Identification of alluvial gold mining in the Colombian Choco region using radar satellite images

Abstract

Alluvial gold mining is a growing activity in Colombia that severely impacts the loss of vegetation cover with high environmental value. The detection of this activity is essential to define strategies focused on reducing this impact. In previous studies that have used optical satellite imagery, continuous cloud cover, especially in tropical regions, has been a limiting factor that can be overcome thanks to the ability of radar sensors to capture information regardless of whether or illumination conditions. In this project, time series analysis, textures metrics and supervised classification were implemented to identify areas affected by alluvial gold mining in the department of Choco. The approach involves two main steps: In the first, Sentinel-1 radar images with dual polarisation (HH and HV) and acquisition date between November 2020 and November 2021 were analysed, and in the second step, a TanDEM-X image with single polarisation (HH) was used. To evaluate the accuracy, classification results are compared with polygons identified in the region in 2021. The overall accuracy obtained was 97% with Sentinel-1 images and 98% with TanDEM-X. These results suggest that images used are suitable for identifying affected areas by alluvial gold mining. However, the discrimination of this affectation from the bare soil generated by other activities requires the inclusion of additional spatial criteria like the presence of previous affectation, beneficiation ponds and elevation data.