Rediscovering Rural Appalachian Communities with Historical GIS

Rediscovering Rural Appalachian Communities with Historical GIS GEORGE TOWERS Concord University From the late 19th century until World War Two, pográficos históricos utilizando SIG identifica ad- agrarian southern Appalachia was a patchwork ecuadamente los límites pasados de los barrios of small, close-knit farm communities. This his- agrícolas del sur de los Apalaches. Utilizando la toric rural settlement pattern is locally recorded función de análisis de asignación de costos de in community case studies by ethnographers and ArcGIS, regiones de menor costo son generadas historical geographers but has not been mapped alrededor de los nodos de los barrios, basadas en systematically. This paper explores the hypothesis el costo de la energía de viajes a pie con respecto a that GIS analysis of historic topographic maps distancia y pendiente. Estos vecindarios agrícolas adequately identifies the boundaries of bygone prospectos se asemejan a las descripciones espa- southern Appalachian agricultural neighbor- ciales de los etnógrafos y los geógrafos históricos. hoods. Using the ArcGIS cost allocation analysis Cartografiar barrios agrícolas históricos en los function, least cost regions are generated around Apalaches provee una importante base para la neighborhood nodes based on the energy cost of comparación de pasados y presentes patrones de foot travel relative to distance and slope. These asentamiento. Este método de investigación es sig- prospective agricultural neighborhoods closely nificativo porque es fácilmente replicable y puede match ethnographers and historical geographers’ ser empleado a través de los Apalaches del Sur y spatial descriptions. Mapping historic Appala- del siglo pasado. chian agricultural neighborhoods provides an important basis for comparison with past and present settlement patterns. The research method is significant because it is easily replicated and may be extended across southern Appalachia and the key words: historical GIS, Appalachia, agricultural neighborhoods, topographic maps, West Virginia, landscape, social history, farming past century. Desde finales del siglo 19 hasta la Segunda Guerra Mundial, el sur agrario de los Apalaches era un mosaico de comunidades agrícolas pequeñas y muy unidas. Este patrón histórico de asentamiento rural es registrado a nivel local en estudios de etnógrafos y geógrafos históricos sobre casos comunitarios, sinembargo no ha sido cartografiada de forma sistemática. Este trabajo explora la hipótesis de que el análisis mapas tosoutheastern geographer, 50(1) 2010: pp. 58–82 introduction This research assesses the hypothesis that historical GIS (HGIS) may be used to map an extinct and iconic American landscape: the southern Appalachian agricultural neighborhoods of a century ago. HGIS enables researchers to ask geographical questions of history and supports its answers with maps and spatial analy- Rediscovering Rural Appalachian Communities sis. Over the last two decades, HGIS has evolved from a research method to a wellrecognized interdisciplinary field of study (Baker 2003; Colten et al. 2005; Gregory and Healey 2007; Knowles 2008). Occupying five or six square kilometers each, agricultural neighborhoods of a few dozen farm families ordered southern Appalachia’s rural social landscape. ‘‘Preindustrial mountain society had been based upon a system of small, independent family farms, clustered together in diffuse open-country neighborhoods’’ (Eller 1982, p 194). Neighborhoods, according to James S. Brown, a leader in mid-20th century southern Appalachian ethnography, are defined by social solidarity, interdependence, and a shared community of interests (1988). Throughout the region, anthropologists and sociologists reported that neighborhood solidarity was cemented through family ties and Protestant fundamentalism while subsistence agriculture engendered the neighborly interdependence that fostered a community of interests (Pearsall 1959; Stephenson 1968; Kaplan 1971; Beaver 1976; Photiadis 1980; Martin 1984). Ethnographers’ emphasis on social organization led them to the label ‘‘kinship neighborhoods.’’ The current research, however, focuses on the cultural landscape and will instead use the term ‘‘agricultural neighborhoods’’ to distinguish this settlement pattern from other regional rural communities like hamlets and coal camps while retaining an emphasis on local social integration. Agricultural neighborhoods were a passing phenomena, existing between the Civil War and World War Two. Previously, agricultural communities spread them- 59 selves over much more territory. For example, early 19th century farm neighborhoods in Tazewell County, Virginia of 25 to 30 households took up 25 to 65 square kilometers (Mann 1995). By the late 1800s, the labor demands of low technology subsistence agriculture had sustained population growth sufficient to crowd the countryside (Salstrom 1994; Billings and Blee 2000). Farms were subdivided among family members. For instance, a re-visitation of eastern Kentucky’s ‘‘Beech Creek’’ neighborhood found that the three farms in the area in 1850 had multiplied more than tenfold by 1942 (Billings and Blee 2000). Agriculture also expanded to exhaust arable land. Martin (1984) provides a case study of this process in his description of Kentucky farmers bringing the isolated Head of Hollybush Hollow into agricultural production in the early 1880s. Coexisting with encroaching coal camps in the first decades of the last century, farm neighborhoods emptied out in the 1940s and 1950s. Farmers and their children found factory jobs and the Great Society of the 1960s declared war on the vestigial Appalachian culture of poverty (Eller 2008). A primarily residential presence— rural sprawl—has since settled over the old landscape of agricultural production. Invaded, abandoned, and obscured, the traditional agricultural neighborhood has ‘‘disappeared from the map’’ (Howell 2003, p 122). A case study of Summers County, West Virginia assesses the hypothesis that HGIS may identify the boundaries of historic southern Appalachian agricultural neighborhoods (see Figure 1). The primary data for this study are century-old 1:62,500 scale U.S. Geological Survey (USGS) topographic maps covering 15 minute quad- Figure 1. Summers County, West Virginia and southern Appalachia. Source for southern Appalachian boundaries: Salstrom 1994. Rediscovering Rural Appalachian Communities rangles of latitude and longitude. HGIS methods are used to convert territory surrounding neighborhood nodes—the country schools and hamlet centers shown on the historic topographic maps—into potential agricultural neighborhoods. HGIS analysis compares the spatial arrangement of the houses, country schools, and churches in the prospective agricultural neighborhoods with the consistent neighborhood settlement patterns described in ethnographic case studies made across southern Appalachia. If corroborated by evidence from