About the impact of using weather forecasts in statistical weatherrouting studies

Ship weather routing is particularly at stake in the shipping industry to meet decarbonization objectives. Besides, as wind propulsion is one of the solutions for effective decarbonization, new applications of weather routing appear. First of all, technically, few algorithms are versatile enough to deal with any kind of propulsion (sail, motor, and hybrid). In this context, D-ICE developed an efficient and robust multi-objective weather routing algorithm that reaches this goal and is embedded onboard Canopée, the first modern sail-assisted cargo. 
Then, as hybrid-propelled ship performances are directly linked to weather routing, route optimization has to be integrated directly during the design phase, to accurately assess the viability of a project and quantify the associated return on investment. To do so, D-ICE is often acting as a third party for ship owners to compute statistical weather routing studies, meaning simulating hundreds of voyages in the past years, using ECMWF reanalysis weather datasets.

However, a potential bias in those statistical studies is the computation of optimal paths based on historical weather data, meaning that the route is optimized knowing in advance the weather evolution, which is not the case in operation as only forecasts are available. This is a critical issue for the relevancy of such studies in the decision-making process of ship-owners.

Here we present a study whose objective is to quantify this bias in statistical routing studies. To do so we access ECMWF archives of forecasts and simulate hundreds of past voyages following an operational strategy: meaning a daily weather routing, computed with available weather forecasts at that time. The savings associated with this route optimization strategy are then compared to the results given based on the historical weather data. Some insight about weather routing robustness and weather conditions that could lead to uncertainties will be highlighted, and a compromise will be discussed between computing time (drastically increased with such procedure) and results accuracy.

Keywords: weather routing, wind propulsion, multi-objective shortest path algorithms, voyage optimization, historical reanalysis dataset, operational archives