Adding error covariance methods based on forecast differences

Error covariance methods based on forecast differences have been implemented:

• NMC (differences of varying length forecasts valid at the same time)
• Canadian Quick Method (differences based on forecast lags)

Development of code to estimate representation errors

In addition to the calculation of HL error covariances, the scope of this code repository has increased, also containing codes to estimate representation errors associated with sub-grid scale variability (Oke and Sakov, 2008). The existing code was refactored, so functions are now more generic, especially the ones associated with IO and masking.

Development of HL method and function fitting

This is the first release of the ocean error covariance code repository. In v1.0.0 we include the Hollingsworth and Lonnberg method to compute observation-minus-forecast data. We also include the respective function fitting code to estimate the magnitude and length-scales of the errors.