Application de systèmes de prévision hydrodynamique à haute résolution pour la prévision des conditions océaniques dans des ports du Canada et leurs approches

Abstract

Port-scale Ocean Prediction Systems (POPS) were developed for six high priority ports across Canada (Vancouver Harbour, Fraser River, Kitimat, Saint John Harbour, the St. Lawrence Estuary, and Strait of Canso) for use in support of dynamic electronic navigation and marine emergency response. The POPS configurations are based on a common base technical configuration. The design/configuration of each POPS was reviewed and generally considered appropriate for the intended purpose, within limitations noted below. Ability of the six POPS in reproducing real-world conditions was assessed against available observational data. The POPS generally show skill in predicting water levels, currents, and water properties at a scale that do not currently exist in available operational models. The models provided coverage for some areas not covered by the current version of the Coastal Ice-Ocean Prediction Systems (CIOPS), and generally matched or outperformed CIOPS in most comparisons against observations considered for areas of overlap between the models. Tidal and total water levels comparative statistics are both generally improved relative to CIOPS at water level gauge stations, and performance outside areas of CIOPS coverage is similar to that where comparison with the parent model is possible. Non-tidal water level performance is comparable to that of CIOPS. Storm surge was assessed using limited time periods spanning significant storm events in each POPS domain, and surge signals are clearly visible in both CIOPS and POPS. Storm surge comparative statistics in POPS generally match or improve upon those in CIOPS. Comparative statistics for currents are generally improved with increasing resolution of the models, particularly where flow around complex topography is present and more accurately resolved. In general, tidal currents are improved more significantly in the POPS than non tidal currents. The seasonal cycle of sea-surface temperature is well captured by the high-resolution models (POPS) with small biases matching or outperforming the larger scale models. Ferry track data generally shows equal or better performance in POPS relative to CIOPS. Regional biases in temperature and salinity are generally reduced, and the POPS skillfully reproduce changes in these water properties, with minor exceptions. Currents provided by the POPS yield slightly improved, or equivalent, predictions of drifter trajectories. The accuracy of trajectory predictions may be limited by the accuracy of available wind forecasts and/or limitations in the trajectory model itself. The POPS model simulations were stable for 5-6 year hindcast periods, which included extreme events in several domains. Fallback mechanisms for missing observational input data are used where relevant. Forecast evaluation focused on measuring the degradation of model skill as a function of forecast lead time. Model skill was not found to degrade substantially over the 48-hour forecasts, as error growth trends for water level and sea surface temperature were typically small. It is recognized that the POPS have certain limitations: Model accuracy is impacted by limits of the model inputs. Quantitative assessments of skill are limited to regions of the POPS domains where observations are available. Observational data are focused on the inner (port) domain and are typically sparser in the coarser outer domain. Current data are also more limited than water property and water height data. The ocean model version used in all POPS does not correctly represent intertidal areas, and consequently the model results are not considered usable in intertidal areas. Biases in temperature and salinity observed in specific locations could lead to errors in surface currents that could not be assessed with available data. The POPS are not coupled to a complete ice model, which could impact surface currents and potentially other variables for regions that experience ice cover. Port systems include region-specific features that are under-resolved and/or not reproduced in the models and may include portions of the domain that are not reliable.