Page 82 - GIS for Science Technology Showcases
P. 82
SPATIOTEMPORAL MACHINE LEARNING Modeling transshipment patterns
Orhun Aydin, Esri
The term transshipment refers to the transfer of cargo, crew, or supplies from one vessel to another. A common type of transshipment occurs between fishing vessels and a refrigerated vessel, also known as a reefer. This practice allows shipping vessels to operate more effectively by reducing round trips to port. Despite its advantages for fishing, transshipment is also linked to human trafficking and the practice of forced labor, because it forces crews to stay onboard for extended hours. From a sustainability point of view, transferring a catch from one fishing vessel to another can obscure the actual location of the catch, making it relatively easy to evade quota requirements and regulations. Transshipment also poses a growing challenge for managing fisheries due to undocumented transactions of catch in international waters, which in turn facilitates illegal catch entering the seafood market. Transshipment activities often occur in international waters where the policy to counter this practice or regulate it is cumbersome.
Exploring and wrangling transshipment data
Understanding the spatial and temporal patterns of transshipment data requires exploring two distinct types of movement patterns of vessels: encountering and loitering. An encountering event is characterized as two vessels remaining close and moving together slowly for minutes or hours. A loitering event occurs when a vessel capable of transporting goods to the port travels at low speed, waiting for other vessels to approach.
Spatial clusters of transshipment
Defining data-driven spatial clusters of transshipment events reveals that some areas of international waters need monitoring. These clusters can be used to derive actionable maps for policy. Understanding “transshipment hotbeds” raises a pertinent question: In which areas are transshipments concentrated? Extensive transshipment occurs when a reefer can meet multiple vessels with ease. So it is vital to understand areas where encountering and loitering events are concentrated. In GIS, visualizing concentrations of this type can be accomplished using density- based clustering and HDBScan tools in ArcGIS Pro.
Using ArcGIS Pro to analyze data provided by Global Fishing Watch, encountering and loitering events are displayed as a data clock or a radial histogram to visualize the events temporally.
Spatiotemporal colocation of encountering and loitering vessels
Two ships must be at the same location at the same time to transfer goods. Data science methods that can mine proximity from two space-time data sources are required. Spatiotemporal colocation of encountering and loitering events summarize space-time neighborhoods where two events occur significantly and frequently. If the spatiotemporal configuration indicates significant clustering in space and time, significant colocation is indicated. In contrast, events that occur significantly far from each other indicate significant isolation. In transshipment, isolation areas correspond to movement corridors where vessels and reefers are progressively separated in space. Colocation corresponds to areas where significant time has been spent when vessels were nearby, pointing to catch transfer locations.
216 GIS for Science
Data clocks are used to analyze patterns in the temporal distribution of data in multiple frequencies.
Encountering events Number of encounters
≤ 51
≤ 100 ≤ 193 ≤ 343 ≤ 583
Loitering events Duration in hours
≤ 9.31 ≤ 12.46 ≤ 17.06 ≤ 29.42 ≤ 53.34