Mapping Moonshine in Hell Hole Swamp: Preliminary Modeling of Clandestine Liquor Distillation Sites in Coastal

Katherine G. Parker and Jordan L. Schaefer

SEAC 2019 Still sites have received limited archaeological attention under the assumption that they cannot contribute to significant research questions. However, several still sites related to clandestine commercial-scale production of liquor in the twentieth century were determined to be eligible for the National Register of Historic Places in 2015. These sites were identified in areas classified by traditional site predictive models as having low archaeological potential. Using this network of sites as a case study, this paper will examine preliminary attempts to predict additional still site locations in order to build a more robust data set for these historically under-investigated sites.

In order to explain the reasons why we’re attempting to model still site locations, I’m going to start by defining still sites and their common attributes, and how these remains have contributed to the under-identification of still site as archaeological resources. This paper will then discuss how predictive models shape the ways in which surveys are conducted and, as a result, what kinds of sites are likely to be identified. Finally, we will conclude by comparing attributes of a widely used predictive model in the South Carolina Lowcountry to site selection priorities for clandestine distillation in order to explore important considerations for locating future still sites.

I. Introduction Clandestine liquor production in South Carolina can be dated back as far as the 1750s, when it was predominantly produced and consumed by small-scale family operations that faced occasional and incidental interference from the law (Miles 2015; Teal and Wallace 2005). Spikes in clandestine production during the 18th and early 19th century were largely responses to taxation on liquor that intended to offset the costs of maintaining the colonial government or financing postwar reconstruction efforts (Meacham 2013; Okrent 2010; Teal and Wallace 2005). The market for moonshine saw a further increase in 1919 with the ratification of the 18th amendment to the Constitution, which prohibited the sale and consumption of alcoholic beverages nationwide. Urban centers in South Carolina such as Charleston put up strong resistance to the economic domination and limits to personal liberty created by prohibition policies (Teal and Wallace 2005:26). Rather than temper the desire for alcohol, these restrictions created a demand that moonshining was well-positioned to fill.

Due to the historically illegal nature of moonshine production, archaeological evidence of clandestine liquor production is often ephemeral (Espenshade 2016; Moonshine Heritage 2011; Parker 2018). Common indicators are the remnants of both operational and concealment elements. Operational elements consisted of three main parts related to the function of the still: the furnace or firebox, containers and condensers, and piping. The firebox was built below the main still apparatus and was used to heat the fermented grains, or mash, to produce a spiritous vapor. Fireboxes could either be dug out below the still or built up, and common construction materials included clay, stone, brick, and/or handmade cinder blocks. Condensers and containers represent the most common operational element identified across still sites. These include the main still apparatus, which housed the mash, as well as components that held water used to condense the spiritous vapors and collect the condensed byproducts of the distillation process. Wooden barrels were common for condensers and containers, particularly in earlier stills, though a wide range of materials including large steel drums, boilers, and buckets were also ubiquitous. Copper, which was also commonly used to construct the still apparatus, was highly valuable and as a result was often removed from the site when the still became defunct (Moonshine Heritage 2011; Shuler 2007; Wigginton 1972). Piping was used for a variety of functions, such as distributing water and distillates to various parts of still and dispersing tell-tale smoke produced by the furnace (Wigginton 1972).

Concealment elements varied widely based on the size, length, and intensity of the operation as well as by the inherent elements of the landscape that afforded natural cover (French 2019; Parker 2018; Wigginton 1972). The degree of investment in concealment efforts appears to have had a positive correlation to the degree of investment, temporally or volumetrically, in the operation. Less-intensive operations typically employed expedient concealment measures such as hollowed-out stream banks or screens created from vegetation, whereas intensive operations exploited more durable forms of concealment such as outbuildings, caves, or subterranean structures to minimize attention (French 2013; Wigginton 1972).

