Estimating Forest Floor Carbon Content in the United States

The USDA Forest Service Forest Inventory and Analysis (FIA) program conducts an annual forest inventory which includes measurements of forest floor and soil carbon content. Samples are collected on a systematic nation-wide array of approximately 7,800 plots where each one may represent up to 38,850 ha. Between 10 and 20 percent of these plots are measured on a recurring basis, and soil sampling includes measurements of both the forest floor and mineral soil (0-10 and 10-20 cm). In the United States, the current method of reporting for C stocks to international parties includes mathematical models of forest floor and mineral soil C. Forest type maps are combined with STATSGO soil survey data to generate soil C storage by forest types, but STATSGO possesses known shortcomings, particularly with respect to forest C estimation. STATSGO data are based largely on agricultural soils, so the data consistently underestimate C storage in forest floors. FIA's national-scale inventory data represent an opportunity to significantly improve our modeling and reporting capabilities because data are directly linked to forest cover and other geospatial information. Also, the FIA survey is unique in that sampling is not predicated on land use (e.g., hardwood versus softwoods, old-growth stand versus reverted agriculture) or soil type, so it is an equal probability sample of all forested soils. Given these qualities, FIA's field-observations should be used to evaluate these estimates if not replace them. Here we combined forest floor measurements with other forest inventory observations to impute forest floor C storage across the United States using nonparametric k-nearest neighbor techniques; resampling methods were used to generate estimates of uncertainty. Other predictors of forest floor formation (e.g., climate, topography, and landscape position) will be used to impute these values to satellite pixels for mapping. The end result is an estimate of landscape-level forest floor C stocks from plot-level observations. Future work will include comparisons of these imputed results with simpler models currently used by the US greenhouse gas inventory.