ethnography and historical quantitative data, the HGIS methodology may be extended across the southern Appalachian region. Subsequent research may lead to construction of regional settlement pattern datasets for comparative temporal and spatial analysis. study area The Summers County portions of the 1912 Big Bend and Meadow Creek 15 minute USGS quadrangles serve as the study area for historical and geographical reasons (see Figure 2). Historically, these quads were the first mapped in the Appalachian plateau of southern West Virginia. They have been recently scanned and georeferenced by the West Virginia Department of Environmental Protection and the West Virginia GIS Technical Center (Dawson et al. 2007). Dominating the rural landscape depicted on the Big Bend and Meadow Creek maps were diversified family farms. While corn occupied half of the cultivated acreage, farmers also grew wheat, oats, and hay and tended vegetable gardens and fruit trees. Livestock included milk cows, hogs and sheep (Unrau 1996). The farm 61 population of 11,008 was 82 percent of the county’s rural total and 84 percent of rural dwellings were farmhouses. In Forest Hill, Jumping Branch, and Pipestem, the three rural southern magisterial districts without the railroad and without sizable unincorporated villages, more than 90 percent of people lived on farms (U.S. Census 1913). HGIS allows local case studies to be integrated with regional scale investigation, offering the opportunity to assess whether Summers County is representative of late 19th and early 20th century agricultural southern Appalachia. Cunfer (2005) provides an example of this approach by supporting his localized longitudinal case studies of farming practices on the Great Plains with a region-wide HGIS dataset derived from agricultural censuses. The regional boundaries shown in Figure 1 have found general agreement among historians of southern Appalachia (Salstrom 1994; Williams 2001) and are the basis for a county level HGIS dataset developed from decadal census data that ranges from 1880 through 1940 and speaks to farm size and farm density. Variables include the number of acres per farm, the number of farms per square mile, the percent of county land in farms, the growth rate of farms, and the rate of change in average farm size. For each southern Appalachian county, these variables were standardized with z scores (see Tables 1 and 2). This resulted in seven z scores for the farm size and the two farm density measures and six z scores for the two rate of change variables. The absolute values of z scores in each category were then averaged to create a single comparative index of each county’s correspondence to regional norms. By this mea- Figure 2. Study Area: The Summers County portions of the 1912 Big Bend and Meadow Creek Quadrangles. Rediscovering Rural Appalachian Communities 63 Table 1. Agriculture in Summers County and Appalachia, 1880–1940. 1880 1890 1900 1910 1920 1930 1940 Mean farm size (ha): Regional mean 72.1 61.5 47.4 41.7 40.5 37.2 32.8 Mean farm size (ha): Summers County 68.4 57.5 41.7 38.9 39.3 39.3 34.4 Mean farm size (ha): Summers’ Z score –0.13 0.11 –0.19 –0.32 –0.21 –0.09 0.13 Farms/sq. km: Regional mean 1.12 1.30 1.71 1.87 1.83 1.77 2.03 Farms/sq. km: Summers County 1.10 1.39 1.98 2.17 2.12 2.06 2.34 Farms/sq. km: Summers’ Z score 0.39 0.34 –0.04 0.15 0.43 0.43 0.42 Pct. of land in farms: Regional mean 72 72 73 71 67 60 60 Pct. of land in farms: Summers County 75 80 83 84 83 81 80 Pct. of land in farms: Summers’ Z score 0.18 0.44 0.51 0.65 0.80 1.04 1.02 Source: University of Virginia, 2004. Table 2. Percent Decadal Change in Agriculture in Summers County and Appalachia, 1880–1940. 1880– 1890– 1900– 1910– 1920– 1930– 1890 1900 1910 1920 1930 1940 Total farms: regional mean 17 33 12 –2 –2 Total farms: Summers County 26 43 9 –2 –3 Total farms: Summers’ Z score 0.53 0.46 –0.08 –0.02 –0.05 17 13 –0.18 Mean farm size: regional mean –13 –22 –10 –3 –8 –12 Mean farm size: Summers County –16 –27 –7 1 0 –12 Mean farm size: Summers’ Z score –0.20 –0.46 0.22 0.36 0.61 –0.01 Source: University of Virginia, 2004. sure, Summers County ranked first among the 43 West Virginia counties in Appalachia and third among all 175 Appalachian counties whose boundaries remained unchanged during the study period (see Table 3). The territorial prevalence of agriculture and the trajectory of agricultural change in Summers County fits the region’s historical experience very closely. data The principle data for this research are the building locations shown on the two 1912 USGS maps. Figure 3 shows detail on the Big Bend map. The identification of agricultural neighborhoods is predicated upon the likelihood that nearly all unidentified buildings in rural areas are farmers’ residences. This assumption may be assessed with manuscript records from the 1910 U.S. Census. In 1910, census takers organized people’s responses by dwelling. Their ledgers record not only the people in each dwelling, but also whether the dwelling was a farm. Therefore, farm populations and farm houses are readily tabulated in absolute and relative terms from the manuscript census. 64 george towers Table 3. Summers County’s Agricultural Representativeness in Appalachia and West Virginia, 1880–1940. Summers County’s Summers County’s Category rank among 43 Summers County’s rank among 175 Appalachian average absolute Z Appalachian counties in West score, 1880–1940 counties Virginia Average farm size Farms per square kilometer Pct. of land in farms Pct. change in number of farms Pct. change in average farm size Average among categories 0.17 0.31 0.66 0.22 0.31 0.34 th 1st th 4th th 18th 4 th 2nd 8 th 1st rd 1st 6 20 80 3 Source: University of Virginia, 2004. The six magisterial districts served as census enumeration districts, inviting comparison between census figures and map counts. Incomplete overlap between the 1912 USGS maps and district boundaries limits evaluation to the Greenbrier, Green Sulphur, and Forest Hill districts. The 1912 USGS maps cover the entire Greenbrier district, 90 percent of Green Sulphur, and 83 percent of Forest Hill. Within the Greenbrier district, assessment is confined to the unincorporated land outside the Hinton and Avis city limits. The tenth of Green Sulphur shown on the Clintonville quad of 1921 was sparsely populated: of the 681 structures mapped in the district as a whole, 96 percent are on the 1912 Meadow Creek map. The onesixth of the Forest Hill district mapped in 1932 contains 15 percent of the district’s structures. Larger, more populated portions of the other three districts were mapped after 1912. Correspondence between unidentified mapped buildings and census dwellings is mediated by complications. There are two compelling reasons to expect that mapped buildings will outnumber dwellings. First, not all mapped buildings were occupied dwellings. For example, vacant houses were mapped but would not have been tallied by census takers. Second, undercount afflicted censuses of the late 1800s and early 1900s. An oftcited estimate of undercount in the 1910 Census is 6.5 percent (Robinson 1988). Whether undercount was higher in the countryside or in the city is debated by historians. Most suggest that rural areas, home to fewer transients and immigrants, were better reported (Parkerson 1991; King and Magnuson 1995). Others, however, draw a finer distinction, asserting that cities and rural areas were both underenumerated relative to small towns (Winkle 1991). As their supervisors warned, census takers could easily miss secluded rural homes set back from main roads (King and Magnuson 1995). Omissions of this type are documented where census manuscripts can be matched with the 1912 USGS maps. For example, Green Sulphur Figure 3. Detail view of the area around the Low Gap School on the 1912 Big Bend Quadrangle. 