Ouranosinc / xsdba / 26414010969

79%

push

github

Fix 278 - add n_last_quantile_filter (#279)

<!-- Please ensure the PR fulfills the following requirements! -->
<!-- If this is your first PR, make sure to add your details to the
AUTHORS.rst! -->
### Pull Request Checklist:
- [x] This PR addresses an already opened issue (for bug fixes /
features)
  - This PR fixes #278 
  - This PR fixes #289 
- [x] (If applicable) Documentation has been added / updated (for bug
fixes / features).
- [x] (If applicable) Tests have been added.
- [x] CHANGELOG.rst has been updated (with summary of main changes).
- [x] Link to issue (:issue:`number`) and pull request (:pull:`number`)
has been added.

### What kind of change does this PR introduce?

* The idea here is that, with RCMs, I am getting some extremes that have
never been seen in the reference data and when I apply the ajustement
factor to it, it explodes to really unrealistic extremes (>1650 mm/day).
So, I am adding an arg to say if the value I want to adjust is 10 (or
whatever n) times larger than the highest quantiles. I should not be
able to do quantile mapping, so I will leave it as is.
* I used to do this filtering outside of xsdba, but if I do it inside
after the adapt freq in adjust, I can fix the issue #278. In that case,
I have mostly dry days (q99 is very low and the af are basically noise).
In my adjust adaptfreq, I can come and add some wet days that can go up
to 20 mm/day. When I try to adjust this 20 mm/day with afs learned below
1 mm, I can get something very large. If the filter is activated, i will
get the adapt, but not the adjustement to crazy levels.

Let me know if you think this make sense!

I'll add some more documentation and changelog, if you agree with the
concept.

### Does this PR introduce a breaking change?
no, though dqm_train might be slower because I am computing the raw
quantiles also

### Other information:
@coxipi the idea I had on slack to modify adapt_freq to not go above the
last quantile did not work.

50 of 50 new or added lines in 4 files covered. (100.0%)

2880 of 3645 relevant lines covered (79.01%)

8.48 hits per line