Internal Model Variability of Ensemble Simulations With a Regional Coupled Wave-Atmosphere Model GCOAST

Abstract

Ensemble simulations are performed to quantify the internal variability of both regional atmospheric models and wave-atmosphere coupled model systems. Studies have shown that the internal variability in atmospheric models (e.g., wind or pressure fields) is increased during extreme events, such as storms. Comparing the magnitude of the internal variability of the atmospheric model with the internal variability of the coupled model system reveals that the internal variability can be reduced by coupling a wave model to the atmospheric model. While this effect is most evident during extreme events, it is still present in a general assessment of the mean internal variability during the whole study period. Furthermore, the role of this wave-atmosphere coupling can be distinguished from that of the internal variability of the atmospheric model since the impact of the wave-atmosphere interaction is larger than the internal variability. This is shown to be robust to different boundary conditions. One method to reduce the internal variability of the atmospheric model is to apply spectral nudging, the role of which in both the stand-alone atmospheric model and the coupled wave-atmosphere system is evaluated. Our analyses show that spectral nudging in both coupled and stand-alone ensemble simulations keeps the internal variability low, while the impact of the wave-atmosphere interaction remains approximately the same as in simulations without spectral nudging, especially for the wind speed and significant wave height. This study shows that in operational and climate research systems, the internal variability of the atmospheric model is reduced when the ocean waves and atmosphere are coupled. Clear influences of the wave-atmosphere interaction on both of these earth system components can be detected and differentiated from the internal model variability. Furthermore, the wave-atmosphere coupling has a positive effect on the agreement of the model results with both satellite and in situ observations.