search

gher-uliege / DIVAnd.jl

DIVAnd performs an n-dimensional variational analysis of arbitrarily located observations
GNU General Public License v2.0
72 stars 11 forks source link

[👆] Getting lists of data points that are used for each depth layer #118

Closed ctroupin closed 1 year ago

ctroupin commented 1 year ago

Describe your objective

Users would like to know which data points are used for an interpolation on a given layer, for instance winter at 500 m depth, knowing the vertical correlation length.

Explain why it is not currently possible to do it?

I guess in the loop it would be a matter of writing files (such as the ones written by saveobs) storing all the obs.

[Optional] Describe how you would solve that

Also would be nice to have an illustration of the vertical correlation length (in a notebook for example) because it seems to cause confusion with several users. For instance, if one increases the vertical correlation length, is it expected to get more or less variability.

jmbeckers commented 1 year ago

If the direct solver is used, ALL data points are used though only those within the reach of the length scales have a significant impact.

If you use diva3d with vertical tiling, then of course you limit the data to those within the vertical window

In principle, increasing L filters more and reduces variability for homogenous data coverage . But maybe with a lot of data near the surface and only a few in deep layers, increasing Lv could actually increase variability because you extrapolate more.

ctroupin commented 1 year ago

yes that's what I was thinking about the influence of Lv, maybe in some cases it can increase variability.

Also for the data used I think there might have been a confusion (now solved) with the parameter searchz (used in the fitting) but not in the interpolation/analysis