A Comprehensive Forest Biomass Dataset for the USA Allows Customized Validation of Remotely Sensed Biomass Estimates.

There are several new and imminent space-based sensors intended to support mapping of forest structure and biomass. These instruments, along with advancing cloud-based mapping platforms, will soon contribute to a proliferation of biomass maps. One means of differentiating the quality of different maps and estimation strategies will be comparison of results against independent field-based estimates at various scales. The Forest Inventory and Analysis Program of the US Forest Service (FIA) maintains a designed sample of uniformly measured field plots across the conterminous United States. This paper reports production of a map of statistical estimates of mean biomass, created at approximately the finest scale (64,000-hectare hexagons) allowed by FIA's sample density. This map may be useful for assessing the accuracy of future remotely sensed biomass estimates. Equally important, fine-scale mapping of FIA estimates highlights several ways in which field- and remote sensing-based methods must be aligned to ensure comparability. For example, the biomass in standing dead trees, which may or may not be included in biomass estimates, represents a source of potential discrepancy that FIA shows to be particularly important in the Western US. Likewise, alternative allometric equations (which link measurable tree dimensions such as diameter to difficult-to-measure variables like biomass) strongly impact biomass estimates in ways that can vary over short distances. Potential mismatch in the conditions counted as forests also varies greatly over space. Field-to-map comparisons will ideally minimize these sources of uncertainty by adopting common allometry, carbon pools, and forest definitions. Our national hexagon-level benchmark estimates, provided in Supplementary Files, therefore addresses multiple pools and allometric approaches independently, while providing explicit forest area and uncertainty information. This range of information is intended to allow scientists to minimize potential discrepancies in support of unambiguous validation. [ABSTRACT FROM AUTHOR]