duplicate bad entry in the bgc-s index

[eeholmes]

I am not sure where to report this and it is a minor issue.

For wmo = 4902627, cyc = 18, the bgc-s index https://data-argo.ifremer.fr/argo_synthetic-profile_index.txt.gz file has a duplicate extra file row with bad data:

                                       file                date  latitude  \
30   meds/4902627/profiles/SR4902627_018.nc 2024-03-04 17:25:00    42.484   
31  meds/4902627/profiles/SR4902627_018D.nc 2024-02-23 23:52:00       NaN   

    longitude ocean  profiler_type institution  \
30    -60.269     A            836          ME   
31        NaN   NaN            836          ME   

                                          parameters parameter_data_mode  \
30  PRES PSAL TEMP DOXY CHLA BBP700 PH_IN_SITU_TOTAL             RRRRARR   
31              PRES PSAL TEMP DOXY PH_IN_SITU_TOTAL               RRRRR   

           date_update      wmo  cyc  
30 2025-08-27 21:56:11  4902627   18  
31 2025-08-27 21:56:00  4902627   18  

When using argopy (and user mode is standard), this WMO is filtered out because it chooses the 2nd (bad) row and then CHLA_DATA_MODE becomes "".

I wasn't able to test the whole index for other duplicates. Perhaps there are others.

MCVE Code Sample

import numpy as np
from argopy import DataFetcher as ArgoDataFetcher

wmo = 4902627
cyc = 18
region = [-70, -40, 20, 60, 0, 1000, "2024-03-01", "2024-04-01"]

# 1) Fetch in expert mode (so nothing gets dropped by standard-mode filtering)
fetcher = ArgoDataFetcher(
    mode="expert",
    ds="bgc",
    src="erddap",
    params=["CHLA", "PRES"],
).region(region)

# 3) Show what the Argo synthetic-profile index says for the same float/cycle
idx = fetcher.fetcher.indexfs
idx.query.wmo([wmo])
idf = idx.to_dataframe(completed=False)
idf[(idf["file"] == target_file) | ((idf["wmo"] == wmo) & (idf["cyc"] == cyc))]

shows output above

[gmaze]

Hi @eeholmes

Thanks for reporting this, my answer:

1. The 2nd row from the index is not a duplicate, it is a "Descending" profile, as it can be seen from the file path SR4902627_018D.nc , with a different timestamp but no values for latitude and longitude.

2. When fetching data from the Erddap only points where latitude and longitude are NOT nans are returned. This is the reason why this descent profile is not returned by the DataFetcher, not because it is a duplicate (which is not).

3. May be in 'expert' mode even those points should be return, what do you think ?

Hope this clarify things
G