Assessing the Synergistic Use of Sentinel-1, Sentinel-2, and LiDAR Data for Forest Type and Species Classification

Abstract

The design of effective forest management strategies requires the precise characterization of forested areas. Currently, different remote sensing technologies can be used for forest mapping, with optical sensors being the most common. The objective of this study was to evaluate the synergistic use of Sentinel-1, Sentinel-2, and LiDAR data for classifying forest types and species. With this aim, a case study was conducted using random forest, considering three classification levels of increasing complexity. The classifications incorporated Sentinel-1 and Sentinel-2 monthly composites, along with LiDAR metrics and topographic variables. The results showed that the combination of Sentinel-2 monthly composites, LiDAR, and topographic variables obtained the highest overall accuracies (0.90 for level 1, 0.80 for level 2, and 0.79 for level 3). The most important variables were identified as Sentinel-2 red-edge and NIR bands from June, July, and August, along with height-related LiDAR and topographic variables. Although not as precise as Sentinel-2 at the species level, Sentinel-1 enabled the classification of broad forest types with remarkable accuracy (0.80), especially when combined with LiDAR data (0.83). Altogether, the results of this study demonstrate the potential of combining data from different Earth observation technologies to enhance the mapping of forest types and species.