WyGISC Friday Forum Series: Predicting forest stand attributes from remote sensing data—to use polygons, plots or pixels? Andrew T. Hudak USFS Rocky Mountain Research Station, Moscow, Idaho
In this case study, Oregon Department of Forestry (ODF) managers of the Tillamook District in northwestern Oregon have LiDAR collections that cover a combined 75% of the study area as well as multitemporal Landsat data and 10m DEMs. Our objective was to use metrics derived from these spatially continuous data sources as predictor variables to impute forest structure attributes, summarized from stand exams at both the stand and stand subplot levels. We developed imputation models to predict multiple stand structure attributes across the study area. Equivalence tests revealed no significant biases between imputed stands and reference stands, or between imputed pixels aggregated to the stand boundaries, and reference stands. This study demonstrates that stand-level reference data may be imputed to pixel-level target cells and then aggregated to the stand boundaries, providing reliable estimates of stand attributes.