Correlation Between Tree Diversity and Spectral Variables of Sentinel-2 Images: A Study of Langtang National Park, Nepal

Anisha Puri; Babatunde Adeniyi Osunmadewa; Ram Asheshwar Mandal; Dr. Elmar Csaplovics

Abstract

This study was objectively conducted to explore the tree diversity of Langtang National Park (LNP), Nepal and show the correlation between spectral variables and tree biodiversity. Sentinel-2 MSI of LNP in Nepal was acquired from https://scihub.copernicus.eu/. Total 60 samples were collected from forest. Pearson correlation was established to show relationship between tree species diversity and reflectance values. Bagging LASSO algorithm was used to validate the model. Result shows that Shannon-Weiner index was ranging from 2.9 to 3 and it was the highest in dense forest. Altogether, there was 18 tree species in LNP. The highest importance value index was 79.64 of Rhoderndron arborum and it was the least around 3.44 of Prunus cornuta. Classified map of tree species of LNP showed high Kappa statistics with 0.7. The NIR band B6, B7, B8 and B8A performed high separability for tree species. High reflectance values were recorded of Betula alnoides, Qurecus semicarpofolia and Castonopsis hystrix but it was the lowest of Querus lamellose. The higher reflectance value was recorded of Juniperus indica and Sorbus cuspidate in case of shortwave-infrared spectrum (SWIR, Band 11) and SWIR2 (Band 12). The highest negative correlation was recorded between Simpson’s index and B7 with -0.386 and it was the highest positive correlation between Shannon-Weiner Index (H’) and B8 of Sentinel-2 with 0.361. The highest correlation was found between Simpson index and B8A_Entrophy with -0.365. Bagging LASSO model performed better prediction to show correlation between biophysical characteristics and reflectance value of remotely sensed data.