Supplementary Material

(Access here)

Figure 1: Example of ubiquitous buoy positions. The two red points correspond to two distinct positions provided by the same buoy at the same timestamp (October 10, 2016 at 00:33).Figure 2: Example of an isolated buoy position. The red points correspond to the position of a buoy separated from its closest neighbors by an inconsistent distance (the speed required to achieve this distance is far greater than the speeds of both tuna purse seine vessels and ocean currents).Figure 3: Land positions. The white points corresponds to sample of buoy positions detected on land using a 0.05° buffered shoreline data from the GSHHG database (Global Self-consistent, Hierarchical, High-resolution Geography; Wessel and Smith, 1996). The red dashed line represents the buffer zone around the shoreline.Figure 4: Boxplots of on-board and at-sea buoy speeds (in knots) from the training data used to build the random forest classification algorithm. The training data consisted of location data provided by buoys equipped with sensors that automatically detect their immersion in seawater.Figure 5: Schematic description of the kinetic classification algorithm (KiC). (A) The green points represent the different positions with undetermined status, recorded along a buoy trajectory. The length of the black arrows roughly reflects the value of the speed associated with the position. (B) The red points correspond to buoy positions classified as “on board” after the first step of the KiC algorithm, given their buoy speed above 6 knots. (C) The step 2 of the KiC algorithm relies on the comparison of changes in buoy speed with those found for “constant” and “transition sequences. Here, the value of the speed change between the first undetermined position following a classified position is consistent with a transition sequence. The undetermined position is therefore classified as “at sea”. (D) The operation is performed along the buoy segment, classifying positions from neighbor to neighbor. (E) The same procedure is then carried out backwards (from the end of the trajectory to the beginning), considering the remaining unclassified positions.Table 1: Standard data format for buoy location data.

© Y. Baidai et al., Published by EDP Sciences 2022