As archaeological resources, the remnants of still sites have frequently been mistaken as modern trash dumps—there’s even a website titled, “5 Ways to Tell If That Pile of Garbage You Found in the Woods Is a Moonshine Still” (Moonshine Heritage 2011). Stills have also been mistakenly recorded as other site types, such as site 31JN13** in Jones County, North Carolina, which was recorded as a pottery kiln despite the lack of definitive evidence for pottery production (Cao et al. 2017). Even when identified correctly as still sites, these archaeological resources have been presumed to lack significance via National Register of Historic Places (NRHP) criteria. This presumption is problematic for two reasons: one, it reflects an emphasis on using Criterion D, the potential to contribute significant data, as the dominant consideration for archaeological resource significance, while minimizing the consideration that still sites have connections to significant historical events in the past or significant people in the community; and two, it presumes that stills are ubiquitous and relatively modern, and are therefore well-understood. As a result, the lack of formally and systematically documented still sites means that limited examples exist by which to compare subsequently identified sites (Espenshade 2016; Parker 2018). Without these examples, how can we accurately assess morphological relationships between still sites? How can we evaluate how intensively, both for scale and time, that these sites were used? How can we compare the material culture between these sites with regards to economic investment, access, and status? There are countless other questions that remain unanswerable without the existence of a more substantial dataset; the issue, then, is how to develop this dataset, and whether present site location strategies are effective for this site type. II. Still Sites in the FMNF The National Forest, which spans more than 259,000 acres in the South Carolina Coastal Plain, has had more than 2,000 archaeological sites recorded in the last 60 years, yet as of 2019, only five of these resources are still sites (South Carolina Institute for Archaeology & Anthropology (SCIAA) 2004; USDA 2019). Two of these sites were identified during a Phase I survey conducted by USFS archaeologists in 1982, while the remaining three were documented during a Phase I survey conducted by Archaeological Consultant of the Carolinas, Inc., in 2015 (Pasquill 1982; Stewart et al. 2017). One of the stills from the more recent survey, site 38BK3020, was actually recommended eligible for the NRHP under Criteria A and B, though consideration under Criterion D required further investigation beyond the scope of the initial methodology (Stewart et al. 2017:127). The extant remains [shown here] indicate intensive modifications to the landscape and a relatively extensive debris scatter related to both the construction and concealment of the distillation operation. Despite the degree of material evidence for this site, site 38BK3020 was nearly missed during the initial survey because it was located at the edge of the survey boundary in an area classified as having a low potential for archaeological resources by traditional site location models (Parker 2018; Stewart et al. 2017).

III. Methods In order to demonstrate how an overreliance on one particular model can lead to dismissal of still sites, we compare two predictive models: one made from ethnographic information and O’Donoughue’s (2008) model for archaeological site potential, which is frequently referenced as a guideline for site surveys in Francis Marion. Predictive models are a means through which we can identify areas in the landscape that have a high potential for containing archaeological sites. The process usually involves isolating specific locations that tend to be favorable for a particular site-type, creating layers for these variables, adding these layers together, and then identifying where they overlap. This produces a map that highlights regions with multiple factors that are favorable for that specific site. Predictive models are particularly useful for projecting known patterns or relationships into past times or places (Warren and Asch 200:6). These models were made using Esri’s ArcGIS software and the tools available within it.

The study area roughly covers the boundary of the Cane Gulley 3 survey area, which was a survey carried out in 2015 that located three out of the five still sites. It is located immediately northwest of Hell Hole Swamp, a known hotbed for moonshining in the region, and covers the property of the Villeponteaux family, who were known to be prominent moonshiners. One foreseeable problem for this project is the fact that there are only five known stills in the study area. To truly gauge the efficacy of a predictive model, it helps to have a robust sample size through which statistical testing can be done. The lack of still sites could mean that our resulting model is speculative, but it nonetheless demonstrates the need for additional still sites to be identified and recorded so that these preliminary models can be refined.

Our model (The Moonshine Model), uses criteria derived from ethnographies that describe ideal locations for stills. Oral histories indicate that still sites were typically placed within one mile of a road, relatively close to streams or running water, and situated in inconspicuous topographic locations. To determine which regions are close to roads, a 1940 map of Berkeley County from the South Carolina State Highway Department was georeferenced and used as a source to determine which roads would have existed at the time. A one-mile buffer was created around these roads and converted into a raster file, creating the first variable. Proximity to streams was defined by measuring whichever site held the longest distance from a stream and creating a buffer at that value. In this case, 38BK2978 was the farthest from any given stream at 0.42 miles, so a half-mile buffer was formed around the streams and converted to a raster layer, creating the second variable.

