Using 15deg dewpoint temp to identify ITD is not correct over western regions e.g. Senegal, need to additionally take account of winds.
Doug's suggestion for identifying ITD / Congo Air Boundary as confluence lines:
I think I have a neat solution for this, which emerged from thinking about velocity potential. Because these features are geographically defined to be E-W (ITD) and N-S (CAB), you should be able to identify them by analysing velocity fields across the domain in the lat-lon directions separately.
For the ITD I am suggesting you can integrate the equation
d\phi / dy = v
by finite differences, starting with \phi=const (e.g. \phi=0) at a fixed latitude comfortably south of the ITD.
Then the convergence-based ITD will by where \phi is maximum in y (by definition, v=0 at this maximum).
Then, for the CAB the process can be repeated, for latitudes south of the ITD (south of min latitude of ITD), solving
d\phi / dx = u.
I think this should work, and because it's an integral, it should be robust. The only problem will come if there is a jump in the latitude (of ITD) from latitude to latitude. In that case maybe the ITD can be broken if it jumps more than a couple of gridboxes?
Notes from SWIFT Science meeting 2/11/2021 (during EA TB3).
Strong signal CAB 11am
Plot confluence line - ITCZ/CAB
Integrate in wind and find min (min advection in one direction)
Using 15deg dewpoint temp to identify ITD is not correct over western regions e.g. Senegal, need to additionally take account of winds.
Doug's suggestion for identifying ITD / Congo Air Boundary as confluence lines: