import cdsapi

c = cdsapi.Client()

c.retrieve(
    'efas-reforecast',
    {
        'format': 'grib',
        'product_type': 'ensemble_perturbed_reforecasts',
        'variable': 'river_discharge_in_the_last_6_hours',
        'model_levels': 'surface_level',
        'hyear': '2007',
        'hmonth': 'march',
        'hday': [
            '04', '07',
        ],
        'leadtime_hour': [
            '0', '12', '18',
            '6',
        ],
    },
    'efas_reforecast.grib')

import cdsapi
from datetime import datetime, timedelta



def get_monthsdays(start =[2019,1,1],end=[2019,12,31]):
# reforecast time index
start, end = datetime(*start),datetime(*end)
days = [start + timedelta(days=i) for i in range((end - start).days + 1)]
monthday = [d.strftime("%B-%d").split("-") for d in days if d.weekday() in [0,3] ] 

return monthday



if __name__ == '__main__':
c = cdsapi.Client()


# station coordinates (lat,lon)
COORDS = {
"Thames":[51.35,-0.45]
}

# select date index corresponding to the event 
MONTHSDAYS = get_monthsdays(start =[2019,7,11],end=[2019,7,11])

YEAR = '2007' 

LEADTIMES = ['%d'%(l) for l in range(24,1128,24)]

# loop over date index (just 1 in this case)
for md in MONTHSDAYS:

month = md[0].lower()
day = md[1]

# loop over station coordinates
for station in COORDS:

station_point_coord = COORDS[station]*2 # coordinates input for the area keyword

c.retrieve(
'cems-glofas-reforecast',
{
'system_version': 'version_2_2',
'variable': 'river_discharge_in_the_last_24_hours',
'format': 'grib',
'hydrological_model': 'htessel_lisflood',
'product_type': ['control_reforecast','ensemble_perturbed_reforecasts'],
'area':station_point_coord, 
'hyear': YEAR,
'hmonth': month ,
'hday': day ,
'leadtime_hour': LEADTIMES,
},
f'glofas_reforecast_{station}_{month}_{day}.grib')

## === retrieve GloFAS Seasonal Reforecast ===

## === subset South America/Amazon region === 

import cdsapi

if __name__ == '__main__':


    c = cdsapi.Client()

    YEARS  = ['%d'%(y) for y in range(1981,2021)]

    MONTHS = ['january', 'february', 'march', 'april', 'may', 'june', 'july', 'august', 'september', 'october', 'november', 'december']

    LEADTIMES = ['%d'%(l) for l in range(24,2976,24)]
    
    for year in YEARS:
        for month in MONTHS:

            c.retrieve(
                'cems-glofas-seasonal-reforecast',
                {
                    'system_version': 'version_2_2',
                    'variable':'river_discharge_in_the_last_24_hours',
                    'format':'grib',
                    'hydrological_model':'htessel_lisflood',
                    'hyear': year,
                    'hmonth': month,
                    'leadtime_hour': LEADTIMES,
                    'area': [ 10.95, -90.95, -30.95, -29.95 ]
                }, 
                f'glofas_seasonal_reforecast_{year}_{month}.grib')

import xarray as xr
import pandas as pd

parameter = "dis24"

ds = xr.open_dataset("glofas_historical.grib", engine="cfgrib",backend_kwargs={'time_dims':['time']})
df = pd.read_csv("GRDC.csv")

total = len(df)

rows = []
count = 0
for lon, lat, id in zip(df.long, df.lat, df.grdc_no):
    extracted = ds.sel(longitude=lon, latitude=lat, method="nearest")[parameter]
    df_temp = extracted.drop_vars(["surface"]).to_dataframe().reset_index()
    df_temp["grdc"] = str(id)
    df_temp = df_temp.set_index(["grdc", "time"])
    rows.append(df_temp)
    count += 1
    print(f"progress: {count/total*100} %")

out = pd.concat(rows)
out.to_csv("extracted.csv", index="grdc")

When transforming from lat/lon (source coordinates) to projected LAEA (target coordinates), you need to consider that the number of decimal places of the source coordinates affects the target coordinates precision:

An interval of 0.001 degrees corresponds to about 100 metres in LAEA.

An interval of 0.00001 degrees corresponds to about 1 metre in LAEA.

EFAS's x and  y coordinates, when converted from GRIB to NetCDF, are not projected coordinates but matrix indexes (i, j), It is necessary to download the upstream area static file that contains the projected coordinates and replace it in EFAS.