Repository navigation
SOM-based tomographic bin definition + n(z) #274
Description
Activity
- added a parent issue
on Jul 15, 2026 It should be checked whether sp_validation is the good place for this or if it will be produced within CosmoPipe.
we currently can't file issues on the UNIONS PhotoPipe so this was the best i could do for now. see this Slack thread for more. but i agree this is CosmoPipe/PhotoPipe territory
Define tomographic bins and n(z) via a SOM. Owner: Klara.
Plan (from the Bochum telecon with Angus Wright, 21 Sep; 22 Sep call notes): follow KiDS-Legacy. One SOM (one magnitude + four colours) trained on UNIONS photometry in W3, with the spectroscopic samples there. Gold weighting rather than a gold cut. Constant-shear image sims. Clustering redshifts (DESI overlap) cross-check the mean. The two UNIONS z filters are mapped onto one system. If W3 turns out unrepresentative of the survey, fall back to two SOMs with a transfer function (DES Y6). Pipeline: CosmoPipe. Next: a first n(z) on current data.
Open: equipopulated vs a-priori bin boundaries.
Context: Paris meeting notes