66 george towers Table 4. Dwelling Estimates Census Dwellings Mapped Dwellings (Adjusted for 6.5% Undercount) (Adjusted for 4.7% Vacancy) Forest Hill 310 304 Greenbrier 256 257 Green Sulphur 608 618 1,174 1,179 District Total Table 5. Farmhouse Estimates Census Farm Houses Estimated Mapped Farm Houses (Adjusted for 6.5% Undercount) (Adjusted for 4.7% Vacancy) Forest Hill 284 280 Greenbrier 201 219 Green Sulphur 522 516 1,007 1,014 District Total enumerator James E. Hensley made unusually detailed entries. He listed dwellings by roads named for creeks or mountains, features that enable identification of matching roads on the Meadow Creek map. While using the map to assign buildings to these roads is inherently imprecise, there were clearly more mapped buildings than enumerated dwellings along these roads (U.S. Census 1910). Adjusting for vacancy and undercount allows for comparison of census dwellings with mapped buildings. Housing vacancy rates were first recorded by the U.S. Census in 1940. These late rates are serviceable for 1910, however, because agricultural neighborhoods in West Virginia did not dissolve until after World War Two (Photiadis 1980). Indeed, the Great Depression had pushed many West Virginians back to subsistence farming (Armentrout 1941; Thomas 1998) and in Summers County the number of farms peaked in 1940 at 2,168. Given the crowded countryside, the 1940 vacancy rate of 4.7 percent for Summers County was probably relatively low and makes for conservative estimates of previous vacancy. The USGS maps show 1,237 unidentified rural structures in the three comparable districts. Applying the 4.7 percent vacancy rate produces an estimate of 1,179 occupied rural dwellings. The 1910 census counted 1,102 rural dwellings in the districts. As there is unresolved disagreement whether rural undercount was exceptional, I applied the general undercount estimate of 6.5 percent which increases the dwelling total to 1,174. Table 4 shows that the remarkable match between the adjusted total figures is replicated within each district. These comparisons may be extended to farmhouses. To estimate how many unidentified buildings were farm houses, I first excluded buildings clustered together in villages from my calculations. I then Rediscovering Rural Appalachian Communities multiplied the remainder by 92 percent, the proportion of farm dwellings recorded by the Census in the entirely rural Forest Hill, Jumping Branch, and Pipestem districts. Making adjustments for undercount and vacancy generates a close fit between census figures and map counts (see Table 5). Based on the results of these rough comparisons, census data supports the use of building locations on historic USGS maps to study rural settlement patterns. methodology and analysis HGIS methods allow for the analysis of historical settlement patterns with precise modern topographic data. For example, anthropologists and geographers use topographic HGIS to analyze site selection within settlement systems (Gragson and Bolstad 2007; Hunter 2009). Of particular relevance to the current research, archeologists employ HGIS to model past peoples’ movement across their landscapes with cost allocation analysis (Wheatley and Gillings 2002; Conolly and Lake 2006). Topographic cost allocation analysis determines the cost of travel given the relative impediments presented by slope, elevation, and aspect. Cost allocation analysis may suggest likely locations for prehistoric pathways (Bell and Lock 2000), or, it may assign least cost regions to destinations based on the incurrence of travel cost. In this case, destinations will be country schools and hamlet centers and the resulting least cost regions will be prospective agricultural neighborhoods. The application of topographic HGIS is predicated upon ethnographers’ shared conclusion that mountainous terrain exerted a powerful influence on community formation. 67 ‘‘Neighborhoods develop among people who have frequent and regular contacts, and in this region topography has helped to determine social relationships of the inhabitants and to form neighborhood groupings’’ (Brown 1988, p 52). While proximity brought people together, rugged land created boundaries between neighborhoods. Brown reports that as of 1942 all travel across the mountains was on foot or horseback. Consequently, steep slopes minimized contact between adjacent neighborhoods separated by ridges (Brown 1988). This generalization is supported by subsequent ethnography and oral history. For instance, Howell writes that Tennessee mountain neighborhoods ‘‘were defined largely by the drainage system’’ (2003, p 111); and, Martin notes that mountains ‘‘continued to separate’’ Hollybush Hollow from adjacent settlements throughout its history (1984, p 3). In Summers County, early 20th century cross-country transportation was equally primitive. For example, in 1906 the 30 mile trip between Hinton and Princeton, the seat of neighboring Mercer County, took 10 hours by horse and carriage. The county’s notoriously poor dirt roads were not substantially improved until the late 1940s (Cottle 1997). ESRI’s ArcGIS cost allocation analysis function recreates the defining role of topography on agricultural neighborhoods. The cost surface is represented by the 30 meter raster cells of the USGS National Elevation Dataset (NED). Each of the millions of cells in the NED is assigned an elevation which allows for the calculation of the slope across neighboring cells. Travel cost is a function of distance and the en- 68 george towers ergy cost of walking at slope. For each destination cell representing a country school or a hamlet center, a region is generated from the surrounding cells for which the travel cost to that destination is the least. Hypothetically, these least cost regions approximate the boundaries of the southern Appalachian agricultural neighborhoods of a hundred years ago. Hamlet centers and country schools are destinations. Hamlets offered essential commercial and social services to the surrounding countryside. For example, a typical early 20th Century hamlet of 100 people may have had a post office, a church, a grocery store, a