COS 45-8 - Estimating albedo from UAVs: A novel method of obtaining versatile, small-scale albedo estimates

Tuesday, August 8, 2017: 10:30 AM
B114, Oregon Convention Center
Charlotte Levy, Ecology and Evolutionary Biology, Cornell University, Ithaca, NY

Estimations of small-scale differences in albedo across forests of different canopy structure, composition, and snow-layer condition are necessary to improve models of the global energy budget. Although widespread coarse scale measurements (via satellite) and select point measurements (via flux towers) are available, there is a need for small-scale, mobile albedo estimates. Here we present and validate a novel method of obtaining albedo estimates using unmanned aerial vehicles (UAVs). A custom quadcopter was used to estimate seasonal albedo of a deciduous hardwood forest in Tully, NY. Results were compared to local satellite estimates obtained from the MODIS satellite. Results were also referenced against tower and satellite albedo estimates from similar forested sites in the region (Durham, NH (DUR), Bartlett, NH (BAR), and Petersham, MA (PSH)).


Summer albedo at solar noon was estimated at 15% ± SE 0.5 by UAV for Tully, NY. Similar forested sites in the Northeast were estimated from flux towers as 12% ± SE 0.0 (DUR), 17% ± SE 0.0 (BAR), and 13% ± SE 0.0 (PSH). These site-level differences are likely due to differences in forest composition and canopy structure across the sites. Site to site variation was significant (chi-squared = 3045.3, p-value < 0.001), with the average deviation across sites amounting to 1.4%. We also compared each ground estimate to satellite estimates, and found that satellites detected a much smaller site to site difference (chi-squared = 76.369, p-value < 0.001) with the average deviation across sites amounting to 0.3%. In order to validate our method of using the quadcopter to estimate albedo over the forest canopy, we looked at the difference between UAV/tower estimates and satellite estimates at each site. We found an average deviation between ground and satellite albedo estimates of 0.4%, 0.9%, 0.9%, and 1.1% at TUL, DUR, BAR, and PSH respectively. Our results suggest that ground estimates, including estimates made by UAV, capture more site to site variation in albedo than do satellites, and that estimates made by UAV are at least as comparable to local satellite estimates as estimates made by fixed towers at other sites. Our continuing work using this method will examine seasonal and small-scale within site variation in forest albedo. We anticipate that UAVs may present an important tool for future estimation of albedo for global modeling.