Determining bias correction and covariance parameters for Optimal Estimation of Sea Surface Temperature
In the context of the OSI SAF Visiting Scientist Program, Christopher Merchant from Univeristy of Reading worked at Météo-France in February 2019 on determining bias correction and covariance carameters for Optimal Estimation of Sea Surface Temperature.
Optimal estimation SST is currently undertaken with error covariance matrix assumptions based on heuristics and expert assumptions. Incorrect specification causes the “optimal” retrievals in fact to be sub-optimal. This study demonstrates new methods to estimate appropriate parameters for OE, adapting ideas from Kalman filtering and Desroziers diagnostics in data assimilation. These methods are shown to improve SST error statistics and retrieval sensitivity. Additional insights into the nature of prior and forward model biases, including the degree of cross-channel simulation-error covariance and angular dependencies, are further obtained.