feed store, a mill, and a school (Hart 1975). Assuming an average household size of five or six, a dozen or two houses would have complemented the handful of community and commercial buildings. Hamlets were named on the 1912 USGS maps. A central point within each of the study area’s 36 named hamlets was digitized with ArcGIS. Consistent with the premise that these were small service centers, 33 of the 36 hamlets had post offices in 1912 and two of the other three had post offices that were closed before 1912 (Helbock 2004). Within the study area, the cluster of some twenty structures at Green Sulphur Springs is a representative hamlet. Farm families from the surrounding area regularly traveled to Green Sulphur Springs to trade at the store and attend church (Newcomb 2008). Smaller hamlets in the study area also served as central places and shift size expectations downward. For example, True, which lay at the confluence of the Bluestone and New Rivers until it was flooded by the construction of the Bluestone Dam in 1948, was a tiny hamlet offering commercial services and river ac- cess to communities up Pipestem Creek and on adjacent Tallery Mountain. One hundred years ago, True ran a kilometer or two along the south bank of the Bluestone. Only five structures, however, including a mill, store and post office, formed the hamlet at the mouth of Pipestem Creek. Four additional structures, presumably farmhouses, were located along the floodplain in the True vicinity (Summers County Historical Society 1984; Sanders 1992). Similarly, Warford was a New River hamlet comprised of a four building cluster that included a country store, post office, and a blacksmith shop (Sanders 1992). Like True, a half dozen residences were scattered through the surrounding neighborhood. Country schools were community nodes for the agricultural neighborhoods that filled the countryside between hamlets. Agricultural neighborhoods, as discussed above, were small kinship-based communities. A representative example from the study area is the River Ridge neighborhood. River Ridge rises sharply between Pipestem Creek and the New River. The Lane, Keaton, Farley, and Pettrey families established a tightly knit neighborhood on the ridge in the early 1800s after valley land was taken. The population pressure representative of the region certainly applied to River Ridge: an early family of Lanes included 15 children and a late 19th century Keaton fathered a dozen with two wives. Community buildings, Ridge School (see Figure 4) and a log church, were built in the 1870s in a central location (Summers County Historical Society 1984; Sanders 1992). Peaking in the first decades of the last century, country schools were an expedient means of providing mandated public Rediscovering Rural Appalachian Communities 69 Figure 4. Ridge School. education to rural residents in the preauto era. In 1913, the year after the Big Bend and Meadow Creek maps were published, America’s 212,000 one room rural schools enrolled more than half of the nation’s schoolchildren (Gulliford 1996). In West Virginia, the number of schoolhouses more than doubled from 2,142 in 1880 to 4,819 in 1905 (Ambler 1951). In Summers County, the number of country schools grew from 16 in 1871 to 119 in 1890 to 160 by 1908 (Miller 1908). Country schools were loci of functional regions in two important ways. First, they were locally administered. Upon statehood in 1863, West Virginia established a highly localized hierarchy of country school administration. From the county scale, administrative space was divided among the magisterial districts which were in turn divided into school districts containing a single country school (Ferguson 1950; Ambler 1951; Trent 1960). In this way, each agricultural neighborhood was formally recognized as a functional region. Second, country schools were central locations for neighborhood activities and symbolized neighborhood identity. As the only public property belonging to the typical rural neighborhood, schools housed not only classes but also a variety of community events including elections and entertainment (Dunne 1977; Reynolds 1999). Neighboring focused on the school and the school came to symbolize the community (DeYoung and Lawrence 1995; Howell 2003). For example, James New 70 george towers comb, who attended Summers County’s Red Spring country school in the 1920s, vividly recalls the pie suppers and cakewalks that gathered his neighbors at the school on Friday evenings (Newcomb 2008). In short, ‘‘The schools housed the activities that joined people into a community, and the identity of rural communities became inextricably linked with their schools’’ (Gulliford 1996, p 35). Consistent with their function as community centers and in order to minimize children’s walk to class, schools were centrally located within agricultural neighborhoods. Anecdotally, Newcomb relates that the Red Spring School was sited so that community members were no more than a mile (1.6 km) walk from the school (2008). A 1929 West Virginia Department of Education survey of 29 rural school districts, including one in Summers County, found that in almost two-thirds of the districts, more than 70 percent of students lived within 1.6 kilometers of their school. Conversely, in 87 percent of the districts, less than 20 percent of students lived more than 2.4 kilometers from school (Holy 1929). Centrally located and regularly distributed, country school locations comprise a spatial catalog of functional nodes required for the GIS analysis. The 74 schools located in the countryside away from named places serve as potential neighborhood nodes. Country churches also organized agricultural neighborhoods. Ethnographers attest that neighborhood churches were an important element of community organization (Stephenson 1968; Photiadis 1980). In some neighborhoods, congregations met in schoolhouses (Brown 1988); in others, freestanding churches occupied central community locations (Beaver 1976). Twenty-five churches appear on the USGS maps in the study area. Of these, 22 are in hamlets or near a school and were not considered as unique nodes. Three churches, however, were alone amidst linear settlement patterns along streams and were included as destinations in the cost allocation analysis. Two assumptions derived from the above discussion support the importance given to walking at slope in constructing the cost surface. First, since schools were sited within walking distance of students, people regularly walked to these destinations. Second, following ethnographers’ reports, steep slopes bounded communities. The cost surface recreates neighborhood boundaries by attaching a high energy cost to traversing steep slopes on foot. From NED elevation data, I generated a raster layer measuring slope as the percentage of vertical rise over horizontal run. To create a walking energy cost surface from slope, I departed from archeologists’ convention of using a physics-based trigonometrical formula (Bell and