Analysis of multispectral RPA images for vegetation mapping in riparian areas

Abstract

Riparian vegetation is important in ecological maintenance along river banks. Vegetation Indices (VIs) and Leaf Area Index (LAI) are indices that can be indirectly correlated with plant development and health. Therefore, this study aimed to test the effectiveness of using Remotely Piloted Aircraft (RPA) in acquiring multispectral images for creating VIs and indirectly obtaining LAI. Orthoimages were obtained of 3 sample areas along the Pardo River Hydrographic Basin (PRHB), which comprises 3 classes of vegetation, the forest at the source of the Pardo River, a grassland vegetation area and a forest area. The sampling areas were cut into 9 plots of 9 x 9 m and distributed across the orthoimages. An RPA of DJI's Phantom 4 Multispectral model was used to obtain multispectral images. From the orthoimages, VIs were generated, such as NDVI (Normalized Difference Vegetation Index) and SAVI (Soil Adjusted Vegetation Index), then the corresponding LAI was generated. The results show that simple linear correlation analyzes identified LAI as a regression-dependent variable, demonstrating a high significance with the independent variables NDVI and SAVI. It was possible to verify that the vegetation classes and their structural heterogeneities influenced the adjustments of the LAIs. It is concluded that the images obtained by multispectral RPA presented very high spectral, spatial and temporal resolution, being suitable for the management and constant monitoring of permanent preservation areas (PPA).