Variable Selection Strategies for Small-area Estimation Using FIA Plots and Remotely Sensed Data

Abstract

The USDA Forest Service’s Forest Inventory and Analysis (FIA) unit maintains a net- work of tens of thousands of georeferenced forest inventory plots distributed across the United States. Data collected on these plots include direct measure- ments of tree diameter and height and other variables. We present a technique by which FIA plot data and coregistered remotely sensed raster data were used to predict the basal area of deciduous trees at a spatial resolution of 30 m. Results varied, generally indicat- ing that culling putatively unrelated variables did not improve estimates over those obtained using all the potential variables in the model.