Semi-Automated Technique for Vegetation Analysis in Sentinel-2 Multi-Spectral remote sensing images using Python

Abstract

Satellite imagery provides a lot of information that can be analysed for a variety of objectives. The vegetation analysis in a region for a particular period of time to process a large amount of data is remaining as a bottleneck and involves a series of timeconsuming steps that delay the output. To overcome this challenge, this study aims to automate the process and estimate the area of sparse and dense vegetation of a certain area of interest and to assess the vegetation dynamics in this region by using the Sentinel-2 data. Python with its opensource libraries are utilized for downloading and processing the satellite data. Mandal (sub-district) level Mean NDVI, area under sparse and dense vegetation are estimated at the peak vegetative growth stage in Rabi season (February) from 2017 to 2020. In the assessment of satellite data, it was observed that the Godavari delta region has shown a decrease in the sparse vegetative area (11094 ha.) and an increase in dense vegetation area (3272 ha.) in 2018 as compared with 2019 assessment. However, in the Krishna delta region, it was observed that the sparse vegetation area was decreased (90600 ha) and an increased dense area (161915 ha.) in 2020 as compared with 2017. The process involves downloading Sentinel-2 data using SentinelHub and SentinelSat API, pre-processing and segmenting NDVI images to classify vegetation areas and the calculation of Mandal (sub-district) level statistics.