The global archaeological record along coastlines is in peril from sea-level rise. Archaeological modeling of this loss is important, and although general models have been developed to assess the impact of rising sea levels, few offer sophisticated modeling of when and the mechanism of how those sites are likely to become damaged or destroyed at a site-specific scale (Anderson et al. Reference Anderson, Bissett, Yerka, Wells, Kansa, Kansa, Myers, Carl DeMuth and White2017; Rick and Sandweiss Reference Rick and Sandweiss2020; Rockman and Hritz Reference Rockman and Hritz2020; Rockman et al. Reference Rockman, Morgan, Ziaja, Hambrecht and Meadow2016). There are a number of reasons why the coastal archaeological record is important, including allowing for an understanding of ecological baselines, documenting the loss of sacred sites to Indigenous peoples and descendant communities, and demonstrating the role that such information can play regarding the nature of ecosystem restoration, among others (Erlandson and Rick Reference Erlandson and Rick2010; Jackson et al. Reference Jackson, Kirby, Berger, Bjorndal, Botsford, Bourque and Bradbury2001; Van de Noort Reference Van de Noort2013).

In the southeastern United States, and in Georgia specifically, recent studies demonstrate the importance of archaeology and paleobiology to understanding oyster reef loss and resilience as well as the Native American inhabitants’ role in shaping these environments over the past 5,000 years, often sustainably (Savarese et al. Reference Savarese, Walker, Stingu, Marquardt and Thompson2016; Thompson et al. Reference Thompson, Rick, Garland, Thomas, Smith, Bergh and Sanger2020). Native American communities—and other populations, such as the Gullah Geechee—still maintain a connection to vulnerable archaeological sites in the region. Similar to coastal scientists and conservationists around the world (Cook et al. Reference Cook, Johnston and Selby2019), we in the Southeast are in a race against the rising seas to develop better tools and strategies to more effectively understand and protect these landscapes (Cook Hale Reference Cook Hale, Thulman and Garrison2019; Cook Hale and Sanger Reference Cook Hale and Sanger2020; Faught Reference Faught2004; Garrison and Cook Hale Reference Garrison, Cook Hale, Thulman and Garrison2019; Garrison et al. Reference Garrison, Cook Hale, Faught, Lang and Sayer2013; Howland and Thompson Reference Howland and Thompson2024; Reeder-Myers and McCoy Reference Reeder-Myers and McCoy2019).

Human activities, especially since the beginning of the European colonization of the Americas and through the industrial revolution, have greatly accelerated the pace of climate-related environmental change (Crutzen and Steffen Reference Crutzen and Steffen2003; Lightfoot et al. Reference Lightfoot, Panich, Schneider and Gonzalez2013). Georgia's coastline is experiencing a suite of deleterious effects from increased global mean sea-level rise (GMSLR) and a warming earth (Binita et al. Reference Binita, Marshall Shepherd and Gaither2015): changing wetlands (Robinson et al. Reference Robinson, Dilkina and Moreno-Cruz2020), increased salinization and acidification, cultural and environmental resource submergence, nuisance and event-based flooding, and shoreline erosion and beach migration (Robinson et al. Reference Robinson, Dilkina and Moreno-Cruz2020). The effects of these global climate changes will be severe and permanent to the many and varied cultural heritage sites on the Georgia coast (Anderson and Bissett Reference Anderson and Bissett2015; Duggins Reference Duggins2012; Erlandson Reference Erlandson2012; Intergovernmental Panel on Climate Change [IPCC] 2014, 2018, 2023; Rudd et al. Reference Rudd, Moore, Rochberg, Bianchi-Fossati, Brown, D'Onofrio and Furman2018). A proximity map of the study area, including an Optimized Hot Spot Density analysis, was calculated as a percentage of the total known cultural heritage sites in the area of interest, with data obtained from GNAHRGIS to depict the concentrations of archaeological surveys that have identified the location of cultural heritage sites—not the true reality of the presence of a resource (Figure 1).

Figure 1.

Red indicates a hot spot of a relative concentration of known cultural heritage sites, whereas blue indicates the presence of cultural heritage sites but in a relatively low concentration. Transparency indicates a statistically not significant relative density of cultural heritage sites.

Predictive modeling of coastal change allows anticipation of climate-related impacts to the coast. In this article, we compare the efficacy between results from traditional passive “bathtub modeling” with hyper-local dynamic modeling to examine their ability to predict the impact of sea-level rise on cultural heritage sites. Using these methods, we also pose the following questions:

1. (1) Which archaeological sites on the Georgia coast will be threatened, damaged, or destroyed?

