Improvement of Coupled Model Skills Through Bias Corrections with Additional Surface Boundary Layer Considerations

The forecast capabilities of coupled atmospheric-oceanic numerical models is dictated mainly by the ability of the included parameterizations in representing the surface forcing, and their complex relationships. For increasing accuracy, one must reduce errors by iteratively adjusting each variable involved according to certain parameters, and re-evaluate each time the outcome at global level. In the particular case of the ESPC (Earth System Prediction Capability) system, we aim here to adjust the air-sea fluxes, with different spatial resolution in the atmospheric and oceanic side, and also the WAVEWATCH III output, both fundamental pieces for the system accuracy.

For the air-sea fluxes, an ensemble reanalysis of 16 members for one year was created, and biases were determined by comparison with the NFLUX system, a satellite-derived gridded-product developed by the US Navy. The variables are then separately averaged over different periods, i.e., 1, 5 and 15 days, to determine the ability of a particular average to represent the system behavior. Then, a correction was developed for each variable, for each month, to reduce the fluxes errors in real time.

For the wave model instead, we focused in the bias in the significant wave height (H_s) when compared to in-situ observations. We are presenting a purely empirical method to reduce the model error by modifying the wind speed at the pre-processor stage, to produce fields of ‘effective wind speeds.’ Such formulation is based on historical biases in Hs and correlation with ASTD. We then compare the model results obtained with the two different approaches to account for the effects of air-sea temperature differences: 1) using existing methods available in WAVEWATCH III^® versus 2) using this new and empirically-based ‘effective wind speed’. An alternative approach using equivalent neutral winds is also considered.