Methods

The goal of this project was to learn more about the placement of Roman temporary camps in Scotland, as a proxy for trying to understand something about how the Romans might have decided these locations. I particularly wanted to examine how the actual placement of these camps would compare to what the military maxims and guidelines of the day say these placements should be like.

Overall, the purpose of my research can be summed up in two questions.The first, which I consider my primary research question, is: At what distance are Roman marching camps typically located from land features, such as roads, water sources, and elevation, that were considered important by sources from the time? I also have a more qualitative secondary research question, which I have attempted to address as well: Do the locations of Roman marching camps tend to follow the maxims and guidance of military manuals from the time?

My first steps involved importing all of the necessary data (as described in the Tools and Data section) into ArcGIS. All told, this went easier than I expected, as I initially thought that I would need to convert much of the data into a single projection and coordinate system. However, most of my data was plotted using the British National Grid Reference system, and the rest were integrated automatically by the ArcGIS software. I did not immediately trust that ArcGIS had done a good job with the integration, so I did my best to ensure that my data had been placed correctly, but I encountered no significant problems of this type.

The first import of data I did in my final project file was the Roman camp locations, which had to be transformed from a CSV file into an Esri shape file to use in ArcGIS. After trimming out some camps (which I described in the Tools and Data section), I used ArcGIS’s Add XY Data feature to import the data and then created a shape file from that data.

At this point, I was still learning about how to use and analyze the elevation portion of my data, so for the time being I focused on roads and rivers. After layering river and road data over my base map of Great Britain, a striking detail I noticed immediately was the vast disparity between the number of Roman roads and the number of waterways, particularly in Scotland.This might make sense if we consider that Rome’s grip on Scotland was never all that firm1 — one might expect a more complex road network if the Romans had fully conquered and controlled their portions of Scotland for longer periods of time.

Early Screenshot of my map, with only Scotland(yellow), roads(black), and waterways(blue) shown.

After editing the color scheme and line widths for the imported data to be more easily understood, I decided to create some buffers of various sizes around the roads and water features to study the relative placement of the camps in a visual manner. I noticed that many of the camps fell within the buffers around roads, but even more fell within the buffers around water features. In order to analyze this data bit more statistically, I used ArcGIS’s Select by Location tools to select all of the camps within various distances of each feature, effectively counting all the camps within each buffer for me. I recorded these results in a table, which I will discuss further in the Analysis section.

A table of the number of camps falling with various ranges of roads and water sources, and the percentages relative to the total number of camps (179).

Next, I needed to import the digital elevation model to study the relationship between the camp locations and elevation. This data was stored in TIFF files, which is a raster data type (meaning the data is stored as pixels or cells), whereas the rest of my data up to this point had been vector data (stored as vertices that make up points, lines, or polygons)2. Vector and raster data can be hard to use in concert because of their differences, but using the Extract Values to Points tool, I was able to find the approximate3 elevation of each temporary camp and append that data to the camp’s row in the ArcGIS attribute table. This was somewhat complicated by the fact that the SRTM data was divided into pieces, and I had some issues with trying to combine them, so a small subset of the camps at the southernmost end of Scotland would have to be analyzed separately in this instance.

Screenshot of the map with the SRTM elevation data as a background rather than the political boundaries of Scotland. Camps are shown as pink dots, and elevation goes from light blue at low elevation to reds and whites and high heights.

Once I had that information, I decided to use ArcGIS’s Statistics tool to determine at which percentile of relative elevation the lowest, highest, and average camp were located within the DEM piece that they were found4. I did this by finding the lowest, highest and average relative elevations for both groups of camps, and then constructed a ratio of the number of unique locations at or below each height to the total number of unique locations in the DEM piece I was studying. I also recorded these results in a table, which will also be examined in my analysis.

A shortened version of the table containing percentiles for the lowest, average, and highest elevations at which a Roman camp has been found.

I also wanted to see whether these camps were on high elevations compared to their immediate surroundings rather than compared to the whole of Scotland. I was not able to find a good way to analyze this statistically, but I was able to create contour lines from the SRTM elevation data using the Create Contour Lines tool so that I could inspect some of the campsites visually. These contour lines tend to blot out the rest of the features of the map, so they are not included in most of my maps, but some screenshots showing them are available in the Static Maps section of the site.

Notes

  1. Maxwell, Gordon S., The Romans in Scotland (Edinburgh: J. Thin, 1989, Print): 26-37; see also Breeze, David J, Roman Frontiers in Britain (London: Bristol Classical Press, 2007, Print).
  2. See this StackExchange Forum post or this article on the British Ordnance Survey website for a more in-depth explanation of vector and raster data.
  3. I say approximate for several reasons — firstly, the SRTM elevation data is only accurate to 90m squares, and secondly, some values would have needed to be interpolated based on the elevations in the pixels around it.
  4. I would have liked to also do these calculations for the median camp, as the average could be affected by outliers, but unfortunately the Statistics tool does not report the median — *** MAY BE CHANGED LATER ***.