04.12.11: Link between SST gradient and BVF and statistics of BVF¶
We look first at the relationship between the gradient of sea surface temperature (SST) and near-surface Brunt-Vaisala frequency (BVF) in the regional HYCOM simulation (Fig. 1) and over the domain 210°E-208°E and 22°N-25°N. Without any spatial smoothing, the SST gradient and near-surface (10-m) BVF are well correlated during Winter, consistent with visual comparison of the two fields. During Winter, most of the near-surface BVF is near zero except within SST filaments, indicating surface mixed layer instabilities (SMLIs). Thus the high skewness of the distribution of near-surface BVF in Winter is directly indicative of the presence of localized values of high near-surface BVF and SMLIs in Winter (Fig. 2).
In order to see if a similar property is present in the observations, I compute the skewness of the near-surface BVF in Winter and Summer for different depths using the HYCOM simulation and the ARGO dataset for the same region. On average, the distributions in the observations are more skewed than in the model. Except very near the surface, the Winter distribution is always more skewed than the Summer one.
Figure 1: Spatial correlation between SST gradient and near-surface (10-m,) BVF. No spatial smoothing is applied. The dashed line shows the 95% confidence interval.
Figure 2: Skewness of the distribution of near-surface BVF over the domain.
Figure 3: Skewness of the distribution of BVF at different depths in the HYCOM simulation and in the ARGO dataset for the same region: (red) during March-August and (blue) September-February. The 95% confidence interval for the ARGO distributions are shown with dot lines. The interval for the HYCOM simulation is a bit long to compute but it is tiny for the shallowest depth.
- Because the difference in skewness in the observations between Summer and Winter is more pronounced within the mixed layer (say 30 m) than at the shallowest avaiable depth (7.5 m), do we still have a good correlation between the SST gradient and BVF at these depths in the model? The answer is yes as shown in Fig. 4. However, there is much less seasonal variability in the skewness at that depth (Fig. 5).
- Notice that the variability of the skweness follows that of the correlation; this is consistent with the idea that when SMLIs occur, SST gradient and 30-m BVF are correlated resulting in localized high values of BVF and thus high skewness, and inversely. This also suggests that the skewness of BVF at 30 m is a direct signature of SMLIs.
Figure 4: Same as Fig. 1 but for BVF at 30 m instead of 10 m.
Figure 5: Skewness of the distribution of 30-m BVF over the domain.
computed with link_gradSST_N_ano.m in RESEARCH/PROJECTS/MARINE_BIOLOGY/SUBMESOSCALE_PROCESSES/FSLE/analysis/HYCOM/ on the central disk.