TROPICAL DRY FOREST DEGRADATION ESTIMATION AT LOCAL SCALE WITH UAV IMAGES

摘要

Forest degradation is a dynamic process, and its accurate mapping and detection has been limited by the lack of spatial and temporal resolution of conventional remote sensing, especially in tropical dry forests (TDF). The objective of this work is to assess UAV images for mapping and quantifying forest degradation of the TDF at local scale. Firstly, the accuracy of UAV images to estimate forest attributes (canopy height, canopy cover, biomass and frequency of individuals) is evaluated. These attributes are then integrated to estimate the status of forest degradation. UAV images were obtained for both rainy and dry season, also field measurements at 22 plots. UAV images were processed by photogrammetry of motion structure, and a canopy height model (CHM) and a mosaic in RGB are created. The CHM calculates canopy height, in combination with the RGB mosaic, the canopy cover was delimitated through an object-based image analysis. For the estimation of biomass and frequency of individuals, multiple linear regression models are developed, which allows the attributes data from field to be related to the height and canopy coverage estimated by UAV images. Forest degradation states are estimated using a relative degradation index. The preliminary results show that the processing of the UAV images has obtained a good accuracy for the average and maximum canopy height with an error of 0.4 - 3.1 m, respectively. The delimitation of the canopy cover has an overall accuracy of 95%. Forest attributes from UAV images are expected to continue to be calculated reliably, compared to those at the ground level.