In order to figure out which regions have low noticeability, topographic prominence was calculated for the study area. Llobera (2001:1007) describes topographic prominence as “a function of height differential between an individual and his/her surroundings as apprehended from the individual’s point of view.” In other words, it is basically a measure of whether a location is higher or lower than its surrounding area. Areas with low topographic prominence are generally more difficult to see from any given position, making them ideal spots for illicit activities. Topographic prominence was calculated by running a focal statistics tool on a digital elevation model (DEM) of the study area. The tool was configured to return a value for each individual cell based on the mean value of all cells within a 200-meter radius. This raster was then subtracted from the original DEM so as to create a new raster where each cell’s value represents the height differential from those surrounding it. The higher values (white colors) represent spots that have a higher-than-average elevation than their surroundings, while lower values (darker colors) have a lower-than-average elevation. Areas with topographic prominence values lower than 0 were isolated and used to create the third variable for this model. All three variables were added together via the raster calculator, which produced a predictive model highlighting regions with high probability for containing still sites.

O’Donoughue’s (2008) model posits areas of high archaeological potential as having excessively drained or well-drained soils or within 90 meters of a wetland on moderately well drained soils. Areas of moderate potential are more than 90 meters from a wetland on moderately well drained soils; within 90 meters of a wetland on somewhat poorly drained soils; within 90 meters of a road on somewhat poorly drained, poorly drained, or very poorly drained soils; and within 90 meters of excessively drained, well-drained, or moderately well drained soils on somewhat poorly drained, poorly drained, or very poorly drained soils. To determine which regions in the study area meet these criteria, the USGS’s National Landcover Database (NLCD) and Soil Database. The aforementioned criteria were identified and added together, which resulted in a different predictive model for the study area.

IV. Results Both models represent areas of archaeological potential in terms of three categories: high, moderate, and low potential. For our model, 33.8% of the study area (44,801.2 km) was classified as high potential, 48.4% (64,232.1 km) was classified as moderate potential, and 17.8% (23,638.1 km) was classified as low potential. Using this model, two of the still sites exist within the high potential zone, and three exist within the moderate potential zone. Meanwhile on O’Donoughue’s (2008) model, 16.5% was high potential, 42% was moderate potential, and 41.5% was low potential. For this model, two sites reside within the high potential zone, two within the moderate potential zone, and one within the low potential zone.

O’Donoughue’s model identifies a much more confined area than our model, which is certainly useful in situations where we are under budget/time constraints or need to know where to specifically focus our efforts. That being said, it is perhaps too constricting. Even though O’Donoughue’s model was able to capture most of the still sites (4/5), a few were right on the boundary between moderate potential and low potential, meaning they could be perceived as low potential while in the field. By looking at the areas surrounding the still sites, we see that O’Donoughue’s model, for the most part, indicates that the general region in which the known still reside has a relatively low probability for containing archaeological sites. Meanwhile, our model portrays this region as much more likely to contain other still sites. When we cross-reference our model with regions that have been surveyed, we find a noticeable lack of survey coverage in areas around the known still sites. O’Donoughue’s model suggests additional surveys in these areas would not prove fruitful, but our model suggests otherwise.

One other noteworthy section is the eastern portion of the study area, where extensive surveys have been conducted in a region that our model claims has a high likelihood for containing still sites. Despite this coverage, no stills have been formally identified in this region. If we look at the same region in O’Donoughue’s model, it’s mostly classified as low potential. If these surveys were conducted with O’Donoghue’s model in mind, then it’s possible that they were covered at a reduced effort. Had they been classified as high potential and covered with a tighter interval, there is a greater likelihood that potential still sites would have been identified.

V. Conclusion We argue that no single predictive model can account for all archaeological sites. Each site-type has its own set of criteria and logic that dictates where it is set up. We should not expect an Archaic Period settlement and a clandestine historical resource to abide by the same location selection. At the end of the day, it’s only a model. The location for a given site is usually selected for reasons that make sense to those making it, and we therefore need to be aware of these variations in order to build effective predictive models. While we recognize that O’Donoughue’s model has its merits, we contend that people often consider other variables for location selection that are not strictly related to soil. References Cited

South Carolina Institute of Archaeology & Anthropology (SCIAA) 2004 SC ArchSite. Electronic Document, accessed May 1, 2017.