Lock 2000) and instead borrowed the following experimentally-based equation from applied physiology that gauges the amount of energy expended by walking relative to the percentage of slope (Minetti et al. 2002): Cwi = 280.5i 5 – 58.7i 4 – 76.8i 3 + 51.9i 2 + 19.6i + 2.5 (1) where Cw is joule / (body weight in kilograms * meters traveled) and i is percent slope. As 4,184 joules equal one kilocalorie (kcal) and assuming an average body weight of 60 kilograms, the following equation converts this formula into a surface of caloric expenditure: Rediscovering Rural Appalachian Communities kcal = (Cwi * 60 * meters traveled) / 4184 (2) The cost surface was modified to allow the study area’s three swift rivers to bound neighborhoods. Sections of the Bluestone, Greenbrier, and New that were mapped as two dimensional features on the 1912 topographic maps were digitized as polygons interrupted at bridge locations. Assigning the rivers an insurmountably high value of 100,000 calories turned the river polygons into neighborhood barriers. Figures 3 and 5 show how the analysis converts topography to least cost zones. Figure 3 is the portion of the 1912 Big Bend Quadrangle immediately surrounding a node, Low Gap School. Figure 5 overlays the semi-transparent cost surface on the topographic map and shows the boundaries (in white) of the Low Gap School least cost zone. As darker shades indicate greater cost, Figure 5 shows how slopes impart higher travel cost and how zone boundaries tend to follow high-cost steep slopes. The overlay suggests that Low Gap School, at a low spot, or ‘‘gap’’ along Wolf Creek Mountain (the letters ‘‘WOLF CR’’ are splined to follow the ridgeline on the map), was the focal point for a ridgetop agricultural neighborhood. The cost allocation analysis generated an initial least cost rural zone around each of the 111 country schools, hamlet centers, and country churches that were digitized as point features. Of these, 32 were truncated by the study area boundaries and were removed from further analysis. Of the remaining 79, 24 are hamlets organized around hamlet centers; the 53 centered on country schools and the 2 based on churches are presumed to be agricultural neighborhoods. 71 results and discussion The following discussion of these zones’ spatial qualities is based on the preceding demonstration that agricultural neighborhoods in Summers County are representative of those throughout southern Appalachia. As presented above, anthropologists and sociologists found great commonality amongst agricultural neighborhoods across the region and local histories of communities like River Ridge match expectations from ethnography. The preceding census data analysis provides quantitative evidence that Summers County agriculture was emblematic of the region. With the establishment of the study area’s representativeness, the hypothesis that HGIS analysis reveals the boundaries of historic southern Appalachian agricultural neighborhoods may be examined. I compared the zones’ spatial characteristics—building counts, geographic size, and building density—with those reported for southern Appalachian agricultural neighborhoods by ethnographers and historical geographers. Ethnographers’ estimates of residences per southern Appalachian agricultural neighborhood range from 11 to 60 (Pearsall 1959; Matthews 1965; Stephenson 1968; Beaver 1976; Martin 1984). Mid-century ethnography, however, did not involve formal cartography and ethnographers made only passing notice of the spatiality of neighborhood settlement patterns. On the other hand, Wilhelm’s historical geography of Virginia’s Shenandoah National Park is unique for its attention to the spatial detail of southern Appalachian agricultural neighborhoods. Although Wilhelm’s work was in the Blue Ridge physiographic province instead of the Appalachian Plateau Figure 5. Cost surface analysis and least cost region around the Low Gap School. Rediscovering Rural Appalachian Communities province that dominates the study area, he is confident that ‘‘the geometric patterns of settlement, much more difficult to change [than other aspects of material culture], became prototypes for the rest of the Mountain South’’ (1978, p 206). His meticulous diagrams indicate that neighborhoods contained between 11 and 49 farms, closely corresponding to ethnographic reports (Wilhelm 1978). For example, Brown’s Beech Creek census of 164 people in 31 houses and Martin’s Hollybush Hollow count of 150 people living on 30 farms agree and are representative (Martin 1984; Brown 1988). The agricultural neighborhoods derived from the HGIS analysis performed here return representative building counts. The 55 agricultural neighborhoods averaged 17 structures and 43 neighborhoods (78 percent) were within Wilhelm’s range of 11 to 49 homes. The remaining twelve zones were smaller, containing 10 or fewer buildings. A twostructure zone that contained a single structure and a schoolhouse was reallocated to adjacent zones, leaving 54 agricultural neighborhoods. The other 11 small zones contained 4 and 9 farmhouses around a country school, enough for several dozen relatives to form a kinship neighborhood. Figure 6 displays the final 54 agricultural neighborhoods and 24 hamlets within the study area. The empty buffer inside the study area and partially surrounding these 78 shaded regions was originally occupied by the 32 least cost zones that crossed the study area boundaries. Figure 6 also serves a reference map showing the places named in the above discussion. The secondary literature suggests that there should be little difference between 73 agricultural neighborhoods and hamlets in absolute terms of structures and geographic size. Hamlets and their surrounding communities in the study area averaged 20 structures. Large hamlets, like Green Sulphur Springs, had around 50 structures within their vicinities and the smallest, like True, had a half dozen. Multiplication of Wilhelm’s range of 11 to 49 farmhouses per neighborhood by the average 1910 Appalachian farm size of 42 hectares, suggests that agricultural neighborhoods should have ranged in size from 462 to 2,058 hectares. By 1940, farms averaged 33 hectares, lowering the range to between 363 and 1,617 hectares. Generally consistent with these calculations, the 54 zones centered on schools and churches averaged 606 hectares of potential neighborhood area with a median of 579 hectares and ranged from 250 to 1,445 hectares. Unlike size measures, density ratios directly address the contrast between dispersed neighborhoods of farmsteads and clustered hamlets. Historical researchers concur that regardless of geomorphological setting, farms were dispersed within neighborhoods. In linear hollows, farm houses spread about 800 meters apart along streams (Wilhelm 1978; Brown 1988). In fan-shaped hollows, farmsteads at headwaters and stream confluences were 150 meters from one another. In coves, several dwellings clustered at the stream outlet and the rest were dispersed around the basin’s periphery. Ridge settlement was linear with about 150 meters separating farmers’ residences (Wilhelm 1978). These observations establish a range of 150 and 800 meters between farms. Calculating dispersion based on small Figure 6. Agricultural neighborhoods and hamlets derived by HGIS analysis. Rediscovering Rural Appalachian Communities and large Summers County farm sizes leads to nearly identical figures. From 1880 to 1930, about 90 percent of Summers County farms occupied between 8 and 202 hectares. Farmsteads centered within evenly dispersed very small eight hectare farms will be 160 meters apart; those on 202 hectare farms will be 802 meters apart. HGIS analysis converted these distances into four density categories. Two are non-agricultural—‘‘commercial’’ and ‘‘vacant’’—and two correspond to farming —‘‘general agricultural density’’ and ‘‘archetypal agricultural density.’’ Maximum density expectations for agriculture derive from a hypothetical area divided into very small farms of eight hectares. A 40 hectare search area centered on each raster cell accommodates 5 very small farms. Therefore, 6 or more structures within the search area suggest that land use is ‘‘commercial’’ and typical of a hamlet. The minimum agricultural density is that of an area exclusively occupied by very large 202 hectare farms with their farmsteads spaced 800 meters apart. Land further than 800 meters from a structure is not likely to be farmland and is classified as ‘‘vacant.’’ Only 2 percent of the study area, however, was this remote and two-thirds of the least cost zones did not contain any ‘‘vacant’’ land. I characterize land at ‘‘general agricultural’’ density levels as follows: the minimum density is a single house within 800 meters; the maximum density is five houses within the surrounding 40 hectares. A narrower density sub-category, ‘‘archetypal agricultural,’’ corresponds to a landscape of evenly spaced 40 hectare farms, the average farm size in the county from 1900 to 1930. Allowing for slightly 75 uneven spacing expands expectations by a farmstead on either side, or an ‘‘archetypal agricultural’’ density range from zero to two farms within the search area. Two expectations follow from the establishment of these density categories. First, an overwhelming majority of the land within zones assumed to be agricultural neighborhoods should be at ‘‘archetypal agricultural’’ densities. Second, even those zones assumed to be hamlets should be primarily farmland but should also contain a relatively greater minority share of ‘‘commercial’’ density. Remember that as in the cases of Green Sulphur Springs, True, and Warford described above, farms fringed hamlets’ tiny commercial cores, leading to the expectation that agricultural densities predominated within hamlet zones. Table 6 shows that in 61 percent of the agricultural neighborhoods, at least 90 percent of land is at ‘‘archetypal agricultural’’ densities. In more than 80 percent of these zones, as shown in Table 7, there is no ‘‘commercial’’ land. The second expectation finds support from Table 7 in that 58 percent of the hamlets contain ‘‘commercial’’ land. An equally meaningful measure is the density surrounding the set of house locations within each zone. This metric indicates whether houses are located in agricultural settings. The above density categories are adapted to this purpose by simply subtracting one—the house in question—from the number of houses within the search area. Therefore, houses situated amidst ‘‘archetypal agricultural’’ densities will have, at most, a single neighbor within the search area and houses in ‘‘commercial’’ settings will have five or more neighbors. The expectations for this measure are 76 george towers Table 6. Archetypal agricultural density. Percent of archetypal agricultural land 39–49% 50–79% 80–89% 90–95% 96–100% Agricultural neighborhoods, N, (%) Hamlets, N, (%) Total 0 7 14 20 13 54 (0%) (13%) (26%) (37%) (24%) (100%) 2 6 8 7 1 24 (8%) (25%) (33%) (29%) (4%) (100%) Table 7. Commercial density. Percent of commercial land 0% Agricultural neighborhoods, N, (%) Hamlets, N, (%) 1–5% 6–15% 16–44% Total 44 10 0 0 54 (81%) (19%) (0%) (0%) (100%) 10 6 7 1 24 (42%) (25%) (29%) (4%) (100%) straightforward: more houses in hamlets should be in ‘‘commercial’’ settings and more houses in agricultural neighborhoods should be in areas of ‘‘archetypal agricultural’’ density. These expectations are borne out by a variety of calculations. In agricultural neighborhoods, 2 of every 3 houses (596 of 903) are in ‘‘archetypal agricultural’’ settings and only 1 in 50 (15 of 903) are in ‘‘commercial’’ areas. Fortynine of 54 agricultural neighborhoods (91 percent) do not contain any houses in ‘‘commercial’’ areas. In and around hamlets, houses in ‘‘archetypal agricultural’’ settings fall to 41 percent of the total while those in ‘‘commercial’’ areas increase to 28 percent. Of the 793 houses in ‘‘archetypal agricultural’’ settings, three-fourths are in agricultural neighborhoods; of the 151 houses in ‘‘commercial’’ areas, nine-tenths are in hamlets. This consistency within zones in terms of size and density is a function of the even spacing of community nodes and the uniformity of farmhouse density. The regularly dispersed pattern of community nodes has less than a one percent likelihood of occurring randomly according to nearest neighbor analysis. Farmhouse density is also constant: 89 percent of the land in the 78 zones is at ‘‘archetypal agricultural’’ densities. Because both categories of point features are evenly spaced across the landscape, zones are certain to contain homogenous settlement patterns. The equivalencies between farmhouse density and zone sizes with those suggested by census data and ethnographic observations assure these patterns’ fidelity to expectations from secondary sources. In other words, the above analysis merely translates the organizational logic of agri- Rediscovering Rural Appalachian Communities cultural neighborhood settlement patterns into numerical terms. conclusions This research presents an HGIS methodology that reliably locates the Appalachian agricultural neighborhoods of a century ago. The least cost zones generated by HGIS analysis of cultural features recorded on early topographic maps share the spatial signature of the early 20th century southern Appalachian agricultural neighborhoods described by ethnographers and cultural geographers. The methodology presented here is also significant for its replicability. The primary data sources, geo-referenced historic topographic maps and modern topographic coverages, are freely downloadable for HGIS analysis. The principal analytical method, cost allocation analysis, is a standard, transparent GIS tool that requires only modest GIS proficiency. Supported by the regional representativeness of the study area, the method may be applied by scholars across the social and environmental sciences to reconstruct historic southern Appalachian rural social space and extend