Forest and Shrub Canopy Structure from Multiangle and High Resolution Passive Remote Sensing

Abstract:

The 3-D structure of forest and shrub canopies can be mapped using diverse technologies, with the most advanced being lidar and interferometric radar. Other approaches include various modes of interpretation of multi-angle imagery, high-resolution stereo photogrammetry, plant identification, delineation, and measurement from high-resolution panchromatic imagery, and image texture metrics. While active remote sensing will revolutionize mapping of canopy structure, there are currently limitations. High precision lidar will remain limited geographically until the launch of NASA's innovative Global Ecosystem Dynamics Investigation to the International Space Station in 2019 but even this mission will not see high latitude boreal forest, taiga, or shrubs in tundra because of the orbit. Radar-based methods must be calibrated using high quality data. Imagery from passive imagers acquired at a range of scales therefore has much value if it can be used to provide structure data at broader geographic and temporal scales. Here we report on canopy mapping at scales from 0.5 m to 250 m using high-resolution panchromatic imagery from satellite imagers and NASA’s Multiangle Imaging Spectro-Radiometer (MISR), respectively. MISR-based 250 m aboveground biomass maps for the southwestern U.S. were assessed against the radar-derived North American Carbon Program National Biomass and Carbon Dataset 2000, showing good agreement (R²=0.80, RMSE=31 Mg ha^-1 for the validation data set; and 0.76 and 18 Mg ha^-1, respectively, for 1013 random points). For Oregon forests the best and worst cases were R²=0.90, RMSE=42 Mg ha^-1 and R²=0.78, RMSE=62 Mg ha^-1, respectively. For improved validation, the CANAPI algorithm was used to interpret high-resolution panchromatic imagery. In Sierra National forest, California, canopy cover estimates agreed well with those from field inventory (R²=0.92, RMSE=0.03). Height estimates gave R²=0.94 and relative RMSE=0.25 m for the range 3 m - 60 m, vs. lidar estimates. In Alaskan Arctic tundra, CANAPI estimates of shrub parameters for 26 sites of 250 m × 250 m yielded R²=0.83 for fractional cover, R²=0.81 for mean crown radius, and R²=0.54 for total number of shrubs, vs. field measurements. CANAPI also allows the generation of canopy maps as well calibration/validation data, with some limitations.