Distribution of Aboveground Biomass and Productivity of US Forests from Fusion of Multi-scale Remote Sensing

Comprehensive assessment of the North American Carbon budget requires a more precise and spatially resolved distribution of carbon stored in the aboveground live biomass (AGLB) of forests and their productivity or sequestration capacity (AGP). This distribution will improve the large uncertainty in estimates of AGLB in US forests, which currently range from 13 to 38 Pg C, will provide more precise estimates of the annual rate of C sequestration in temperate forests in North America, and will help to quantify the size and location of the missing terrestrial C sink. Using biomass estimates from 3.41 × 105 plots that represent 9.94 × 106 trees inventoried as part of the US Forest Services' Forest Inventory and Analysis (FIA) program, we construct a wall-to-wall map of forest biomass and productivity in the conterminous US. Environmental variables that control interannual variability in C uptake and loss in US forests act at different spatial scales. For example, soil moisture varies at a fine spatial scale and wind-throw events result in C sources in small patches in mature forest. In contrast, temperature, precipitation, and radiation that controls photosynthesis by incident light vary at coarser scales. However, previous efforts to model the distribution of AGLB in US forests have considered only a single spatial scale. The objective of this analysis is to generalize previous AGLB models by formulating a multi-scale regression model called a hierarchical linear model (HLM). In the HLM, the response variable is biomass at the 250 m resolution and the predictor variables are remote sensing data acquired by MODIS, QSCAT, SRTM, Radarsat, etc. capturing the landscape and seasonal variability of vegetation structure, phenology, and moisture at different spatial resolutions. We develop the model using AGLB and AGP from FIA data as both the training and test set over the US. The results are assessed both in terms of their magnitude and spatial scale and are discussed based on their regional and ecosystem significance.