County-scale biomass map comparison: a case study for Sonoma, California

Abstract

ABSTRACTThe amount of carbon stored in forests affects a wide range of regional- to global-scale climate change processes. However, current maps often show large differences in carbon accounting. In this study, we present a framework to evaluate and compare multiple recent biomass maps at the county scale (4119 km2). We first compare the differences in the forest and non-forest areas at the pixel and county levels from multiple maps. Map-based estimates of county-level mean and total biomass are compared to the United States Forest Service (USFS) sample-based estimates. Comparison of raster-based biomass products shows differences in mean and total biomass at both pixel- and county-levels. Despite all the maps using USFS's plot data for model training, only the three active sensor derived products compare well to USFS's estimates of total biomass (within 10%), while the three passive sensor derived map products underestimated total biomass by as much as 47%. Our evaluation demonstrates that the biomass map generated using combined Light Detection and Ranging (LiDAR) and auxiliary data achieve accurate estimates at plot-level (R2 = 0.67; RMSE = 97.9 Mg.ha-1). This comparison study confirmed that missing direct height information either from active sensors tends to underestimate total biomass and mean biomass density at county-level.