How can I change the units 'm' to 'mm' for ERA5 precipitation dataset.

[CGL5230]

Generic Issue

• xclim version:0.46.0
• Python version:3.11
• Operating System:Linux

Description

I download a ERA5 daily_mean precipitation dataset. I want to use this dataset to calculate the AnnualPrecipitation. I use the command as below. Here is the information of precipitation dataset after I revise the units. Before the changing activity, the units is 'm'.

What I Did

pre_ERA5.tp.values = pre_ERA5.tp.values*24*1000
pre_ERA5['tp'].attrs['units'] = 'mm'

What I Received

ValidationError: Data units millimeter are not compatible with requested [precipitation].

What I found

I don't konw why I got this error, even I changed the units to 'mm'. I want to know how can I fix this problem.
Then I search the xclim user guide and find the function xclim.core.units.convert_units_to(source, target, context=None)
Can I change the units of precipitation like:

pre_ERA5 = xclim.core.units.convert_units_to(pre_ERA5, 'mm', context=None)

Code of Conduct

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

[aulemahal]

Hi @CGL5230 !
This is because xclim calls "precipitation" the precipitation flux. We follow language from the "Climate and Forecast" (CF) conventions where "flux" is a mass per area per time, usually "kg m-2 s-1". Xclim can also receive a "rate", i.e. "length per time", like "mm/day", because the transformation from rate to flux is always the same for liquid water : multiply by the density of water which is 1000 kg/m³.

In your case, the data you have seems to be a total thickness of liquid water for the current timestep, which seems to be a day. If you confirm that the dataset is daily then an explicit and easy way out would be to patch the units like this:

pre_ERA5['tp'].attrs['units'] = 'm / day'

Xclim will convert to "kg m-2 s-1" under the hood.

But in this precise case, convert_units_to should be able to handle it, you simply need to request a precipitation flux unit:

pr = xclim.core.units.convert_units_to(pre_ERA5.tp, 'kg m-2 s-1')

Be aware however, that this works because the input has all needed metadata for xclim to understand how to handle the conversion! Xclim will read the standard_name and understand that this variable speaks of liquid water. It will also look at the frequency of the data. It will then enable the conversion from "[length]" to "[mass]/([area][time])".

[CGL5230]

Thanks for quickly reply. I use the command as pre_ERA5['tp'].attrs['units'] = 'm / day' to change the units. And I got the waring like below when I run the ERA5_rx1day = xclim.indicators.atmos.max_1day_precipitation_amount(pr='tp',freq='YS',ds=pre_ERA5). I think the reason is the units and variable name not the standard name for xclim. So,the result will be correct for R1XDAY index ?

/home/anaconda3/envs/py311/lib/python3.11/site-packages/xclim/core/cfchecks.py:41:
UserWarning: Variable does not have a `cell_methods` attribute.
 _check_cell_methods(
/home/anaconda3/envs/py311/lib/python3.11/site-packages/xclim/core/cfchecks.py:45: UserWarning: Variable has a non-conforming standard_name: Got `lwe_thickness_of_precipitation_amount`, expected `['precipitation_flux']`
 check_valid(vardata, "standard_name", data["standard_name"])

[CGL5230]

I want to clarify this command pre_ERA5['tp'].attrs['units'] = 'm / day'. I got the daily statistic precipitation dataset from ERA5 data pool. You can find the calculate step from document. The precipitation dataset is a daily mean dataset.

Notice my comand:pre_ERA5.tp.values = pre_ERA5.tp.values*24*1000. I change the precipitation dataset as a daily sum dataset and now the unit of values actually is 'mm'. So I use this command to convert the units.

pre_ERA5.tp.values = pre_ERA5.tp.values*24*1000
pre_ERA5['tp'].attrs['units'] = 'mm / day'

But I still got the waring like before.

/home/anaconda3/envs/py311/lib/python3.11/site-packages/xclim/core/cfchecks.py:41:
UserWarning: Variable does not have a `cell_methods` attribute.
 _check_cell_methods(
/home/anaconda3/envs/py311/lib/python3.11/site-packages/xclim/core/cfchecks.py:45: UserWarning: Variable has a non-conforming standard_name: Got `lwe_thickness_of_precipitation_amount`, expected `['precipitation_flux']`
 check_valid(vardata, "standard_name", data["standard_name"])

Hi @CGL5230 ! This is because xclim calls "precipitation" the precipitation flux. We follow language from the "Climate and Forecast" (CF) conventions where "flux" is a mass per area per time, usually "kg m-2 s-1". Xclim can also receive a "rate", i.e. "length per time", like "mm/day", because the transformation from rate to flux is always the same for liquid water : multiply by the density of water which is 1000 kg/m³.

In your case, the data you have seems to be a total thickness of liquid water for the current timestep, which seems to be a day. If you confirm that the dataset is daily then an explicit and easy way out would be to patch the units like this:

pre_ERA5['tp'].attrs['units'] = 'm / day'

[aulemahal]

The warnings are indeed simply because the metadata of your dataset is not exactly what xclim expected, especially the standard_name that becomes wrong after our unit change.

If you use convert_units_to, the standard name will be modified as well and the warning¹ should go away. But it's only a warning, and if you know that your data is valid, then it's safe to ignore.

I don't have access to the first link you gave, but from the metadata you showed, it seems that pre_ERA5['tp'] is the total precipitation accumulated over one timestep. So the liquid-water-equivalent thickness, in "m". To transform to a rate, you simply need to divide by the length of time of that timestep, which here is a single day.

precipitation_rate = total_precipitation  / 1 day

Which is why simply changing the units from "m" to "m/day" should work. I'm not sure I understand why you wanted to multiply by 24 ? That would mean the value is the mean hourly precipitation accumulation, which is not what I infer from the attributes.

Footnotes

1. That is the warning about the standard_name. The other one about the cell_methods will stay, but again, this is safe to ignore, it's just xclim being a bit over-careful. ↩

[CGL5230]

Thanks for your quick reply. I offer the document link to show the precipitation dataset actually is daily mean dataset which would mean the value is the mean hourly precipitation accumulation. Because I want to get a daily-sum precipitation but hourly precipitation. The daily-sum precipitation can be used to calculate the annual precipitation.

Which is why simply changing the units from "m" to "m/day" should work. I'm not sure I understand why you wanted to multiply by 24 ? That would mean the value is the mean hourly precipitation accumulation, which is not what I infer from the attributes.