Building a Web Map, part 4

Well, the most difficult and time consuming part of the project is done; the DEM is finished. It's not perfect, but the original data wasn't perfect either. Here is the finished product, followed by an explanation.

The first thing I did was I identify points where I knew the exact elevation. There were just a few of these such as mountain peaks. For others I had relative elevations, such as the keep on the borderlands. the contour lines were marked as change of 25ft in elevation, but I had no base elevation. At other places I had general descriptions or average elevation. Also there were the maps from B10, Night's Dark Terror which had ridge lines of mountains depicted. This was useful for knowing the direction of slopes and adding more details to those areas.

In order to determine which method of interpolation was best, I did some testing with the keep on the borderlands map, which can be seen in a previous post. Interpolation is a mathematical method of predicting what the values of a certain geographical point will be based on the known values of surrounding points. I ended up working with the Topo to Raster tool which uses the ANUDEM method developed by the Australian National University Fenner School of Environment and Society. It is essentially a modification of the standard spline interpolation method, which can be described as taking a rubber sheet and spreading it over a given area and making certain it passes though specific points. The ANUDEM method is unique in that it allows for the input of more than just points; you can include rivers, cliffs, contour lines, coastlines, hydrological sinks, and lakes.

I first created contour lines, cliffs, and rivers. Then to create elevation points I created a single point for each hex from various maps, whether they were 6mph, 3mph, 2mph, or 1/2 mph. then I selected different hexagons within a certain area and generated random values within a certain range based on descriptions in the text. For the mountain ranges shown in B10, I detailed the values a little bit more meticulously. Though this may sound fairly easy and simple, it was extremely time consuming and took a lot of fine tuning to get just right. In the end there were over 8000 input points! I often had to go back and run test interpolations to catch anomalies and fix those points to better match the description of the text. Below is a screenshot of all my input data. From the final result I created a hillshade which I then placed over the DEM as a transparent layer and got what you see above. I will upload it to the web version shortly.