Estimation of annual harvested wood products based on remote sensing and TPO survey data

Abstract

ABSTRACT Accurate estimation of Timber Product Output (TPO) is important for carbon budget accounting, since wood products can act as delayed release carbon pools. However, the existing timber harvest data in the US relies on Forest Service’s TPO survey, and the survey does not happen every year. In this study, we proposed a methodological framework to produce annual TPO volume estimates for seven southeastern states (North Carolina, South Carolina, Alabama, Florida, Georgia, Mississippi, and Tennessee) of the US by integrating TPO survey data, Landsat Time Series Stacks (LTSS), and National Land Cover Database (NLCD). First, a forest disturbance product was derived based on Vegetation Change Tracker (VCT) algorithm using LTSS from 1985 to 2016. Then, by linking the predictor variables derived from the disturbance data and the TPO survey data, two regression algorithms were tested and compared, and Random Forest was selected to create TPO estimation models for different wood types. The results show that from 1986 to 2015, the region produced more than 5 × 109 m3 wood products, including 3.7 × 109 m3 softwood products and 1.6 × 109 m3 hardwood products. The derived TPO estimates had large spatial variations among the counties within each state as well as large temporal variations across the study period. The TPO data derived through this study can provide an observational basis for calculating the amount of C transferred from the standing biomass to the wood products through logging and for partitioning of harvested C among different wood product pools.