A Comparison of Machine Learning and Geostatistical Approaches for Mapping Forest Canopy Height over the Southeastern US Using ICESat-2

The availability of canopy height information in the Ice, Cloud, and Land Elevation Satellite-2’s (ICESat-2’s) land vegetation product, or ATL08, presents opportunities for developing full-coverage products over broad spatial scales. primary goal this study was to develop a 30-meter map southeastern US, Southeastern Plains ecoregion Middle Atlantic Coastal ecoregion. More specifically, work served compare well-known modeling approaches upscaling from ATL08 wall-to-wall product. Focusing on only strong beams nighttime acquisitions, h_canopy parameter extracted data. Landsat-8 bands derived indices (normalized difference index, enhanced modified soil-adjusted index) along with National Cover Database’s cover digital elevation models were used extrapolate ICESat-2 tracks regional level. Two different techniques, random forest (RF) regression kriging (RK), applied estimating height. RF model estimated coefficient determination (R2) value 0.48, root-mean-square error (RMSE) 4.58 m, mean absolute (MAE) 3.47 bias 0.23 independent validation, an R2 0.38, RMSE 6.39 MAE 5.04 −1.39 when compared airborne lidar-derived heights. RK heights 0.69, 3.49 2.61 0.03 R 0.68, 0.47, 5.96m, 4.52 −1.81 results suggest feasibility implementation method larger extent potential combining other remote sensing satellite data future monitoring dynamics.