Deep Learning Framework for Analysing Land Use and Land Cover Classification and Change Detection

Land Use (LU) refers to the way humans utilize land. Land Cover (LC) describes visible features on the land surface. Land use includes artificial surfaces and agricultural areas, while land cover includes forests, semi-natural areas, wetlands, and water bodies. Land Use and Land Cover (LULC) can be studied using remote sensing techniques, particularly satellite imagery, due to its broad geographic coverage and high temporal resolution. However, LULC classification and change detection are challenging due to the limited availability of high-quality labelled satellite data and noise and variability in the data. This research proposes a deep learning framework to classify the type of land use and land cover by using deep learning techniques enabling the monitoring of land-use changes over time to ensure sustainable land resource management in support of sustainable management of the ecosystems (United Nations Sustainable Development Goal (SDG) 15) and supporting urban planning (SDG 11). The proposed framework combines two key components: the LULC Classification Model and the LULC change Detection Model. The LULC Classification Model is implemented using ResNet50 architecture and it is trained on the Sentinel-2 MultiSpectral Instrument (MSI) - Level-1C remotely satellite imagery with a maximum of 5% cloud cover. The input dataset comprises 17112 images, which are classified into five different classes such as artificial surfaces, agricultural areas, forest and semi-natural areas, wetlands, and waterbodies. Results demonstrate that the LULC Classification Model achieved an accuracy of 92.38%, a precision of 92.41%, and a kappa statistic of 0.91. The LULC Change Detection Model is implemented by subtracting the area covered by each land use and land cover class between the two study years – 2018 and 2024, allowing for the identification and quantification of changes in land use and land cover over time. This research study found that agricultural land covered 74.67% in 2018 and increased to 75.27% in 2024, while artificial surfaces grew by 1.24% as forest areas declined by 3.20% in the area of application. This research aims to aid urban planners, enabling them to make informed decisions regarding sustainable land resource management.