC, and less strongly correlated with AmpRC counts. The trace metals all were strongly correlated with each other and less correlated with TC and AmpRC concentrations (Figure 3).

The coliforms (TC) and AmpRC observed in the floodwaters in this study can originate from many sources, as observed in similar coastal areas (Selvakumar and Borst, 2006). The detection of AmpRC in the nearby tidal creeks, Shem Creek and Kushiwah Creek (DeLorenzo et al., 2012; Emery et al., 2022), also confirms that one of the main sources of AmpRC in coastal waterways is the floodwater generated within the urban watersheds. Other cities in the United States have found coliforms and antibiotic-resistant bacteria (ARB) in different urban waterbodies from stormwater discharge (Singh et al., 2019) and antibiotic-resistant microbes, an emerging public health threat (Scott et al., 2016). These bacteria are discharged directly into the waterbodies via stormwater drains, negatively affecting the marine ecosystem and human health (Parker et al., 2010). In addition, there is the potential for antibiotic-resistant genes, such as ampicillin-resistant genes, to be shared among different microbes (Uyaguari et al., 2011, 2013; Horton, 2022) such as Vibrio spp. which are highly prevalent and known human health threats to seafood safety and contact recreation. The mechanism for resistance to ampicillin is typically due to β-lactamase activity. Genes for this enzyme are typically encoded on mobile genetic elements allowing for ease of horizontal gene transfer to other environmental bacteria, which can impact human health (Uyaguari et al., 2011, 2013). Increased surveillance of AmpRC as well as these genetic elements would allow for targeted interventions to prevent environmentally acquired antibiotic resistance infection (Djordjevic et al., 2023).

Recent studies of urban floodwaters in the aftermath of tropical cyclones have shown an increased presence of contaminants, including coliforms and antibiotic genes, which exacerbates potential AmpRC presence in floodwaters (Presley et al., 2005; Yang et al., 2021). The possible sources for the transfer of coliforms during storms and sunny-day flooding include potential septage leaks from aging sewer lines, dog waste, and wildlife (Whitlock et al., 2002; DeLorenzo et al., 2012; Almakki et al., 2019; Powers et al., 2020). The AmpRC analytical method used in this study may be used to identify pollution sources from pets, as ampicillin is often prescribed for skin disorders in dogs and cats. Elevated levels of ampicillin-resistant E. coli were found at sites with high concentrations of pets, such as urban dog parks (Kelsey et al., 2003). In this study, the HF site had nearly 51.5% of the TC in the form of AmpRC bacteria (or, AmpRC/TC ratio was 51.5%) and is near a city park frequented by pets. The AmpRC/TC ratio was much lower at the other sites in this study, ranging from 7.6%−18.9%.

The rise in coastal sea level is exacerbating the flooding for many communities along the coasts across the world, including the United States. The global average sea level in 2022 was 101.2 mm higher than the 1993 level, with high tide flooding across the United States being over 300–900% more frequent than it was 50 years ago (Lindsey, 2022). The increased impervious surfaces in coastal cities, coastal erosion, groundwater pumping, and subsidence exacerbate flooding and, thus, increase pollution of floodwater and coastal waterways (Chithra et al., 2015; Lindsey, 2022; Ohenhen et al., 2023). Climate change has been increasing the number of storms, including short-duration storms and major storms like hurricanes (Lau et al., 2022), which can also coincide with high tides as more intense compound flooding (Bevacqua et al., 2020).

The conductivity was used as an indicator to identify if the floodwaters originated from freshwater sources (rainfall) or brackish (tidal) sources. For freshwater, the conductivity that was <700 μS/cm was typically associated with rainfall. Brackish water typically has a range between 700–25,000 μS/cm, which would come from tidal flooding and not directly from the ocean (saltwater) (Rusydi, 2018).

The USGS's land coverage (NLCD raster) data from 2019 was used to represent the changing landscape of Charleston (USGS, 2021). Combining these land cover data with high-resolution (1-m) ground elevation (LiDAR), data allowed us to closely examine the impact of impervious surfaces on the contamination of floodwater. The percentage of impervious surfaces showed urbanization surrounding the study sites correlated inversely with the coliform data. On average, the MC site had fewer coliforms than the other three study sites which had more impervious surfaces. This high-resolution LiDAR watershed delineation identified where stormwater can accumulate and flow toward each of our sampling sites but did not directly address tidal flooding. Tidal conditions impacted how quickly the storm drains were emptied or if they backed up—these aspects were not addressed in our watershed delineation analysis and would need to be addressed in the future.

