Developing an EnKF Data Assimilation System for the Canadian MEOPAR Project, the Benefits and Costs of High Resolution

This study discusses the current status of development of a High Resolution EnKF (HREnKF) data assimilation system for the Canadian MEOPAR project (Marine Environmental Observation Prediction and Response Network). This system will eventually be part of a full Atmosphere-Ocean re-locatable forecast system that can be used within hours of a marine emergency (such as a search-and-rescue incident or oil spill) anywhere along Canada's coastline. The HREnKF system itself currently runs at 2.5 km horizontal resolution and generates analyses every hours. Increasing the resolution of atmospheric models and assimilation systems can be a doubles-edged sword. On the good side, higher spatial resolution allows a better representation of topography and physical processes. Higher temporal resolution also allows for short-lived/small-scale features to be represented in analyses. Unfortunately, the sparsity of observations often constrains the accurate depiction of such features. Even when available in sufficient density, surface observations are notoriously difficult to assimilate. Documenting the added value of high resolution and justifying the associated computational costs is therefore an active research topic within the MEOPAR research group presently. Our HREnKF system is nested with a research version of Environment Canada's Regional EnKF (REnKF) system that runs on the Canadian Meteorological Center's supercomputers in Dorval at 15 km horizontal resolution and generates analyses every 6 hours. Our verification efforts concentrate on the differences between the REnKF and HREnKF systems in terms of analyses and short-term forecasts up to 12h. As a first stage of our research program, for the most part, these two systems ingest the same types of observations. Both systems were run for a period of 14 days in the winter of 2011. Verifications against surface and radar observations will be shown at the conference. The relative impacts of increased spatial resolution versus assimilating observations more frequently will also be discussed.