Mapping the landing-zone probability of cocaine-carrying boats in Guatemala and Costa Rica

Keywords: Boat wake, spatially explicit mapping, YOLOv4, illicit supply chain
Abstract Type: Paper Abstract

Abstract

Over the last two decades, illegal cocaine trafficking has seen a global geographical expansion and economic growth, especially in the U.S. and Europe. To meet the growing demands, more than 23,000 metric tons of cocaine have been trafficked through the Central American ‘transit zone’ between 2006 and 2018. Due to the large area of the transit zone and limited interdiction resources, only a small portion of cocaine-carrying boats are intercepted by law enforcement authorities. However, taking advantage of their small size and clandestine nature, most illegal boats successfully land at unknown locations throughout the transit zone. Here we estimate the probabilities that spatially explicit locations will be the landing zones for cocaine-carrying boats near the shores of Guatemala and Costa Rica. To overcome the challenge of missing small boats in satellite imagery, we detected "boat wakes" applying a Yolo4 machine learning model that provided high-accuracy detections. Detected boats were then filtered using AIS and Fishing Effort data to remove legal activity, leaving only the boats that are likely to be trafficking cocaine. Moreover, vessel speed is calculated using boat wake length and categorized into different types (go-fast, fishing, semi-submersible, pangas). Finally, spatially explicit landing-zone probability maps for the landing zones for cocaine-carrying boats are created. Our results demonstrated that high-probability landing zones were concentrated near remote coastal areas with low-population density presumably to take advantage of decreased seizure risk. We anticipate that our research is a starting point toward more spatially explicit and disaggregated cocaine supply chain mapping.