Ship‐icing prediction methods applied in operational weather forecasting

Sea‐spray wetting of ships operating in cold environments imposes a great safety risk, due to icing. For this reason, marine‐icing warnings have been a part of operational weather forecasting for the last five decades, yet verification of such warnings has only been done sparingly. This article evaluates different ship‐icing methods applied in operational weather forecasting. The methods are tested against a unique dataset from a single ship type from Arctic–Norwegian waters and two screened datasets from several ship types from Alaska and the east coast of Canada. Missing and uncertain parameters in the latter datasets are supplemented by reanalysis data from different sources. Continuous icing‐rate verification and sensitivity tests are presented for the physical icing models alongside categorical icing‐rate verification, which is applied in order also to evaluate icing nomograms, which are still used by several forecasting agencies. Furthermore, a newly proposed definition of the boundaries between icing‐rate severity categories is applied in the categorical verification procedure. The overall best verification scores for continuous and categorical icing rates are obtained by the Marine Icing model for the Norwegian COast Guard (MINCOG) and a physically based Overland model, updated from its initial version with more realistic heat transfer. Finally, sensitivity tests highlight that very low air and sea‐surface temperatures rarely occur over sea areas together with high waves, due to fetch limitations, even for strong winds. For this reason, models and nomograms that do not treat wind speed and wave height separately will provide inaccurate predictions of the icing rate in such areas. Consequently, it is preferable that methods applied in operational weather forecasting are replaced with methods capable of taking this effect into account.