SPEI : can't figure it out

[aulemahal]

Discussed in #1589

See comment : #1589 (comment) for a description of the bug.

Issue comes from this line :

xclim/xclim/indices/stats.py

Line 765 in bd6a6f8

da = da.rename(day="time").reindex(time=da_ref.time.dt.dayofyear)

Having freq='D' (or freq=None with a daily dataset) results in the step previous to the above line grouping the data with time.dayofyear. Thus at the line above, the "temporal" dimension is named dayofyear and not day, which is why this fails.

Code of Conduct

• I agree to follow this project's Code of Conduct

[coxipi]

I thought I had a test for daily-SPI, apparently not.

I've made the changes, and wanted to add tests to compare with monocongo, but the results don't match. It's a bit hard to find a 3rd party for a tie-breaker between xclim and monocongo since the daily SPI is not that common. I wanted to try with this one: https://github.com/e-baumer/standard_precip, but it only partially obtain the correct values I've established for monthly SPI where monocongo and xclim do match.

[aulemahal]

I think #1607 fixed that issue? Can you confirm @coxipi ?