Evaluate impact of adding Argo float data

[jemrobinson]

Problem / Goal

How much does the Argo float data help? What accuracy/precision etc. did we have before? What do we have afterwards?

What does good look like?

• Can quantify impact of adding Argo float data

Related context

No response

[IFenton]

As part of this, we should test our assumptions about the time window (4hrs) / water depth (0-50m) that we are using - are they the optimal values?

[IFenton]

Some more thoughts on testing this:

• We’ve already discussed better ways of gridding data - the method we use is very crude
• To do a better comparison with our models we should be including a gridded version of sea surface temperature / salinity in the data as otherwise we’re not really comparing like with like
• I also suspect we should check again after we’ve investigated the grid reprojections / masking. Have we actually checked that adding ERA5 makes improvements to our model in the current setup? It should, but my current impression is that everything is dominated by the SIC at the moment.

[aranas]

To evaluate if change in performance is meaningful, increase the density of argo float data incrementally and see if the reduction in error behaves as expected (eg error roughly decreases with increasing amount of obs data).