The basic concept of a GIS is that one creates a computerized "layer cake" of spatial information, where each layer represents a single, spatial attribute such as roads, streams, soils, elevation, etc. Information in the form of points (such as archaeological sites), lines (such as roads or steams), or polygons (such as a soils or geology map) can be entered into the system through various means. These individual data layers (from whatever sources) are all georeferenced, meaning that they are all entered and stored in the system with the same coordinate system, such as UTM (Universal Transverse Mercator) or latitude/longitude. Once the various "map layers" are entered into the system, one can conduct a wide variety of display, measurement, and analysis operations on the data in ways that are much harder, if not impossible, using traditional methods and paper maps. Useful analysis such as area and distance measurements, coincidence tabulations between layers, and extensive spatial statistical functions and modeling of various layers can be conducted routinely. Theoretical models can be tested quantitatively. Hard copy output in the form of traditional paper maps at various scales can easily be produced consisting of various information. Such maps can represent a single layer, or have various line or point data overlayed (such as roads, streams, UTM coordinates, or archaeological sites). Data can also be entered directly from GPS receivers in the field. A primary benefit of the GIS approach is that the data are available for a variety of research users, and the database is cumulative over time.
GIS layer cake
GRASS GIS Database Development and Analysis
The GIS and image processing system used for the remote sensing and GIS analysis is the GRASS GIS. GRASS, the Geographic Resources Analysis Support System, is a general purpose GIS system that was developed by the U.S. Army Corps of Engineers Construction Engineering Research Laboratory (USA-CERL) in Champaign, IL. It was originally developed for environmental and land use planning at U.S. Army bases, including conducting cultural resources and other environmental land-use analysis. This powerful, integrated raster/vector GIS and image processing software resides in the public domain so that it is freely available. It is available with source code, so that enhancements and alterations can be easily made to it. A new GRASS development and support center that now has a Windows version of GRASS is located at Baylor University.
The Project GIS Data Base
The GIS database covers an area of about 35 by 60 km, covering the majority of the Arroux River Valley and its immediate environs. The current basic raster and vector layers of the GIS data base include: elevation (generated from the French digital elevation data and manually digitized 1:25,000 contours) aspect (derived from the digital elevation data) slope (derived from the digital elevation data) SPOT images (20 meter false color infrared) SPOT images (10 meter panchromatic) land use/land cover maps (derived from spot image data) geology (generated from 1:80,000 geology map of the upper 2/3 part of the region) faults (from the same 1:80,000 geology map) hydrology (from the three 1:50,000 topo maps and 1:25,000 maps) modern roads (from the three 1:50,000 and 1:25,000 topo maps) ancient roads (from project information and old maps) known Celtic hillforts (from project information and old maps), archaeological sites and field survey transects (from project surveys and other sources), and more.
Additional derived data layers showing different distance categories, or buffer zones, from: roads, streams, faults, archaeological sites, hillforts, and ancient roads were then generated from the data above. Additional data have recently been added that were derived from the 1:25,000 maps, including reclassifications and distance measurements from sites, hillforts, ancient roads, and hydrology. In all there are currently over 100 point, vector, and raster data layers in the database. These data are used to conduct a variety of analyses, including Line-of-sight modeling of the hillforts and the development of predictive models of archaeological sites of different periods.
.GIS Data Examples
At below left is the Digital Elevation Model (DEM) that was created by manually digitizing contours from a 1:25,000 map. These contours were then interpolated into a 20 meter raster array. The white areas are the highest elevation, with Mt. Buevray in the upper left. The hillforts are the white areas along the river valley which is the dark area in the center. In the middle is the aspect, or direction of slope face map. On the right is a slope map, showing the steepness of the slope. These were both derived from the DEM. Flat areas are yellow and orange, steep slopes are pink and blue. The river bottomlands are clearly visible in yellow.
Elevation ................Aspect ................Slope......... ..... ....... Geology (L) Archaeology sites and survey transects (R)
Numerous derived data layers have also been created. These are created by processing basic data layers with operations such as distance zones, etc. The following three images are distance to ancient roads (on left), distance to hydrology (center), and distance to the hillforts (right). The field survey segments are visible at left, along with archaeological sites from different periods.
Distance zones to ancient roads,.hydrology, and hillforts .........
One of the more interesting results on the GIS analysis was the line-of-site analysis (Madry and Rakos 1996). This GIS technique allows you to determine what is visible from any given location, and demonstrated that the old Celtic road network connecting the hillforts of the area tended to follow within the line-of-sight of the hillforts, rather than take more direct paths (as originally proposed in Madry and Crumley 1990). We have run the same line-of-site analysis from four locations on each of the known Celtic hillforts in the research area. These four line-of-site maps for each hillfort were then combined to generate a map of complete inter-visibility from each entire hillfort, assuming that the forts were manned by watchers from each of the four "corners" of the ramparts. An eye height of 5 meters above the terrain was used, assuming that towers of just over 3 meters height were strategically located around the ramparts (and that the eye level was just under 2 meters above that height). These individual line-of-site maps were then combined to produce a map which shows the total portion of the research area and each ground transect that is within line-of-site of each hillfort, and for the total network of hillforts in the region. This analysis shows clearly that the Celtic roads definitely tend to follow paths that stay within the view of the hillforts, even if they are less direct or require a steeper climb.
Additional work was done to model the location of Celtic and Gallo-Roman roads where we do not know the exact location of segments. Environmental and cultural variables were analyzed along the routes of various road segments and models developed to replicate and then extend the known segment to their destinations.
Line of sight analysis
Site location modeling is a useful GIS product. It allows us to model where archaeological sites of a given period may be located, based on the known site locations and various environmental and cultural data in the GIS. Statistical analyses were run on the archaeological sites of different periods with most layers in the GIS to look for pattern. Various predictive models were developed using the site data generated from field surveys conducted in 1978 and 1979. Models based on environmental and cultural data were created that accounted for 78.9% (45 of 57) of all Gallo-Roman sites in only 29.2% of the total area that was field surveyed. The same model also included 69.2% (36 of 52) of the Iron Age sites and 80.3% (49 of 61) of the Medieval period sites located in the field survey. This model was then generalized to include a much larger area surrounding the transects (4.7 times as large). New layers in the GIS containing these locations were produced and new maps showing the areas with the highest probability of site locations were created. These areas of higher probability of archaeological sites have a high correlation with areas that are threatened by current gravel mining activities in the area. Current research involves analysis of aerial photography and site surveys using GPS in these areas predicted to have higher site potential. We also are continuing to seek additional sites to increase the site sample (which is small) to create improved models.
Predictive model overlaid on SPOT satellite image
The Roman era sites tend to cluster along the rivers in flat bottomlands, and also along the Roman roads, which also tend to follow the lower elevations. Celtic sites tend to be more in uplands and near the Celtic hillforts on the hilltops. The largest single area of high probability of Gallo-Roman sites in this model is the area shown below, next to the river, where the Roman villa was located and later destroyed. The gravel operations continue to work in the area, and researchers are trying to locate additional sites using these techniques before they are destroyed.