Classifying and Examining Deforestation Patterns and its Environmental Implications

Abstract

Deforestation has surged significantly in recent years, driven by factors such as agricultural expansion, livestock grazing, mining, industrial activities, construction, transportation infrastructure, and forest fires. Forests cover approximately 31% of the Earth’s surface, providing essential ecosystem services including oxygen production and carbon dioxide (CO2) sequestration, while also housing nearly 80% of terrestrial biodiversity. Maharashtra, the second most industrialized state in India, faces severe environmental and socioeconomic impacts from deforestation. This research aims to classify the 36 districts of Maharashtra into deforested and non-deforested areas. The study is bifurcated into two parts:

In the first part, satellite imagery datasets are processed using the Siamese Algorithm, a deep learning model optimized for analyzing paired images to detect changes and patterns. In the second part, multiple geospatial factors—such as distance from roads, construction activities, forest fire occurrences, elevation, and river erosion—are considered. The count of green pixels within the imagery is converted into numerical values and, alongside the geospatial data, input into the AdaBoost algorithm. AdaBoost, a robust machine learning classifier, optimizes these inputs to enhance classification accuracy.

By integrating both image-based and geospatial data, this approach offers a comprehensive assessment of deforestation patterns. The results are displayed through an interactive web-based application featuring maps, charts, and graphs for effective visualization. The dataset comprises Sentinel-2 satellite images spanning six years (2017-2022), capturing critical geographical features. The Siamese Neural Network exhibits high validation accuracy (96.15%), and the AdaBoost algorithm demonstrates exceptional classification performance with an accuracy of 97.82%. This study not only advances the methodology for deforestation detection but also provides valuable insights for sustainable forest management in Maharashtra, aligning with the United Nations’ Sustainable Development Goals.