2. (2) When, by what process, and how rapidly will those sites be threatened, damaged, or destroyed?

In addition, we implement language adopted from the National Park Service (Rockman et al. Reference Rockman, Morgan, Ziaja, Hambrecht and Meadow2016) to create an archaeological triage assessment (ATA) based on impacts of shifting National Wetland Inventory (NWI) categories over time obtained via results from SLAMM. Our results demonstrate the need for fine-grain regional ecological modeling to be included in assessments of impacts to coastal heritage sites. In this article, we first contextualize the impacts of wetland reallocation and more nuanced effects of local sea-level-rise estimates, provide a step-by-step guide on SLAMM, introduce the ATA, and suggest next steps in the research of local impacts of anthropogenic climate change in coastal and wetland environments. Figure 2 provides a broad overview of the process to use SLAMM output data to inform an ATA to determine where to deploy archaeologists to endangered or at-risk areas that contain or are likely to contain cultural heritage resources.

Figure 2.

A summary flowchart of the steps necessary to conduct a SLAMM analysis.

Although rising water levels are the most direct long-term threat to coastal archaeological sites and other aboveground cultural resources, the majority of more rapid effects to archaeological sites come from wetland reallocation and shoreline erosion, often from significant storm events (Environmental Protection Agency 2024; Reeder-Myers and McCoy Reference Reeder-Myers and McCoy2019). Once the context of an archaeological site is destroyed, the potential to interpret the information it contained is nullified (Hambrecht and Rockman Reference Hambrecht and Rockman2017). The loss of information around the artifacts negates the potential for meaningful and comparative interpretation: without the context embedded within a site, artifacts themselves may mean little (Beavers et al. Reference Beavers, Babson and Schupp2016; Boivin et al. Reference Boivin, Zeder, Fuller, Crowther, Larson, Erlandson, Denhami and Petraglia2016; Dawson et al. Reference Dawson, Hambly, Kelley, Lees and Miller2020; Hollesen et al. Reference Hollesen, Callanan, Dawson, Fenger-Nielsen, Max Friesen, Jensen, Markham, Martens, Pitulko and Rockman2018; Rockman and Hritz Reference Rockman and Hritz2020). The first step in protecting cultural heritage is knowing when coastal managers need to deploy conservation and mitigation efforts. An accurate timeline of site-specific climate impacts is essential to create an effective planning system.

Climate-related changes to the Georgia coast are occurring quickly and in robust, multivariate ways. The coast is extremely vulnerable to SLR, with its more than 160 km (100 mi.) of coastline, 3,772 km (2,344 mi.) of shoreline, 14 barrier islands, nine tidal marsh and open water estuaries, and approximately 5 km (3 mi.) of tidal shoreline. Furthermore, there are also over 1,400 marsh back-barrier islands in the tidal area of the coast, which make up about 12% of the marsh landscape, serving as keystone structures in the ecosystem as well as being the location of many archaeological sites (Georgia Archaeological State Site File [GASF] 2024; Thompson et al. Reference Thompson, Turck, DePratter, Thompson and Waggoner2013). The region is characterized by a low overall topographic elevation, especially on smaller marsh islands. This makes this region vulnerable to increased storm frequency and larger storm surges and sea-level rise, all of which contribute to compressed sediments and sinking landforms (Anderson et al. Reference Anderson, Fletcher, Barbee, Romine, Lemmo and Delevaux2018; Howland and Thompson Reference Howland and Thompson2024; Reeder-Myers and McCoy Reference Reeder-Myers and McCoy2019). Many tools are available, often for free, to allow users to estimate the impact of sea-level rise in a specific area. These include passive inundation models, colloquially referred to as “bathtub” models, which assume equal distribution of increased water volume with measurements derived from global estimates of rising sea levels (Anderson et al. Reference Anderson, Bissett, Yerka, Wells, Kansa, Kansa, Myers, Carl DeMuth and White2017; Marcy et al. Reference Marcy, Herold, Waters, Brooks, Hadley, Pendleton and Schmid2011; National Oceanic and Atmospheric Administration [NOAA] Office for Coastal Management 2017a, 2017b, 2023; Schmid et al. Reference Schmid, Hadley and Wijekoon2011, Reference Schmid, C. Hadley and Waters2014; Spring 2018). However, for the purposes of archaeological research at a site-specific scale, the data input into these models are often too generalized and do not consider the impact of coastal biophysical feedback loops. Consequently, these models often severely underestimate land exposed and lost.

