Residential Development Potential for Wyoming

This dataset is a predictive model of Wyoming residential development potential. The model represents where new housing units are most likely to be built, with lowest development potential represented by values of 0 and highest development potential values represented by 1.To predict housing development into the future, we modeled the change in housing unit density over the past 20 years (1990-2010, US Census data) within R, using Random Forests and spatially-explicit predictor variables previously related to residential development patterns in Wyoming, including services, transportation, natural amenities, and past residential and oil and gas development. A binary model was selected as the best-fitting model and was used to predict a probabilistic output. Model validation was performed using out-of-bag (OOB) testing techniques to produce standard error statistics including Cohen?s kappa, OOB error, and class error. Within the Random Forest models independent boot-strapping (OOB) subsets with many thousands of iterations are used for model validation and each tree is constructed using a different bootstrap sample from the original data.Cohen?s kappa was 0.75, OOB error was 11.7% and model accuracy was 84.9% and 90.5%, respectively, for observed positives (change) or observed negatives (no change) correctly predicted. Additionally, we applied the Boyce index to test the model against residential structures constructed in Wyoming between 2010-2012, using binned versions of the models. Results indicated a highly significant model (Boyce Index = 0.976; P Further modeling details are provided in Copeland et al. (In Review).Citation: Copeland, HE, A Pocewicz, DE Naugle, T Griffiths, D Keinath, J Evans, J Platt (In Review). Measuring the effectiveness of conservation: A novel framework to quantify the benefits of sage-grouse conservation policy and easements in Wyoming. PLoS ONE