Sea Surface Height Variability

The large-scale spatial pattern of sea surface height variability is captured by the model (see Fig. 15-17). The North Atlantic Current extends through the southeast corner of the model domain and is associated with very large variability in sea surface height. The model captures roughly 75 percent of the energy in this region despite the proximity to the model boundary. In fact, the boundary conditions only give abnormally low variability within the sponge layer (nearest 5 grid points to the the boundary) without a noticeable effect to the interior. For the year of Oct. 1996-1997, the Topex/Poseidon altimeter observes a maximum SSH variance of nearly $700 cm^{2}$ in the western boundary current extension, while the model predicts maximum values of $500 cm^{2}$. The model also correctly gives lower variability in the center of the Labrador Sea. The point-by-point correlation coefficient between model and altimetry SSH variance in the entire model domain is .75, primarily due to the model's ability to distinguish between high variability in the North Atlantic Current and lower variability in the Labrador Sea proper. However, a closer look at the center of the Labrador Sea reveals slightly more ambiguous results. In this area, the model only captures 40-50 percent the total SSH variance. The spatial mean SSH variance is $40 cm^{2}$ in the model, $80 cm^{2}$ in the altimetry data. Next, the model predicts a local maximum variance in the rim current of $50 cm^{2}$, but much smaller variability in the center of the Labrador Sea's convecting area. The altimeter also sees a slight maximum in the rim current, but there is almost the same amount of variability in the center of the Labrador Sea. The model's spatial pattern is not quite as good in this smaller region, despite its successes on the large scale. The signal in the data is much larger than the noise , which is estimated to be on the order of $4 cm^{2}$ (Fu et al, 1994) and can not possibly explain the discrepancy in the center of the Labrador Sea. It has been speculated that lack of model resolution is responsible for the low background variability away from frontal structures (Stammer and Boning, 1996). This seems to be a logical explanation in regions of convection since such small spatial scales can not be explicitly seen even by this relatively high resolution model. However , this model is eddy-resolving and we should see more variability than lower resolution models if this hypothesis is correct. Indeed, previous studies with 2 degree resolution global GCM's only capture 25-50 percent of the sea surface height variance (Stammer et al, 1996). In energetic regions such as the West Greenland rim current and the North Atlantic Current, this model consistently captures at least 75 percent of the variability. Although sea surface height variability in the convective region is not captured to the same degree, perhaps this should be expected by a model which does not explicitly resolve such small features. The degree of resolution needed can be more adequately discussed after a study of the frequency spectra.