Using data assimilation to improve thermal structure prediction in Lake Erie

Session: Improving Model Predictions Through Coupled System and Data Assimilation (2)

Pengfei Xue, Michigan Tech, [email protected]
Xinyu Ye, Michigan Tech University, [email protected]
Philip Chu, NOAA/GLERL, [email protected]
Eric Anderson, NOAA/GLERL, [email protected]
Huang Chenfu, Michigan Technological University, [email protected]
Gregory Lang, Great Lakes Environmental Research Laboratory, [email protected]

Abstract

The Great Lakes region has a relatively dense and long-term observational record, and these observational data have been widely used for model initialization and verification, but rarely been blended into model simulations to improve short-term forecasts or create reanalyses through data assimilation (DA). In this work, we test and evaluate the application of DA to a case study of the hydrodynamic simulation of Lake Erie. The moored instrument data and satellite data are incorporated into a data-assimilative hydrodynamic model for DA analysis and evaluation. Results show that DA can effectively improve the model performance with limited observational data when the DA formulation is appropriately developed in recognition of the dynamic complexities and anisotropic error covariances of Lake Erie. Furthermore, the data assimilative model also improves forecasting accuracy and restrains the forecasting uncertainty to an acceptable level on a timescale of 1-7 days after unleashed from DA. Lastly, data sampling strategies based on error correlation map is also examined. Results show the method can effectively reduce the sampling efforts while achieving the similar assimilation results with a potential to lead to the optimized design of observation network.