A NEURAL NETWORK APPROACH TO UNDERSTAND FOREST RESTORATION PRIORITIZATION IN COLORADO

Goal: Our primary goal is to identify the key input factors and their relevance (or relative importance) in influencing the current forest restoration prioritization decisions in Colorado at different spatial scales. In case there are different methods to map certain input factor (e.g. different methods to identify WUI), we will also evaluate the relevance of those methods.Objectives:Build NNMs that allow us to take different combinations of input maps, conduct sensitivity analysis, and report factors that have significant influenceon forest restoration decisions at different locations and spatial scales.Test the hypothesis that data, models, and assumptions may have significant influence on the real-world restoration treatment decisions in Colorado, at least at certain spatial scales.Test the hypothesis that maps of observatory (or raw, un-synthesized) data such as digital elevation map, roads, streams and reservoirs, vegetation, fire history, building locations, etc. may have higher impacts to Colorado's forest restoration practice than synthesized maps produced through simulation models.Test the hypothesis that synthesized data from certain prioritization models may have been better received by restoration decision makers than the other models.Test the hypothesis that social and economic conditions (i.e. population density, treatment costs, accessibility, ownership, incomes, education etc.) may have significant impact to restoration decisions.Propose future data collection effort and identify future research and analytical directions that would be more relevant in supporting the real-world forest restoration practices.