Landfill Suitability Mapping using Satellite Imagery and Machine Learning Techniques: A Case Study for Delhi-NCR, India

Abstract. The improper disposal of municipal waste is a worldwide problem that is dangerous to the environment. They are responsible for soil, water, and air pollution, resulting in ecological imbalance. Thus, effective landfill site selection reduces these adverse effects and supports sustainable urban development. However, it is difficult to identify suitable landfill sites, especially in fast-growing urban areas, because of the complexity of adjusting various environmental, infrastructural, and demographic factors. Moreover, the unavailability of comprehensive datasets and reliable predictive models worsens the problem, leading to frequent suboptimal site selection. This study overcomes these difficulties by combining satellite imagery and machine learning (ML) techniques to develop an integrated web-based framework for assessing landfill site suitability in the Delhi-NCR region. A comprehensive dataset comprising features such as spectral features captured using Landsat-9 satellite imagery, Digital Elevation Model (DEM), Land Use/Land Cover (LULC), proximity to roads, railways, rivers, industries, restricted zones, and settlements, along with Land Surface Temperature (LST), population density, Soil-Adjusted Vegetation Index (SAVI), Normalized Difference Vegetation Index (NDVI), and Modified Normalized Difference Water Index (MNDWI) was utilized for model development. The potential of ML techniques, such as random forest (RF) and extreme gradient boost (XGB), was assessed for landfill site classification. The experimental results show that RF outperforms XGB (F1-score:0.91, AUC:0.98). This study can help policymakers in sustainable waste management and provide a means for improved environmental sustainability with optimal landfill site selection in urbanizing areas.