GC004-0006
Forest Aboveground Biomass in the Southwestern U.S. from a MISR Multi-Angle Index, 2000 – 2015

Monday, 7 December 2020
Poster
Mark J Chopping, Montclair State University, Montclair, NJ, United States, Zhuosen Wang, University of Maryland, College Park, Earth System Science Interdisciplinary Center, College Park, MD, United States, Crystal Schaaf, University of Massachusetts Boston, School for the Environment, Boston, MA, United States, Michael A Bull, NASA Jet Propulsion Laboratory, Pasadena, CA, United States and Rocio Duchesne, University of Wisconsin-Whitewater, Geography & Geology, Whitewater, United States
Abstract:
Aboveground biomass (AGB, Mg ha-1) was mapped across the forests of the southwestern United States inter-annually from the start of the Terra mission (2000), using data from NASA’s Multiangle Imaging Spectro-Radiometer (MISR). The approach developed here relies on the observation that a multi-angle index (MAI) shows a loge relationship with AGB in the National Biomass and Carbon Dataset 2000 (NBCD 2000) that was derived from radar and other data sets. MISR Level 1B2 Terrain radiance scenes from May-June 2000 - 2015 were converted to surface bidirectional reflectance factors (BRFs) using MISR Toolkit routines and the MISR 1 km LAND product BRFs and the red band BRFs in all available cameras were used to invert the RossThinLiSparseReciprocal BRDF model to retrieve kernel weights, model-fitting RMSE, and Weights of Determination. Cloud-free composites were obtained for each year with min(RMSE) as the criterion and the composited kernel weights were used to generate synthetic MISR datasets in several geometries: mean MISR camera viewing and solar angles; MISR viewing angles in the solar principal plane (SPP); and in the SPP but including specular and hotspot viewing angles. The MAI for MISR cameras (DA/AA)/CF in the SPP was found to be optimal with respect to NBCD 2000 over 19 locations near Mt. Lindsey, Colorado in the range 20 - 190 Mg ha-1, with AGB = 89.157 x ln(MAI)-210.75 yielding R2=0.91 and RMSE=15.4 Mg ha-1 after screening for sites with extensive exposed rock (nadir NIR BRF>0.3). With spectral metrics R2 = 0.07 (nadir NIR BRFs), 0.32 (NDVI) and 0.37 (nadir Red BRFs). To account for the differences in forest types, soils, and topography across the region, the loge model was fitted to NBCD 2000 by adjusting a single coefficient, using the Generalized Reduced Gradient method with AGB = a x log(MAI). The resulting maps were compared with up-scaled ICESat GLAS-derived estimates for 2005; and with up-scaled NASA Carbon Monitoring System lidar-derived estimates for 2013/14. Good agreement was seen vs. lidar-derived AGB in 2005 (R2=0.75, RMSE=59.2 Mg ha-1) and 2013/14 (0.66, 82.8 Mg ha-1). The MISR-derived AGB maps were spatially consistent with the NBCD2000 and 2005 GLAS-based maps across the southwestern U.S.; moreover, trajectories were stable through time, tracking the 2005 and 2013/14 estimates well.