Spatial Estimation of Timber Production and Carbon in Harvested Wood Products Using Remote Sensing

Accurate estimation of the annual production of different kinds of timbers at different locations has many science and policy implications. For example, timber type information is needed for accurate estimation of the amount and life cycle of carbon stored in the harvested wood product (HWP) pool, and possible transport of carbon in wood products through trade. Several attempts have been made to estimate the carbon storage in the HWP, regardless which approach to use, information of the annual timber production are required. A statistic model has been developed to estimate the annual roundwood production at the county level. The inputs of the model includes forest disturbance area calculated using the VCT algorithm derived from the Landsat time series stack, a forest type map, and timber product output (TPO) data collected from wood processing mills by the USFS. The model is applied to North Carolina, a state with a large forestry sector and where harvesting and logging are a primary forest disturbance type. Ten-fold cross validation were done to the preliminary estimation for each type of HWP. The root mean square errors range between 13.6 and 31.5 for hardwood types; and between 1.3 and 55.6 for softwood types. The model is empirical as it depends on the local information on forest disturbance, forest types, and the amount of the roundwood output. However, the approach of the model can be used to apply to other areas with the local information provided. The result can be served as a starting point in spatial estimation of carbon storage in HWP.