The Effect on Atmospheric and Topographic Corrections on the Vegetation Density Mapping using Several Vegetation Indices: A Case Study in Arjuno-Welirang Volcanoes, East Java

This study aimed to evaluate the effects of atmospheric and topographic corrections on the vegetation density estimates based on vegetation index transformation. The research was conducted in Arjuno-Welirang volcanic complex, East Java, using Landsat 8 OLI imagery at 30 m spatial resolution. The image was corrected at two levels, i.e., atmospheric correction to at-surface reflectance using FLAASH method, and topographic correction using SCS-C method. The topographic correction referred to ALOS PALSAR DSM data, which was resampled at 30 m pixel size. The vegetation indices used includes NDVI, SAVI, ARVI, EVI and MSARVI. Fieldwork for measuring vegetation density was carried out by vertical bottom-up photography of the canopy on each sample, supported by observations of vegetation density using high-spatial resolution Google Earth imagery. The results showed that --in comparison with the atmospheric correction-- the topographic correction was able to increase the correlation coefficients between the spectral information and the measured vegetation density in the field, especially for SAVI, EVI and MSARVI transformations. On the other hand, the NDVI and ARVI showed slight decreases. Based on the vegetation density maps generated using regression equations, the SAVI, EVI and MSARVI showed slight increases from atmospheric to topographic corrections, while the NDVI and ARVI showed declines. The rugged terrain condition affected the accuracies of the models due to the difficulty of vegetation density measurement in the field and even distribution of the samples.

Automated summary

This study aimed to evaluate the effects of atmospheric and topographic corrections on vegetation density estimates using Landsat 8 OLI imagery in the Arjuno-Welirang volcanic complex in East Java. The results showed that the topographic correction improved the correlation coefficients between spectral information and measured vegetation density in the field, particularly for SAVI, EVI, and MSARVI transformations, while NDVI and ARVI showed slight decreases.