The region continues to experience population growth, according to the 2020 Decennial Census (US Census Bureau, 2023). This urbanization within the greater Charleston region is expected to increase between 2% and 30% by 2030, with urban sprawl continuing to impact the region (Dickes et al., 2016). Rising sea levels in the coastal areas, increasingly intense precipitation, and the associated rise of the water table increase concerns about the loss of floodwater's ability to seep into the ground (Sweet and Park, 2014; Neumann et al., 2015; Kulp and Strauss, 2019). Loss of permeable ground surfaces increases local flooding and the transport of floodwater-associated contaminants into coastal waterbodies. The HF and MC sites have higher vegetation buffers and less impervious surfaces than the other two sites, which allowed increased seeping of floodwater into the ground instead of localized flooding to some extent. A more systematic study would be required to confirm this observation. The HF and MC sites also had lower TC and AmpRC colony concentrations than the CC and EB17 sites. Studies reported increased fecal coliform and trace metal concentrations in stormwater after an increase in impervious surfaces in urban areas (Bhandari et al., 2017; Zhang et al., 2021). These studies confirm a positive correlation between the increased presence of E. coli in stormwater and increasing urbanization.

Trace metals are commonly associated with urban runoff under a variety of conditions—during normal or extreme precipitation events (Joshi and Balasubramanian, 2010; Barber et al., 2017; Bhandari et al., 2017; Kirker and Vulava, 2022). There are no guidelines or regulations that inform safe levels of trace metals in urban floodwater. Still, there are the USEPA's published guidelines, such as the “National Recommended Water Quality Criteria-Aquatic Life Criteria Table” for acceptable concentrations of water contaminants (USEPA, 2023). Typically, floodwater and urban runoff can be more concentrated compared to the final receiving waterbodies, where contaminants are diluted (Joshi and Balasubramanian, 2010). Three trace metals (As, Cr, and Zn) were detected in the floodwater during this study and none of these trace metals approached levels that would be deemed concerning from aquatic life criteria (USEPA, 2023). We note that all floodwater samples in our study were filtered and analyzed—however, there was a significant suspended sediment load in most floodwater samples. Analysis of composited water samples would most likely have shown a significantly higher concentration of trace metals in the runoff (Jeong et al., 2020). Charleston region has abundant marine-derived phosphate rock deposits (Malde, 1959) and a history of phosphate mining, resulting in elevated arsenic in regional waterbodies of this area and the floodwater in our study. Arsenic pollution in urban watersheds and the estuaries of the southeastern US and many parts of the world is primarily from the leaching of As from phosphate rocks and industrial activities (Chirenje et al., 2003; Cotti-Rausch et al., 2018). Similarly, Cr and Zn in urban floodwaters can be associated with a combination of rock weathering, industrial activities, automobile wear and tear, etc. (Joshi and Balasubramanian, 2010; Cotti-Rausch et al., 2018).

Identifying relationships between different water quality data has given insight into the interaction between the variables. Trace metals as stressors on ARB or the possibility of co-resistance between the two have been frequently studied (Hamilton et al., 2020; Biswas et al., 2021; Horton, 2022), however, further studies are needed in the Charleston region. The relationships between rainfall amounts and tide height with coliforms (TC and AmpRC) are well-known, especially in the urban environment (Liu and Huang, 2012; Kirker and Vulava, 2022). High trace metal concentrations may be toxic to Vibrio and Enterococcus, while low chronic doses stimulate growth and possibly induce antibiotic resistance (Horton, 2022).

The bacteria concentrations at all study sites affirmed the presence of AmpRC in the urban floodwaters of Charleston. Additional studies would be required to identify these bacteria source(s). Coliforms resistant to other common antibiotics may be present in these floodwaters—an analytical method designed to identify the resistance to multiple antibiotics would provide more insights. Multidrug-resistant coliforms have been isolated from a surrounding waterway (Emery et al., 2022). Future assessments of urban floodwaters should assess the prevalence of coliforms resistant to multiple commonly human and animal-prescribed antibiotics.

The trace metals measured in this study did not positively correlate with TC or AmpRC concentrations, but other trace metals that are likely to be associated with suspended sediment may correlate with both types of bacteria. An extended sampling period and additional sampling sites would allow more comprehensive data. Comparisons between trace metal concentrations in tidal creeks and stormwater ponds within the tidal creek watersheds in coastal South Carolina confirmed the presence of Cd, Cu, Cr, and Zn (Sanger et al., 1999; Allen et al., 2019; Horton, 2022).

The PCA method allowed for the simplification of complex variables while still retaining trends (Lever et al., 2017), but using other statistical methods would provide more insight into the relationships with other variables from this study.

While flooding on the Charleston peninsula continues in a changing climate, the utilization of best management practices (BMPs) for the stormwater infrastructure would mitigate contamination of water bodies. For example, bioretention landscapes or rain gardens that include vegetation, filtration media, and drains can retain and filter stormwater (Li et al., 2019). Many people in Charleston are frequently exposed to these contaminated floodwaters. Installing BMPs may alleviate the exposure to contamination and help lower the microbial load in the floodwaters (Ahmed et al., 2019).

To decrease human exposure to these antibiotic-resistant potential pathogens, members of communities with high levels of flooding should be informed. Active public educational outreach about contaminants present in the floodwaters will contribute significantly to positive health outcomes (Meinhardt, 2006; MacDonald and Tippett, 2020). For example, the development of a short one-page fact sheet warning the public of the bacterial and other pollution hazards posed by floodwater would alert the public of exposure. The USEPA uses such infographics to warn the public about pollutants, such as Pb found in drinking water (USEPA, 2017).