A probabilistic approach to ship voyage reconstruction in ICOADS

摘要: The International Comprehensive Ocean-Atmosphere Data Set (ICOADS) provides the main archive for surface marine observations past approximately 150 years. ICOADS ship identifier (ID) information is often missing or unusable, preventing linking of reports to an individual ship. A method reconstruction voyages in presented, by which groups can be associated with track. defines a function representing probability density (pdf) any particular report being group reports. parameters pdf are calculated from data themselves, giving likely variation perpendicular its overall direction travel. For ID information, PDF used associate without known-ID Reports then clustered together form most probable Results shown period 1855–1969. Both percentage tracks and length those increase substantially following tracking. Initial validation results was performed visual inspection: model implementation refined improve results. Confidence tracking increased demonstration that clusters similar sea temperature characteristics. Issues were found one challenges implementing technique. Particular problems encountered included coarse resolution some position information; mispositioned either space time; unidentified duplicate reports; fragmentation between different acquisition sources. Some these effects could ameliorated pre-processing reports, however full reprocessing historical input sources would required make further improvements.