How-To: Sea Levels Affecting Marshes Model (SLAMM)

SLAMM is an open-source program based on a mathematical model that combines lidar-derived bare earth elevation, slope, and National Wetlands Inventory (NWI) categories to estimate potential impacts of sea-level rise to shorelines and wetlands. The program has been in production since the mid-1980s, and modeling within the program began in 2006. Whereas the spatial data processing takes place in a GIS or R software package, SLAMM is a standalone graphical user interface (GUI) that can be manipulated by the user, allowing additional data, such as salinity, overwash scenarios, and historic erosion and accretion rates. SLAMM itself is a network of complex decision trees that “incorporate geometric and qualitative relationships to represent transfers among coastal classes” (Warren Pinnacle Consulting 2017). Although not required, some models, such as the one presented here, adopt data from the marshes equilibrium model (MEM) to estimate changes to carbon sequestration and marsh health with SLAMM results from long-term impacts to shorelines due to sea-level change.

The following section provides an overview guide to processing each major dataset that goes into SLAMM. The overall key to preparing data for the model is to ensure all cell sizes and projections are the same and are based in meters (e.g., UTM and NAVD88). The following steps can be streamlined by setting the environments (e.g., cell size, raster bands, clipping type, and extent) of the map before importing any data or just after importing the DEM or lidar. All data, whether processed in R or ArcGIS, are exported as an ASCII file prior to importation into SLAMM. Completed model files from the SLAMM GUI may be exported in ASCII, CSV, or Shapefile file formats, depending on user preference.

Digital Elevation Model (DEM) and Slope

The majority of data for SLAMM are from DEMs derived from lidar, with cells at the smallest functional size. Lidar data are most often obtained through a state repository system, 3DEP repositories, or USGS Earth Explorer (https://earthexplorer.usgs.gov/). Depending on the size of the area of interest, data must be preprocessed to convert text-based point clouds files to a LAS file format prior to converting lidar into a bare earth—or ground return—DEM raster file (see Figure 3 for a guide to process lidar data into DEM and slope ASCII files). ESRI ArcMap tutorials are valuable resources in learning to process the data (e.g., ArcMap 2024). Because of the high quality of DEM data, elevation uncertainty modeling was not undertaken as part of this project, as recommended in the SLAMM model user guide. If uncertainty elevation data needs to take place, adjust the DEM to the mean tide level as zero using the National Oceanic and Atmospheric Administration's (NOAA) vertical datum transformation tool (NOAA Office for Coastal Management 2024). Slope, a raster-based calculation of the gradient or steepness between nine raster cells, was derived from this DEM using the slope tool in the spatial analyst toolbox.

Figure 3.

The basic process to acquire, convert, process, and prepare DEM and slope files from lidar.

National Wetlands Inventory (NWI) and SLAMM Environmental Categories

SLAMM simulates the processes involved in wetland conversions that will occur with long-term sea-level rise—namely, inundation, erosion, accretion, soil saturation, and barrier island overwash—which is why the National Wetlands Inventory's (NWI) Cowardin classification systems are central to processing input data for the model. SLAMM contains 23 land-cover classes, but not all categories may apply to every project area. To convert the many thousands of potential NWI classes to 23 SLAMM classes, attribute tables from the NWI GIS layer are exported into a CSV Excel file (Warren Pinnacle Consulting 2016:78). The Excel crosswalk function, VLOOKUP, is then used to connect the classes to one of 23 SLAMM categories. NWI categories may not always directly match one of the SLAMM categories, requiring the SLAMM user to perform quality control on each entry. SLAMM categories are then reimported into GIS and merged with the modified NWI vector file. See Figure 4 for a more in-depth flowchart of steps to process the NWI dataset along with the SLAMM Technical Manual (Warren Pinnacle Consulting 2016:78–79).

Figure 4.

A summary of the process of acquiring National Wetlands Inventory data and converting those categories to SLAMM Land Cover classes.

