GIS 5990 Module 4: Predictive Modeling

Weighted overlay depicting a predictive model for site locations.

Module 4, stretching over two weeks, required us to progress through the multi-step process of creating a predictive model for archaeological site location probability. A predictive model combines known site locations with variables (generally environmental) to create a classified raster categorizing the probability of finding sites in different regions. Common variables to consider are elevation, slope, aspect, and distance to water sources; these were the variables used for our project. In order to create data layers that can be used in the Weighted Overlay tool to produce the predictive model, the required data must be located, acquired, clipped, merged, reclassifed, and/or converted. As an example, we must decide how many classifications we will have in our weighted overlay layer; we used three for our assignment. Thus, once we extracted an aspect layer from the DEM using the Aspect tool, we then used the Reclassify tool to create a new aspect raster with only three classifications. We had to similarly edit and create the slope, elevation, and water source distance layers for use in the Weighted Overlay tool.

This assignment, although lengthy, was an enjoyable practice of something I have been doing in my internship with the BLM. Something both this project and my internship have driven home is the need for a good project plan and good data for any predictive model. The sheer number of variables and different potential classifications and weightings for those variables, as well as varying responses to those variables by different human populations in different times and environments, requires careful consideration if one has any hope of producing a useful predictive model.