K-Nearest Neighbor Imputation for Estimating Ponderosa Pine
Forest Structure and Wood Supply in Northern Arizona
Using Forest Inventory and Analysis (FIA) and Landsat TM Data

Over the last century, ponderosa pine forests in northern Arizona have moved toward conditions conducive to severe wildfire behavior. USDA Forest Service Southwestern Region objectives are to restore fire adapted ecosystems via tree thinning and prescribed burning. An analysis was undertaken to estimate wood supply from thinning byproducts over a 2.4 million acre area. Forest Inventory and Analysis plots, digital elevation and Landsat TM data were used to derive forest structural layers using k-nearest neighbor imputation with regression trees. A comparison of imputed wood volume to values observed from plots resulted in an r2=0.53. Integrating imputed forest structural variables including basal area, stand density index and canopy cover dramatically improved imputed volumes (r2=0.83). Wood volume was estimated at 4.5 billion ft3 for the analysis area. An integrated landscape assessment indicated that 0.85 to 1.02 billion ft3 of thinning byproducts are potentially obtainable to aid forest management planning and economic decisions.