Use of Satellite Data for Assimilation into Ocean Models

This is discussed at length in other chapters. Three important issues are emphasized here:

1. There can be large differences in data quality between real time and delayed mode (reprocessed) data sets. Depending on applications, trade-offs between time delay and accuracy often need to be considered.

2. Error characterisation is mandatory for data assimilation and a proper characterisation of error covariance can be quite complex for satellite observations. Data error covariance should always be tested and checked as part of the data assimilation systems.

3. It is much better in theory and for advanced assimilation schemes to use raw data (level 2 or in some cases level 1 when the model can provide data needed for level 1 processing). The data error structure is generally more easily defined. The model and the assimilation scheme should also do a better high level processing (e.g. a model forecast should provide a better background than climatology or persistence). However, in practice, this is not always true. Some data high level processing (e.g. correcting biases or large scale errors, intercalibration) is often needed as it cannot be easily done within the assimilation systems.

0 0

Post a comment