Remote Sensing and Machine Learning Fusion: A Robust Framework for Land Use and Land Cover Change Detection

Determining land use and land cover (LULC) changes in an area is crucial before proceeding with any developmental or restoration plan. Traditionally, field surveys were done to mark significant changes in an area, often compared with toposheets of the region. However, this task was time-consuming and required a substantial human workforce and resources. With the advent of satellite sensor technology, the task of field surveys was minimised by a considerable percentage. The satellite imageries, capable of providing high-resolution and historical data, were analysed for land use/cover changes. The results of this study show a reduction in water bodies by a percentage of nearly 41% in a span of 28 years. Rapid industrialisation and illegal residency have contributed to an increase of 33% in built-up. Barren/fallow land shows a growth of nearly 79%, agricultural practices remain dominant in the region, whereas forest decreases 4%. Despite the time-saving capabilities, satellite imagery requires substantial processing. With the advancements in computer science and Geographical Information Systems (GIS), classification techniques came into the image, and LULC studies gave further refined and near-accurate results. The machine learning techniques, i.e. Maximum Likelihood Classification (MLC), have been used in this study to recognise its benefits in the timely analysis of a problem statement of LULC and will help review and monitor the study area.