VIRS based detection in combination with machine learning for mapping soil pollution

Widespread soil contamination threatens living standards and weakens global efforts towards the Sustainable Development Goals (SDGs). Detailed mapping is needed to guide effective countermeasures sustainable remediation operations. Here, we review visible infrared reflectance spectroscopy (VIRS) based detection methods in combination with machine learning. To date, proximal, airborne spaceborne carrier devices have been employed for detection, allowing large areas be covered at low cost minimal secondary environmental impact. In this way, contaminants can monitored remotely, either directly or through correlation components (e.g. Fe-oxides, organic matter, clay minerals). Observed vegetation spectra has also proven an indicator pollution. Calibration models on learning are used interpret spectral data predict levels. The algorithms include partial least squares regression, neural networks, random forest. processes underlying each of these approaches outlined review. Finally, current challenges future research directions explored discussed.