Repository navigation
monthly-means request with inconsistent product type returns hourly data without warning #161
Description
What happened?
I requested the dataset reanalysis-era5-single-levels-monthly-means, but mistakenly specified product_type="reanalysis" and omitted the time parameter (this used to work for previous reanalysis, with the previous API).
Problem: The request completed without an error or warning. However, the returned data contain instantaneous hourly fields (1st of the month, at 12:00) rather than monthly averages.
I got this conclusion after downloading and comparing the same variable using the correct product_type="monthly_averaged_reanalysis" and time="00:00" (as suggested by the CDS web interface). I also downloaded and compared with the same variable at hourly resolution.
I would expect an unsupported product type for the selected dataset, or a missing required parameter, to produce a validation error rather than return data with a different temporal meaning.
Could request validation reject this combination to prevent users from unknowingly treating hourly fields as monthly means?
I have attached three scripts with API pull request to download a few data (~20Mb) for "hcc" (high cloud cover) for 2022:
- inconsistent monthly-means request
- corrected monthly-means request
- hourly request
I have also attached a plotting script and the three figures showing the returned field, the correctly requested monthly mean, and differences against the hourly reference fields.
Steps:
- execute the 3 scripts to download 20Mb of sample files
- execute the plotting script to create the comparison figures
- check the header of the "good" and "bad" files (check hcc:GRIB_stepType)
What are the steps to reproduce the bug?
example_hcc_compare.py
example_hcc_2022_mmeans_good_pullrequest.py
example_hcc_2022_mmeans_bad_pullrequest.py
example_hcc_2022_hourly_pullrequest.py
Version
0.7.7
Platform (OS and architecture)
Ubuntu 24.04.4 LTS (64 bits)
Relevant log output
/opt/miniconda3/envs/geo/bin/python -m pip show cdsapi
Name: cdsapi
Version: 0.7.7
Summary: Climate Data Store API
Home-page: https://github.com/ecmwf/cdsapi
Author: ECMWF
Author-email: software.support@ecmwf.int
License: Apache 2.0
Location: /opt/miniconda3/envs/geo/lib/python3.12/site-packages
Requires: ecmwf-datastores-client, requests, tqdm
Required-by:Accompanying data
No response
Organisation
University of Maryland Baltimore County (UMBC), MD, USA
Activity