frequency_analysis indicator weirdly declared

[aulemahal]

Setup Information

• Xclim version: master
• Python version: all
• Operating System: all

Description

xclim/xclim/indicators/land/_streamflow.py

Lines 69 to 78 in 1505aac

	freq_analysis = FA(
	identifier="freq_analysis",
	var_name="q{window}{mode:r}{indexer}",
	long_name="N-year return period {mode} {indexer} {window}-day flow",
	description="Streamflow frequency analysis for the {mode} {indexer} {window}-day flow "
	"estimated using the {dist} distribution.",
	units="m^3 s-1",
	title="Flow values for given return periods.",
	compute=declare_units(da=None)(frequency_analysis),
	)

Here you see that the output of frequency_analysis is a discharge (var name q, units m³/s). However, the input variable has a generic name da and no prescribed dimensionality. This would mean that if you pass anything else then a discharge, the unit check will not fail, but the indicator will fail when attempting to convert the output units.

Indeed, the next example fails. But even weirder, is that we have a special CF check that looks for the "water_volume_transport_in_river_channel" standard name. Thus, we already are expecting a discharge.

Steps To Reproduce

import xclim as xc
from xclim.testing import open_dataset

ds = open_dataset('ERA5/daily_surface_cancities_1990-1993.nc')
xc.land.freq_analysis(da=ds.dtas, t=100, dist='norm', mode='max')

warns:

UserWarning: Variable has a non-conforming standard_name: Got `dew_point_temperature`, expected `['water_volume_transport_in_river_channel']`

and fails:

DimensionalityError: Cannot convert from 'kelvin' ([temperature]) to 'meter ** 3 / second' ([length] ** 3 / [time])

The expect error would be:

ValidationError: Data units K are not compatible with requested [discharge].

Additional context

@RondeauG

Contribution

• I would be willing/able to open a Pull Request to address this bug.

Code of Conduct

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

[RondeauG]

Adding to this issue, I think a discharge variable should be added to VARIABLES if we want to add and support more and more discharge-related indicators (which I think is the case?).

As for the name, I'd argue that q is not good, since this is already used for "specific discharge" in hydrology, with units in m3 s-1 km-2. Q is used for discharge in m3 s-1, but I don't think that we want a capital letter?