Cao, Luan, Dawn Reid, and Katherine G. Parker 2017 Archaeological Survey of the Longleaf Timber Sale Analysis Area, Croatan Ranger District, Croatan National Forest, Jones County, North Carolina. Archaeological Consultants of the Carolinas, Inc., Clayton, NC.

Espenshade, Christopher T 2016 Still Sites: Why Even Bother? Paper presented at the Middle Atlantic Archaeological Conference, 2016, Virginia Beach, VA.

French, Kirk 2013 The Moonshine Archaeology Project: Investigating the Social and Economic Impacts of Illicit Whiskey Production in Western North Carolina. Electronic Document, http://moonshinearchaeology.blogspot.com/2013/12/investigating-social-and- economic.html, accessed February 1, 2017. 2019 A Round for All My Friends: Understanding the Economic Impact of Moonshine Production. In The Historical Archaeology of Shadow and Intimate Economies, edited by James A. Nyman, Kevin R. Fogle, and Mary Carolyn Beaudry, pp. 51–66. University Press of Florida, Gainesville.

Llobera, Marcos 2001 Building Past Landscape Perception with GIS: Understanding Topographic Prominence. Journal of Archaeological Science 28:1005-1014.

Meacham, Sarah Hand 2013 Every Home a Distillery: Alcohol, Gender, and Technology in the Colonial Chesapeake. Johns Hopkins University Press, Baltimore, MD.

Miles, Suzannah 2015 Moonshine over Hell Hole Swamp. Charleston Magazine. http://charlestonmag.com/features/moonshine_over_hell_hole_swamp, accessed April 14, 2017.

Moonshine Heritage 2011 5 Ways to Tell If That Pile of Garbage You Found in the Woods Is a Moonshine Still. Electronic Document, http://www.moonshineheritage.com/blog/5-ways-to-tell-if-that- pile-of-garbage-you-found-in-the-woods-is-a-moonshine-still/, accessed February 1, 2017.

Okrent, Daniel 2010 Last Call: The Rise and Fall of Prohibition. Scribner, New York.

Parker, Katherine G. 2018 Shining in the Tar Woods: An Examination of Illicit Liquor Distillation Sites in the Francis Marion National Forest. Paper presented at the 51st Annual Meeting of the Society for Historical Archaeology Conference on Historic and Underwater Archaeology, 2018, New Orleans, LA.

Pasquill, Robert G. 1982 An Archaeological Survey of Selected Timber Stands in Compartment 53, Witherbee District, Francis Marion National Forest. United States Forest Service.

Shuler, William Dennis 2007 Short Stories About Life in Berkeley County, South Carolina. AuthorHouse, Bloomington, IN.

Stewart, James, Katherine G. Parker, and Dawn Reid 2017 Cultural Resource Survey of the Cane Gully Analysis Area Part III, Francis Marion National Forest. Archaeological Consultants of the Carolinas, Inc., Clayton, NC.

Teal, Harvey S, and Rita Wallace 2005 The South Carolina Dispensary and Embossed S.C. Whiskey Bottles and Jugs, 1865-1915. Midlands Printing Company, Camden, SC.

USDA 2019 Francis Marion National Forest. Electronic Document, https://www.fs.usda.gov/detail/scnfs/home/?cid=FSBDEV3_037393, accessed November 1, 2019.

Warren, Robert E., and David L. Asch 2000 A Predictive Model of Archaeological Site Location in the Eastern Prairie Peninsula. In Practical Applications of GIS for Archaeologists: A Predictive Modeling Toolkit, edited by Konnie Wescott and R. Joe Brandon, pp. 5-32. Taylor and Francis, London.

Wigginton, Eliot (editor) 1972 Hog Dressing, Log Cabin Building, Mountain Crafts and Foods, Planting by the Signs, Snake Lore, Hunting Tales, Faith Healing, Moonshining, and Other Affairs of Plain Living. The Foxfire Book. Anchor Press/Doubleday, Garden City, NY.