TITLE:
Estimating Tree Canopy Cover and Identifying Deforestation Patterns in Meghalaya (1990-2021) through ML Classifiers
AUTHORS:
Tapan Dhar, Sengalvarayan Aravindan, Basudeb Bhatta
KEYWORDS:
Meghalaya, Forest Cover, ML, SVM, Python Script, Decadal Change, Biennial Change
JOURNAL NAME:
Journal of Geographic Information System,
Vol.15 No.3,
June
20,
2023
ABSTRACT: This study utilizes ML classifiers to estimate canopy density based on three decades of data (1990-2021). The Support Vector Machine (SVM) classifier outperformed other classifiers, such as Random Tree and Maximum Likelihood. Satellite data from Landsat and Sentinel 2 was classified using a developed python model, providing an economical and time-saving approach. The accuracy of the classification was evaluated through a confusion matrix and area computation. The findings indicate a negative trend in the overall decadal change, with significant tree loss attributed to jhum cultivation, mining, and quarry activities. However, positive changes were observed in recent years due to the ban on illegal mining. The study highlights the dynamic nature of tree cover and emphasizes the need for biennial assessments using at least five time-series data. Micro-level analysis in Shallang, West Khasi hills, revealed a concerning trend of shortening jhum cycles. Automation in canopy change analysis is crucial for effective forest monitoring, providing timely information for law enforcement proposals and involving forest managers, stakeholders, and watchdog organizations.