A data-driven framework for attainable ship speed uncertainty under stochastic weather conditions

This paper presents a data-driven framework for quantifying attainable ship speed uncertainty considering weather forecast uncertainty. The methodology integrates two parallel workflows: weather forecast processing and ship performance simulation. Weather forecast data from NOAA and GFS sources are collected at multiple lead times (0-24h, 24-72h, 72-120h, 120-168h). The data undergo spatial discretisation over a North Atlantic rectangular grid, extracting the main meteorological variables, including significant wave height, peak period, wave direction, wind speed, and wind direction. Ship performance simulations were done using Wärtsilä NaviTrainer NTPRO 5000 and HydroComp NavCad to generate attainable ship speed lookup tables under varying conditions: intended speeds (14.5, 13.5, 12.0 kn), wave heights (0-14 m according to WMO Sea State Codes 0-8), and wave encounter angles (0°-180°). Multiple metrics were used for uncertainty quantification, including RMSE, MAE, Bias, UGR, CRPS, IoA, and FSS for meteorological variables, alongside CMAE for directional parameters. These metrics are subsequently applied to estimated attainable ship speeds, establishing response variable uncertainties. Correlation analysis was conducted between the uncertainty of meteorological variables and the uncertainty in attainable ship speed, providing important insights for estimated time of arrival (ETA) calculations and voyage planning under weather uncertainty.