A Space-Time Data Cube: Multi-Temporal Forest Structure Maps From Landsat And Lidar

In this study, we prototype the combination of samples of airborne LiDAR (LiDAR plots) and Landsat data to characterize the development of forest structure attributes through time. A nearest neighbor imputation model was developed using predictors generated from wall-to-wall Landsat best available pixel (BAP) composites and reference measurements of forest structure derived from LiDAR plots. The imputation model was then applied through time on a study area in Canada's boreal forest, resulting in forest structure maps with a 30 m resolution for the period 1984-2012. We characterize post-disturbance trends in these forest structural metrics following wildfire and harvest and offer insights on the large-area, temporally dense mapping opportunities offered by the synergistic use of samples of airborne LiDAR and Landsat BAP composites.