SSH Frequency Spectra

SSH Frequency Spectra were computed at two sample points in the model domain. In each figure, 3 spectra are shown : 2 spectra from TOPEX at a track crossover point, and 1 model spectra from daily average output (see Figs. 18-19). One sample point is near the North Atlantic Current and the other is very near the Bravo mooring in the convecting region of the interior Labrador Sea. The model fails generally to have enough energy at most timescales but especially at high frequencies. It should be noted that the model run used for comparison here had convective adjustment every 12 hours, not the standard 30 minutes. However, the model fails at high frequencies even outside the convective region. For this reason, I do not believe that the frequency of the convective scheme will greatly alleviate the problems in producing high frequency energy. On the other hand, the model does much better at long timescales ( 80 days - 1 year) . In the case of the North Atlantic Current point, the model actually has as much energy as the altimetric observations. This is a remarkable improvement from previous studies where model energy is consistently lower than observations at all periods. The long period variations in this region probably reflect the model's ability to track the meanders of the North Atlantic Current. By shortening the length of the timeseries, it was possible to isolate the energy in the seasonal cycle. This calculation reveals $50 cm^{2}$ variance or a 30 cm yearly change in sea surface height around the North Atlantic Current. This is consistent with the strength of the current. In the case of the interior Labrador Sea, the results are not as good. Now, model energy is too low at all scales although low frequencies are more accurate. As reported previously, the model's interior Labrador Sea does not capture enough total variance of the sea surface. A quick comparison between the two model spectra shows that the difference in energy comes from the low frequencies. The interior point has a very flat plateau in the low frequency region, periods of 50-100 days. Mesoscale eddy variability can not be directly forced at a point so far removed from the lateral boundaries. Perhaps, the model's rim current needs to be more baroclinically unstable or the surface forcing field needs more variability in this frequency range. To explicitly resolve all eddies in the model, gridpoint separation should be on the order of 4-5 km. This study suggests that complete resolution of eddies will have the effect of increasing the variability at all larger scales even if the variability on the smallest resolvable scales does not increase dramatically. Although examination of an in-situ spectra is not done here, it is expected that the total energy at the timescales of convection is actually small due to the red geophysical spectrum. It is probably more important to capture the eddies that form as a response to convection if the goal is produce enough total variability in the model. It is not obvious that adequate resolution of eddies ensures that their formation may occur from still smaller unresolvable processes. However, the net result of increasing resolution to 4 kilometers is likely to be very substantial in a wide range of frequencies.