our understanding of the region’s historical geography and contemporary cultural landscape. For example, in archeology the inventorying of historic maps to analyze past settlement patterns is a fundamental HGIS application (Harris 2002; Armstrong et al. 2008). As students of the southern Appalachian countryside attest, once ubiquitous major landscape artifacts like the log cabins and company houses represented on old topographic maps are rapidly vanishing (Rehder 2004; DellaMea 2009). Archeolo- 77 gists may find this topographic HGIS method useful as they search for and integrate the remaining traces of material culture representing 19th century southern Appalachian society. Historical geographers might use this research method to explore Francaviglia’s observation that ‘‘one of the greatest visual contrasts in our culture occurs as one crosses the line from agriculture to mining’’ (1991, p 5). This passage resonates with Figure 7, which shows structures on a panel of contemporaneous USGS topographic maps. The coal camps around Winona and those strung between Gentry and Backus comprised the eastern flank of Fayette County’s New River coalfield and stand out from the surrounding farmlands. For southern Appalachia, the juxtaposition of these two landscapes is a dualism that defines the region’s history. The methodologies presented here invite inquiry not only into how the coalfieldcountryside boundary shifted over time and space, but also may inform questions about the complementarity of these settlement patterns. For historical sociologists and social geographers, topographic HGIS analysis might address the social and economic differences long observed between valley and ridge communities. Early on, environmental advantages found socioeconomic expression. Valleys offering access to water and good farmland were settled first and supported the region’s leading rural communities (Wilhelm 1978). Ridge communities, physiographically denied these amenities, were afflicted by the notorious southern Appalachian ‘‘culture of poverty’’ (Weller 1965; Gallaher 1974). Determining whether the topography of socioeco- Figure 7. Structures shown on the Big Bend, Meadow Creek, and Winona quadrangles. Rediscovering Rural Appalachian Communities nomic status persists or has been reversed with rural sprawl as suggested in recent Canadian research (Paquette and Domon 2001) will contribute to our understanding of contemporary Appalachia. Finally, for geographers, planners, and landscape ecologists studying ‘‘rural sprawl,’’ the low density settlement pattern that encircles many small towns and flanks rural roadways (Daniels 1999), topographic HGIS provides important context. Like metropolitan sprawl, rural sprawl is lamented for its encroachment on farmland and wilderness, its infrastructural demands, and its centrifugal effects on community (Daniels 1999; Reeder et al. 2001). While GIS-based assessment of sprawl’s costs is a burgeoning research area, it is typically made on the basis of relatively recent changes (Hasse and Lathrop 2003; Burchell et al. 2005; Wolman et al. 2005). Topographic HGIS puts recent landscape change in historical perspective, allowing for a richer assessment of rural sprawl’s environmental impact. The digitally driven ‘‘democratization of cartography’’ empowers diverse scholarship with GIS (Slocum et al. 2009). Beyond the reconstruction of southern Appalachian agricultural neighborhoods, the goal of this study is to invite others to put topographic HGIS to their research purposes. 79 Armentrout, W.W. 1941. The Low-Income Farm Situation in West Virginia As We Know It. In Bulletin 299: Proceedings of the 1940 Conference on Low-Income Farms, 12–16. Morgantown: Agricultural Experiment Station, West Virginia University. Armstrong, D.V., M.W. Hauser, D.W. Knight, and S. Lenik. 2008. Maps, Matricals, and Material Remains: An Archeological GIS of Late-Eighteenth-Century Historic Sites on St. John, Danish West Indies. In Archaeology and Geoinformatics: Case Studies from the Caribbean, ed. B.A. Reid, 99–126. Tuscaloosa: University of Alabama Press. Baker, A.R.H. 2003. Geography and History: Bridging the Divide. Cambridge: Cambridge University. Beaver, P.D. 1976. Symbols and Social Organization in an Appalachian Mountain Community. Ph.D. Dissertation, Department of Anthropology, Duke University. Bell, T., and G. Lock. 2000. Topographic and Cultural Influences on Walking the Ridgeway in Later Prehistoric Times. In Beyond the Map: Archaeology and Spatial Technologies, ed. G. Lock, 85–100. Amsterdam: IOS Press. Billings, D.B., and Blee, K.M. 2000. The Road to Poverty: The Making of Wealth and Hardship in Appalachia. Cambridge: Cambridge University Press. Brown, J.S. 1988. Beech Creek: A Study of a Kentucky Mountain Neighborhood. Berea, KY: Berea College Press. acknowledegment The author thanks the editors and anonymous reviewers for their very helpful suggestions. Burchell, R.W., Downs, A., McCann, B., and Mukherji, S. 2005. Sprawl Costs: Economic Impacts of Unchecked Development. Washington: Island Press. references Ambler, C.H. 1951. A History of Education in Colten, C.E., P.J. Hugill, T. Young, and K.M. Morin. 2005. Historical Geography. In West Virginia: From Early Colonial Times to Geography in America at the Dawn of the 21st 1949. Huntington, WV: Standard Printing Century and Publishing Company. 149–163. Oxford: Oxford University. , eds. G.L. Gaille and C.J. Willmott, 80 george towers Conolly, J., and LakeM. 2006. Geographical Francaviglia, R.V. 1991. Hard Places: Reading Information Systems in Archaeology. the Landscape of America’s Historic Mining Cambridge: Cambridge University. Districts. Iowa City: University of Iowa Cottle, R.K. 1997. Cemetery Siting in the Bluestone Reservation Area, Summers County, West Virginia: 1750–1997. M.S. Thesis, Press. Gallaher, A., Jr. 1974. The Community as a Setting for Change in Southern Appalachia. Department of Geography, Virginia In Appalachia: Its People, Heritage, and Polytechnic Institute and State University. Problems, ed. F.S. Riddel, 291–304. Cunfer, G. 2005. On the Great Plains: Agriculture and Environment. College Station: Texas A&M University Press. Daniels, T. 1999. What to Do About Rural Sprawl? Paper presented at the American Planning Association Conference, Seattle. Dawson, A., Donaldson, K., Elmes, A., and Dubuque, IA: Kendall/Hunt Publishing Company. Gragson, T.L., and Bolstad, P.V. 2007. A Local Analysis of Early-Eighteenth-Century Cherokee Settlement. Social Science History 31(3):435–468. Gregory, I.N., and Healey, R.G. 2007. Historical Gormont, J. 2007. WV Historical Geospatial GIS: Structuring, Mapping and Analysing Products. Morgantown: West Virginia GIS Geographies of the Past. Progress in Human Technical Center. DellaMea, C. 2009. Southern West Virginia Coalfields. Accessed 31 July 2009 at http:// www.coalcampusa.com/sowv/index.html. DeYoung, A.J., and Lawrence, B.K. 1995. On Geography 31(5):638–653. Gulliford, A. 1996. America’s Country Schools, 3rd Edition. Niwot, CO: University Press of Colorado. Harris, T.M. 2002. GIS in Archaeology. In Past Hoosiers, Yankees, and Mountaineers. Phi Time, Past Place: GIS for History, ed. A.K. Delta Kappan 77(2):104–112. Knowles, 131–142. Redlands, CA: ESRI Dunn, D. 1988. Cades Cove: The Life and Death of a Southern Appalachian Community, 1818–1937. Knoxville: University of Tennessee Press. Dunne, F. 1977. Choosing Smallness: An Press. Hart, J.F. 1975. The Look of the Land. Englewood Cliffs, NJ: Prentice-Hall, Inc. Hasse, J., and Lathrop, R.G. 2003. A HousingUnit-Level Approach to Characterizing Examination of the Small School Residential Sprawl. Photogrammetric Experience in Rural America. In Education Engineering & Remote Sensing 69(9):1021– in Rural America, ed. J.P, Sher, 81–124. Boulder, CO: Westview Press. Eller, R.D. 1982. Miners, Millhands, and Mountaineers: Industrialization of the Appalachian South, 1880–1930. Knoxville: University of Tennessee Press. ———. 