Marshes Equilibrium Model (MEM)

A Marshes Equilibrium Model (MEM) was used to calibrate/generate accretion response curves for salt, brackish, and tidal fresh marshes from all available Georgia rivers. Given that wetland categories change from generally dry lands to environments that will become predominantly tidal swamps, areas of open estuarine and ocean water, and regularly flooded saltmarsh, archaeological sites are subjected to inundation, biotic change, and shoreline migration. For this reason, marsh growth and erosion are important parameters to consider (Morris et al. Reference Morris, Sundareshwar, Nietch, Kjerfve and Cahoon2002, Reference Morris, Edwards, Crooks and Reyes2012). The northern and southern halves of the Georgia coast were processed and then mosaicked together. Wetland vertical accretion in salt, brackish, and fresh marshes were modeled in the Marsh Equilibrium Model 5.4 utilizing data on sediment, biomass, and accretion from the Savannah River, Ogeechee River, South Newport River, Altamaha River, and Satilla River watersheds. The predicted accretion rates were then applied to a spatially explicit model of SLR to model 1 m of sea-level rise on the Georgia coast using the Sea Levels Affecting Marshes Model 6.3 (SLAMM; Warren Pinnacle Consulting 2017) using the vegetation-corrected DEM as the base elevation for modeling. This dynamic modeling suite is founded in quantifying shoreline changes due to multimodal changes driven by wetland conversions (Anderson et al. Reference Anderson, Fletcher, Barbee, Romine, Lemmo and Delevaux2018; Kirwan et al. Reference Kirwan, Temmerman, Skeehan, Guntenspergen and Fagherazzi2016).

Case Study of Georgia's Coast: SLAMM and Archaeological Triage Assessment (ATA)

Our new work provides a temporally sensitive and regionally specific model to estimate which of Georgia's cultural resources will be threatened, damaged, or destroyed based on National Wetland Inventory (NWI) land-cover classes derived from SLAMM crosswalk tables (Cowardin et al. Reference Cowardin, Carter, Golet and LaRoe1979). The analysis spans the full 161 km of the Georgia coastline, state border to state border, and extends inland approximately 30 km from the easternmost edge of the barrier islands. In this regional approach to coast-wide resiliency planning, a hybridized modeling approach was first created to parameterize and calibrate marsh accretion models for coastal Georgia marshes and its five major estuaries. Although we anticipate updated input models to encompass more modern IPCC estimates, recent data from Sweet et alia (Reference Sweet, Hamlington, Kopp, Weaver, Barnard, Bekaert and Brooks2022) and Corkran et alia (Reference Corkran, Potter and Schmutz2023) illustrate high levels of comparability between estimates in the study area.

Archaeological site locations, recorded as point vector data, were drawn from Georgia's Natural, Archaeological, and Historic Resources GIS (GNAHRGIS; GASF 2024). This tool is a secured interactive web-based registry for the natural, cultural, and archaeological resources of Georgia. GNAHRGIS points were used in a GIS with SLAMM models in 25-year increments (2006, 2025, 2050, 2075, 2100). Archaeological site counts within each SLAMM/NWI polygon were tabulated for each SLAMM 25-year time slice for the dynamic model. Table 1 provides conversions from SLAMM land-cover classes to anticipated impacts (no impact, damaged, destroyed) to cultural resources. Once a site is fully underwater, it is assumed destroyed; when a site undergoes consistent wave action, it is assumed damaged or destroyed; a site within an area susceptible to flooding or significant storm surge is considered threatened; sites on dry land are expected to sustain no to minimal impacts from that wetland category type. GNAHRGIS site counts were tabulated to assess the number of archaeological sites threatened according to the passive bathtub model. Passive modeling does not accurately illustrate the details of anticipated shoreline and wetlands changes, whereas dynamic modeling provides the precision necessary to deploy targeted reconnaissance, monitoring, and survey crews to document and protect cultural heritage sites.

Table 1.

NWI Category Conversions to the Archaeological Triage Assessment.

Note: Threat assessments are based on the anticipated damaged to cultural heritage sites that lie within each of the 22 NWI categories (NWI category 21 is not used on the coast). Once a site is fully underwater, it is assumed destroyed; when a site undergoes consistent wave action, it is assumed damaged or destroyed; a site within an area susceptible to flooding or significant storm surge is considered threatened; sites on dry land are expected to sustain no to minimal impacts from that wetland category type.

