The Dependence of Tornado Activity on Terrain: Applying Bayes Law to the Historical Record in Arkansas

Recent research suggests that terrain variability may influence tornado activity, though separating this effect from population density biases is difficult to do in observations. Here a simple conditional probability model is derived purely from data using Bayes Law to calculate the joint dependence of tornado density on both population density and terrain variability in the vicinity of Arkansas for the period 1950-2015. Log-link linear regression is used to quantify the magnitude of the dependence of tornado activity on each predictor. The results show that, conditioned on population density, tornado density decreases 10% for every 10 meters of increase in terrain standard deviation. This result is found to be consistent across averaging area and across 20-year time periods. The research method is useful for developing a general tornado probability model which accounts for both population density and terrain variability.