For the same region studied with the ARGO float (see *this note*), we calculate the percentage of large depth anomalies (DA) for each isopycnal for various categories of sea surface height anomalies (SSHA). “Large” is defined again as “larger than one temporal and spatial standard deviation”.

We observe that: 1. Most large DA occurs either with large SSHA of same sign or with weak SSHA. 2. The cases ‘large DA and SSHA of same sign’ occur more than in the observations, 2. These cases are less frequent near the surface as in the observations, 3. At least 40% of large DA occurs with weak SSHA as in the observations.

Questions are then: 1. Why the percentage of ‘large DA and SSHA of same sign’ is larger in the model than in the observations? 2. Why this percentage decreases upward in both observations and model? 3. What is the physical cause of large DA with weak SSHA in the model?

Note that when twice the standard deviation is used to define what is large, the vertical structure of the percentage in Fig. 1 does not change but their mean position does; now, between 70-90% of large DA occurs with weak SSHA, and 10-30% of large DA occurs with large SSHA of same sign. This shows that the percentage itself is highly dependent on the definition of what is large.