Built-Up and Land Use Land Cover Mapping for Sustainable Planning in Himalayan Region by Using Machine Learning and Sentinel-2 Satellite Imagery

The Himalaya region, known for its ecological sensitivity, also faces recurring natural hazards such as earthquakes, landslides, sinking, glacier bursts, and flash floods, which necessitate the implementation of effective planning and management strategies to ensure sustainable development. This study focuses on Joshimath, a subdivision of Chamoli district in Uttarakhand, India, situated in the Greater Himalaya region, covering an area of 4563.52 km<sup xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">2</sup>, Our objective is to map built-up areas using Machine learning techniques and Sentinel-2 satellite multispectral imagery to support sustainable planning initiatives. Two machine learning models, Random Forest (RF) and Support Vector Machine (SVM), were employed in this study. RF achieved an overall accuracy of 92.62% with a kappa value of 0.911, while SVM achieved an accuracy of 90.557 with a kappa value of 0.886. Precision and recall metrics for the built-up class were calculated, with RF achieving 94% precision and 83.22% recall, and SVM achieving 93.48% precision and 82.24% recall. Additionally, the total area classified as built-up by RF was 236.96 km<sup xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">2</sup>, whereas SVM classified 198.67 km<sup xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">2</sup>. The results demonstrate the effectiveness of machine learning techniques, particularly RF and SVM, in accurately mapping built-up areas in the Himalaya region. This information is crucial for sustainable planning and management practices, enabling stakeholders to make informed decisions regarding land use and development while preserving the ecological integrity of the region.