CMStatistics 2020: Start Registration
View Submission - CMStatistics
Title: Most likely pathways in the ocean Authors:  Adam Sykulski - Lancaster University (United Kingdom) [presenting]
Michael OMalley - Lancaster University (United Kingdom)
Abstract: Spatial statistics for ocean data is a rapidly growing area of research. We provide a methodology for estimating the most likely path taken by a particle between two fixed locations on the global ocean surface. Such pathways are useful for understanding ocean circulation in general, and the movement of ocean-borne objects such as plankton, plastic, oil, and debris. Our methodology is purely data-driven using data from GPS-tracked ocean buoys from the Global Drifter Program. We use this data to construct Markov transition matrices and apply Dijkstra's algorithm to find the most likely paths. The novelty is that we apply hexagonal tessellation of the ocean using Uber's H3 index (which we show is far superior to the standard practice of rectangular or lat-lon gridding). We provide techniques for measuring uncertainty by bootstrapping and applying rotations to the hexagonal grid. We also develop a novel method for calculating the estimated travel time associated with a most likely path, which provides useful maps of global ocean connectivity.