Summary of my recent work (08/19/10)

1. The ultimate goal is to explain the physical process that causes the large events of nutrient injection into the euphotic zone observed by Johnson et al. (2010). Say differently, what causes these vertical velocities.
2. From recent work, two hypothesis can be proposed: eddy pumping, inside eddies, and submesoscale processes, mostly at the edge of eddies or outside eddies within isolated filaments.
3. In the 1/30th-of-a-deg. OFES simulation, energetic submesoscale processes, with large vertical velocities, appear (see Fig. 5 in this note). These features are trapped within the surface mixed layer (SML; see Figs. 12a and 13a in this note) so that they are just above the observed depths of the observed nutrient injections.
4. Submesoscale features, albeit of larger horizonal scale and smaller vertical velocities, are also observed in the 1/10th-of-a-deg. OFES simulation. They are also trapped in the SML (see Figs. 12b and 13b in this note). An immediate question is: Are these submesoscale features associated with density anomaly and/or nutrient injections?
5. If these submesoscale features are relevant to the nutrient injections, one then need to a) find an index computable from ARGO/AVISO data that can localize these features and check that these features are co-located with tracer (nitrate, oxygen and.or salinity) injections observed by ARGO floats.
6. Using the 1/30th-of-a-deg. OFES snapshot, it was found that both positive and negative values of the full and AVISO-like spatial resolution Okubo-Weiss (OW) correspond to submesoscale features defined as large R=ζ/f (Figs. 3a, 4a and 8 in this note) so that OW is not an appropriate index. Task: Compare FSLE with vertical velocity field to see if FSLE can be such an index.
7. If what matters is the vertical velocity below the SML, then it was found that such field in the 1/10th and 1/30-of-a-deg. OFES simulations, is dominated by high-frequency (period smaller than the model output resolution of 3 days) waves that do not significantly alter the depths of isopycnal surfaces. Thus, this wave might not be relevant to explain the nutrient injections. This suggests that the vertical velocity that is relevant needs to be obtained differently. For instance, we could use 1 or 3-day averaged model output or, assuming that the vertical velocity has no diapycnal component, we could estimate it from the vertical velocity of the depth of isopycnal surfaces. However, to confirm that the vertical velocity is mostly non-diapycnal, one would need to see if, using this vertical velocity, the conservation equation of nitrate (that involves not only vertical advection but also horizontal advection and biological processes) can be retrieved.

Meeting with Kelvin (08/19/10)

1. The pattern of salt in Fig. 2 of this note is very different from the time series of salinity measured by the ARGO float of Johnson et al. (2010; their Fig. 3c). Kelvin thinks that in the model and in that region, the salinity is dominated by lateral stirring. Try to find a region where the mean horizontal gradient is weak so that lateral stirring is kept at a minimum.
2. Compare mean stratification between model and observations; maybe in the model, stratification is too strong, preventing the submesoscale vertical velocities to reach the depths 100-150 m at which the vertical injections of nitrate have been observed.
3. Look at the depth of an isopycnal to see if there is any submesoscale structure (as we would conjecture if Johnson et al.’s observations are due to submesoscale processes but not as we would conjecture from various numerical modelling studies, in particular Lapeyre and Klein (2006) and Thomas et al. (2007) that show that the submesoscale signature in density field is weak at best).