2008. Uneven Ground: Appalachia since 1945. Lexington: University Press of Kentucky. Ferguson, L.M. 1950. The Educational Development of Mercer County. M.A. Thesis, Department of Education, Marshall College. 1030. Helbock, R.W. 2004. United States Post Offices, Volume VI—The Mid-Atlantic. Scappoose, OR: La Posta Publications. Holy, T.C. 1929. Survey of Education in West Virginia, Volume III: School Buildings. Charleston: West Virginia State Department of Education. Howell, B.J. 2003. Folklife along the Big South Fork of the Cumberland River. Knoxville: University of Tennessee Press. Hunter, R.W. 2009. People, Sheep, and Landscape Rediscovering Rural Appalachian Communities Change in Colonial Mexico: The Sixteenth- 81 Paquette, S., and Domon, G. 2001. Trends in Century Transformation of the Valle Del Rural Landscape Development and Mezquital. Ph.D. Dissertation, Department Sociodemographic Recomposition in of Geography and Anthropology, Louisiana Southern Quebec (Canada). Landscape and State University. Kaplan, B.H. 1971. Blue Ridge: An Appalachian Urban Planning 55:215–238. Parkerson, D.H. 1991. Comments on the Community in Transition. West Virginia Underenumeration of the U.S. Census, University Bulletin 71:(7/2). 1850-1880. Social Science History King, M.L., and Magnuson, D.L. 1995. Perspectives on Historical U.S. Census 15(4):509–515. Pearsall, M. 1959. Little Smoky Ridge: The Undercounts. Social Science History Natural History of a Southern Appalachian 19(4):455–466. Neighborhood. Birmingham: University of Knowles, A.K. 2008. GIS and History. In Placing History: How Maps, Spatial Data, and GIS Alabama Press. Photiadis, J.D. 1980. The Changing Rural Are Changing Historical Scholarship, eds. Appalachian Community and Low-Income A.K. Knowles and A. Hillier, 1–26. Family: Implications for Community Redlands, CA: ESRI Press. Development. West Virginia University Mann, R. 1995. Diversity in the Antebellum Appalachian South: Four Farm Communities Bulletin 80(9/7). Reeder, R., Brown, D., and McReynolds, K. in Tazewell County, Virginia. In Appalachia 2001. Rural Sprawl: Problems and Policies in the Making: The Mountain South in the in Eight Rural Counties. The Small City and Nineteenth Century, eds. M.B. Pudup, D.B. Regional Community: Proceedings of the Billings, and A.L. Waller, 132–162. Chapel Hill: University of North Carolina Press. Martin, C.E. 1984. Hollybush: Folk Building and Social Change in an Appalachian 2000 Conference 14:199–206. Rehder, J.B. 2004. Appalachian Folkways. Baltimore: Johns Hopkins University Press. Reynolds, D.R. 1999. There Goes the Community. Knoxville: University of Neighborhood: Rural School Consolidation at Tennessee Press. the Grass Roots in Early Twentieth-Century Matthews, E.M. 1965. Neighbor and Kin: Life in a Tennessee Ridge Community. Nashville: Vanderbilt University Press. Miller, J.H. 1908. History of Summers County, West Virginia from the Earliest Settlement to the Present Time. Hinton, WV: James H. Miller. Iowa. Iowa City: University of Iowa Press. Robinson, J.G. 1988. Perspectives on the Completeness of Coverage of Population in the United States Decennial Censuses. Paper presented at the Annual Meeting of the Population Association of America. Salstrom, P. 1994. Appalachia’s Path to Minetti, A.E., Moia, C., Roi, G.S., Susta, D., and Dependency: Rethinking a Region’s Economic Ferretti, G. 2002. Energy Cost of Walking History, 1730–1940. Lexington: University and Running at Extreme Uphill and Downhill Slopes. Journal of Applied Physiology 93:1039–1046. Newcomb, J.C. 2008. Interview with W.J. Daniels and author. Lynco, WV, December 30. Press of Kentucky. Sanders, W. 1992. A New River Heritage, Volume II. Parsons, WV: McClain Printing Company. Slocum, T.A., R.B. McMaster, F.C. Kessler, and H.H. Howard. 2009. Thematic Cartography 82 george towers and Visualization, 3 Edition. Upper Saddle Study/Historic Context Study. Washington: River, NJ: Pearson/Prentice Hall. National Park Service, U.S. Department of rd Stephenson, J.B. 1968. Shiloh: A Mountain Community. Lexington: University of Kentucky Press. Summers County Historical Society. 1984. The History of Summers County, West Virginia. Marceline, MO: Walsworth Press. Thomas, J.B. 1998. An Appalachian New Deal: West Virginia in the Great Depression. Lexington: University Press of Kentucky. Trent, W.W. 1960. Mountaineer Education: A Story of Education in West Virginia 1885– 1957. Charleston, WV: Jarrett Printing Company. United States Census. 1910. Thirteenth Census of the United States, 1910—Population, Microfilm Roll 1694 Pocahontas, Putnam and Summers Counties, West Virginia. Washington: U.S. Census. United States Census. 1913. Thirteenth Census the Interior. Weller, J.E. 1965. Yesterday’s People: Life in Contemporary Appalachia. Lexington: University of Kentucky Press. Wheatley, D., and Gillings, M. 2002. Spatial Technology and Archaeology: The Archaeological Applications of GIS. New York: CRC Press. Wilhelm, G., Jr. 1978. Folk Settlements in the Blue Ridge Mountains. Appalachian Journal 5(2):204–245. Williams, J.A. 2001. Appalachia: A History. Chapel Hill: University of North Carolina Press. Winkle, K. 1991. The U.S. Census as a Source in Political History. Social Science History 15(4):565–577. Wolman, H., Galster, G., Hanson, R., Ratcliffe, M., Furdell, K., and Sarzynski, A. of the United States: 1910—Volume 7: 2005. The Fundamental Challenge in Agriculture: Reports by States, Nebraska— Measuring Sprawl: Which Land Should Be Wyoming, Alaska, Hawaii, Puerto Rico. Considered? Professional Geographer Washington: U.S. Census. 57(1):94–105. University of Virginia, Geospatial and Statistical Data Center. 2004. Historical Census Browser. http://fisher.lib.virginia george towers is a Professor of Geography .edu/collections/stats/histcensus/index and Associate Academic Dean at Concord .html. University, Athens, WV 24712. Email: Unrau, H.D. 1996. New River Gorge National River, West Virginia: Special History [email protected]. His research interests involve the human geography of Appalachia.