Remote Sensing-Derived Spectral Vegetation Indices and Forest Carbon:Testing the Validity of Models in Mountainous Terrain Covered with High Biodiversity

Abstract

Sequestration of carbon through forests is an important aspect in global climate change mitigation. Assessment of carbon in forests using remote sensing and GIS tools is one of the most important aspects of rapid and verifiable methodologies. A number of studies have shown the utility of spectral (vegetation) indices like NDVI in the assessment of forest carbon. However, there are limitations to this approach. The mountainous topography and high biodiversity affect the spectral values in pixels in multiple ways. The present article aims to test the validity of use of vegetation indices in high-biodiversity forests in mountains by modelling the ground based forest carbon measurement with vegetation indices of NDVI, EVI, SAVI and MSAVI in a multi-sensor, multi-season data environment with multiple regression methods like linear, power, logarithmic, polynomial and exponential. It is found that all the regressions have a poor coefficient of determination not even exceeding 0.2. It is concluded that the remote sensing-based spectral vegetation indices alone cannot be a proxy for forest carbon calculators in high biodiversity mountain forests.