Analyzing and mapping forest canopy heights variations across malaysia using spaceborne lidar and synergism of multiple remote sensing platforms

Abstract

Forest canopy height and its spatial variation are key determinants of forest ecosystem services, biodiversity and above ground biomass (AGB). Large scale canopy height models (CHMs) based on previous spaceborne light detection and ranging (LiDAR) suffered from previous instruments limitations and the lack of regional focus. Additionally, studies examining the relationship between canopy height and its environmental and climatic determinants were set at a dormant stage due to the scarcity of accurate measurements at scale. The global ecosystem dynamic and investigation (GEDI) spaceborne LiDAR offers unprecedented sampled observations of forest structure. Herein, this study first assesses GEDI metrics as a proxy of canopy height in Malaysia based on airborne LiDAR. Second, explores the most influential climatic and environmental determinants of canopy height. Third, develops a high resolution spatio-continuous canopy height model by synergizing GEDI with Landsat 8, Sentinel 2 and Sentinel 1 observations and the inclusion of climatic and environmental variables. In this regard, GEDI data were assessed in four different sites. Covariate and multivariate analysis and machine learning algorithms, namely generalized linear regression, random forest (RF) and extreme gradient boosting were used to determine the most influential variables and analyze the relationships among them. Lastly, the machine learning model with the highest performance was used to develop a CHM for the whole tropical forests of Malaysia. Water availability, annual mean temperature and elevation gradients were found to be the most influential determinants at large scale. Moreover, the canopy height relationship with water availability is observed to be less significant at altitudes higher than 1000 m. The final high resolution canopy height model was achieved using the RF algorithm and reported 0.94 for coefficient of determination (R2) and 3.65 m for root mean square error (RMSE). This study provides insights into the influence of environmental and climatic gradients on canopy heights at large scales and supports the development of ecosystem modeling and monitoring forest response to climate change. Furthermore, this study contributes toward a cost-effective framework for large scale forest monitoring and AGB estimation by facilitating the operationalization of space LiDAR-based technology, and hence, pushing towards achieving sustainable forest management practices.