Don ’ t stop ‘ til you get enough-Sensitivity testing of Monte Carlo Iterations for Model Calibration

One of the powerful tools geospatial modeling uses is the Monte Carlo Method. However, little work has been done on measuring the optimal number of Monte Carlo iterations to be performed. In this work, we present the new utility of two metrics for deriving the number of Monte Carlo iterations needed to calibrate the CA-based SLEUTH Urban Growth Model. SLEUTH calibration is the process of choosing the best set of parameters to forecast urban growth into the future. SLEUTH calibration is performed on historical urban layers. The two metrics used are the OSM metric, which is the optimal combination of available SLEUTH metrics (like comparative size and dispersal of urban growth) and the MCAWSderived Diversity metric that accounts for the individual model run area in summarizing the Monte Carlo results. We applied these two metrics on the calibration of three different cities; Tampa, FL, Merced, CA, and the Ellwood region of Santa Barbara, CA. We found that for SLEUTH calibration of historical data sets, one would need between 10 and 25 Monte Carlo iteration for the optimal variation in the calibration process. We discuss the far-reaching consequences of discovering that “less is more” in terms of Monte Carlo iterations in urban modeling, Geocomputation, and beyond.