Scaling ICESat-2 Canopy Heights for Sub-continental Forest Height Mapping

Forests play a vital role in climate change mitigation, and accurate characterization of forest biophysical attributes such as canopy height is crucial to reducing uncertainties in carbon emissions. However, challenges on generating high resolution canopy height products at a regional and global scales still persist due to inadequacies in ground sampling. Relying on height estimates from the Ice, Cloud, and land Elevation Satellite (ICESat-2) spaceborne lidar mission, we demonstrate the production of a 30-m resolution sub-continental canopy height product in the Deep South from Texas to Virginia. To meet this research goal, we estimated canopy heights in 30 m segment lengths using Land and Vegetation product (ATL08) classified point clouds and built and evaluated regression models relating canopy height and ancillary variables. Our ancillary variables included monthly Landsat imagery and LANDFIRE (Landscape Fire and Resource Management Planning Tools Project (http://www.landfire.gov) datasets including existing vegetation height (EVH), canopy height (CH) and forest canopy cover (CC). We developed gradient-boosted regression models based on sample ICESat-2 granules by taking the estimated canopy height (98th percentile height) as the dependent variable, and Landsat surface reflectance and LANDFIRE variables as independent variables. Model performance varied (R2 = 0.44 - 0. 50, MAE = 2.61 - .2.80 m) when individual (per month) Landsat data and LANDFIRE data were used. Improved performance was observed when combined Landsat and LANDFIRE data were used (R2 = 0.70, MAE = 2.09 m). We applied the developed model to produce a gridded canopy height product over our study area, which agreed moderately (R2 = 0.46, MAE = 4.38 m) with independent airborne lidar derived canopy heights. Our product showed better agreement over the 2019 Global Land Analysis and Discovery GEDI Global Forest Canopy Height (Potapov et al. 2021) (R2 = 0.19 MAE = 5.83 m). Our integration of ICESat-2 data with other spatially complete datasets shows promise which is expected to support large-scale operational assessment of forest-related carbon emissions and other forest structure studies.