Results: Passive versus Dynamic Models to Estimate Cultural Resource Vulnerability

Passive and dynamic models are fundamentally different. Although both models estimate the broad effects of SLR, the two models rely on different methods of measuring change over time. The passive model captures blanket rates of SLR, whereas SLAMM captures wetland reallocation. These differences are especially apparent given that this study relies on previously generated estimates of SLR impacts to measure the effect of SLR to cultural heritage sites. The quantity of cultural resources impacted by multiple scenarios of SLR using the SLAMM model are presented in Table 2 and Figure 5. Nearly 40% (n = 1,409) of Georgia's known coastal cultural heritage sites will be inundated or destroyed by 2100, based on the moderate estimate of a local 1.5 m sea-level rise. The majority of resources impacted between now and 2100 were documented in the undeveloped dry lands class; dry wetland habitat categories are environments that will oftentimes become predominantly tidal swamps, areas of open estuarine and ocean water, and regularly flooded saltmarsh.

Table 2.

Summary of the Cumulative Counts of Coastal Sites That Will Be Unimpacted, Threatened, or Destroyed According to Certain Climate Scenarios Using Dynamic SLAMM Estimates.

Figure 5.

SLAMM model output of the Georgia coast with an expected 2 m SLR, illustrating wetland redistribution over time as a response to sea-level rise at 25-year increments to the year 2100.

The overall numerical result from the dynamic and passive models were, ultimately, quite similar (Table 3). The NOAA GMSLR passive model applied to the Georgia coast estimated similar site loss as that predicted by SLAMM, but without the time seriation provided by the SLAMM model. The temporal, spatial, and ecological details that are made available by using a dynamic predictive model as the basis for archaeological planning and mediation lead to more accurate assessments of site vulnerability. Sapelo Island, Georgia, is an excellent case study of the direct difference between passive and dynamic modeling strategies when creating archaeological triage systems (Figure 6). Passive modeling does not accurately illustrate the details of anticipated shoreline and wetlands changes, whereas dynamic modeling provides modeling precision necessary to deploy targeted reconnaissance, monitoring, and survey crews to document and protect cultural heritage sites. The latter is clearly more detailed and provides a much better guide for triaging sites. Rather than estimates that call for total destruction, SLAMM provides both temporal guidelines for changes over time and 10 m horizontal resolution of anticipated change.

Table 3.

Cumulative Estimates of Cultural Heritage Sites Destroyed by a 1 m, 1.5 m, and 2 m Global Mean Sea-Level Rise According to Passive and Dynamic Models.

Figure 6.

Comparison of (A) passive versus (B) dynamic and (C) archaeological triage assessment conversion results for the Georgia coastline, 2 m SLR. The archaeological triage assessment interprets results from dynamic modeling to estimate the impact of specific wetland changes to cultural heritage sites (green = no impacts, yellow = threatened, orange = damaged, red = destroyed). See Table 1 for category conversions.

This comparison between modeling strategies also illustrates the resiliency of the marshes. This critical resource adapts to SLR due to elevation-sediment-vegetation feedback cycles. The adaptability of the marshes and other environmental parameters serve to delay, reduce, or prevent site loss under certain SLR scenarios. SLAMM provides opportunities to measure the degree to which site preservation relies on marshes as protective barriers against storm surge, wave action, flood depth, and other shoreline loss mechanisms.

Although archaeologists now consider many inundated resources as a total or near total loss, we propose that there is more nuance to the mechanism of site degradation. Site destruction need not be binary—preserved or not preserved. Instead, we recommend that in addition to direct observations of the nuance of change, future studies should turn to the submergence of the continental shelf since the Last Glacial Maximum as an example of mechanisms of large-scale environmental change and the impact to site preservation. Archaeologists are now exploring the late submerged glacial- and early postglacial-era continental shelf. As they explore this record, they are finding and assessing the contents and condition of formerly terrestrial sites in submerged settings, as well as creating predictive models for site location (e.g., Anderson and Bissett Reference Anderson and Bissett2015; Faught Reference Faught2004).

In many ways, the inundation of cultural heritage due to sea-level rise is similar to situations archaeologists faced with the large dam projects of the 1930s and through the 1980s. These dam projects, funded by state and federal agencies, destroyed many archaeological sites in Georgia and across the US Southeast—and beyond (such as the Aswan High Dam in Egypt). However, in contrast to what is happening along our coastlines, those infrastructure projects attracted millions of dollars to examine and excavate sites ahead of inundation, leading to critical advancements in our understanding of the history and ecology of these areas (e.g., Anderson and Joseph Reference Anderson and Joseph1988; Hassan Reference Hassan2007). Unfortunately, no such coordinated efforts or funding is being provided as of yet for the archaeological losses along our coastlines. As such, we need to develop more sophisticated models to aid in our efforts to mitigate the effects of SLR on the coast's cultural heritage sites.