Challenges exist in monitoring artisanal and small-scale mining (ASM) activities, given their dynamic and often informal nature. ASM takes form through various techniques and scales, including riverine dredging, which often targets the abundant alluvial gold deposits in South America. Remote sensing offers a solution to improve data collection, regulation, and monitoring of the more mobile and elusive ASM activities and their impacts. Mapping ASM riverine dredges using Synthetic Aperture Radar (SAR) is one of the application areas least explored. Three semi-automated detection approaches using Sentinel-1 SAR are compared on their ability to identify dredges with minimal false positives. The methods are: (i) Search for Unidentified Maritime Objects (SUMO), an established method for large ocean ship detection; and two techniques specifically developed for riverine environments that are introduced in this paper: (ii) a local detection method; and (iii) a global threshold method. A visual interpretation of SAR data with the inclusion of optical high-resolution data are used to generate a validation dataset. Results show it is possible to semi-automatically detect riverine dredge using SAR and that a local detection method provides the best balance between sensitivity and precision and has the lowest risk of error. Future improvements may consider further automation, more discriminatory variables, and analyzing the methods in different environments and at higher spatial resolutions.
Full Text Finder