Gravel mining from river channels is conducted in many countries around the world, yet ground-based monitoring of these activities requires considerable manpower and is not very effective. Therefore, innovative and effective approaches to monitoring gravel mining are urgently required. Deployed as a high spatial resolution sensor in a daily revisit orbit, Formosat-2 has proved to be an ideal satellite for site surveillance. Using one known event of gravel mining in the TsengWen River, Taiwan, between March 2004 and March 2005, we analysed nine Formosat-2 images taken during this period to summarize three characteristics of gravel mining in the images. Based on these characteristics, a standard procedure for processing Formosat-2 imagery to detect gravel mining is proposed. This procedure is validated against ground truthing collected by an unmanned helicopter flying at low altitude. The evolution of this gravel-mining event in the Tseng-Wen River is described by processing all historical Formosat-2 imagery using the proposed procedure. This standard procedure has been successfully incorporated into the Formosat-2 automatic image processing system and has been used to monitor gravel mining on a daily-basis.