There are many situations where a user is only interested in a subset of the dataset spatial domain.

For example, when comparing modelled river flow against observations, it is reasonable to be able to extract the time-series at those point coordinates rather than dealing with many GB of data. Similarly, when focusing on a specific catchment it is likely that you want to deal with only that part of the spatial domain.

In summary, there are two operations of data size reduction that are very popular on CEMS-Flood datasets, area cropping and time-series extraction.

There are two ways to perform those operations:

  • Remotely - Using the CDS API to perform the operation remotely on the CDS compute nodes and retrieve only the reduced data.
  • Locally - Using the CDS API to retrieve the entire data and perform the operation locally.

This section provides scripts for both cases and for both CEMS-Flood products, GloFAS and EFAS.

Set up a Python environment

If you have not done it yet, create a Python virtual environment.

Activate the conda environment and install the additional Python package https://corteva.github.io/rioxarray

conda install rioxarray


Prepare and retrieve data (for local processing)

For the following exercises on extracting time series on the local machine, we are going to use the latitude and longitude coordinates from a tiny subset of the GRDC dataset.

Copy the content of the code block into an empty file named "GRDC.csv", the file should reside in your working folder.

GRDC dataset 
grdc_no,wmo_reg,sub_reg,river,station,country,lat,long,area,altitude
6118010,6,18,CLAIE,SAINT-JEAN-BREVELAY,FR,47.824831,-2.703308,137.0,99.98
6118015,6,18,YVEL,LOYAT (PONT D129),FR,47.993815,-2.368849,315.0,88.26
6118020,6,18,"ARON, RUISSEAU D'",GRAND-FOUGERAY (LA BERNADAISE),FR,47.71222,-1.690835,118.0,68.0
6118025,6,18,"CANUT, RUISSEAU DE",SAINT-JUST (LA RIVIERE COLOMBEL),FR,47.775309,-1.979609,37.0,80.22
6118030,6,18,AFF,PAIMPONT (PONT DU SECRET),FR,47.981631,-2.143762,30.2,119.01
6118050,6,18,COET-ORGAN,QUISTINIC (KERDEC),FR,47.904164,-3.201265,47.7,94.42
6118060,6,18,EVEL,GUENIN,FR,47.899928,-2.975167,316.0,95.16
6118070,6,18,STER-GOZ,BANNALEC (PONT MEYA),FR,47.906833,-3.752172,69.7,85.08
6118080,6,18,MOROS,CONCARNEAU (PONT D22),FR,47.882934,-3.875375,20

Then, retrieve the following datasets into the same working folder.

Get the GloFAS Historical data 
import cdsapi

c = cdsapi.Client()

c.retrieve(
    'cems-glofas-historical',
    {
        'variable': 'river_discharge_in_the_last_24_hours',
        'format': 'grib',
        'hydrological_model': 'lisflood',
        'product_type': 'intermediate',
        'hyear': '2021',
        'hmonth': 'january',
        'hday': [
            '01', '02', '03',
            '04', '05', '06',
            '07', '08', '09',
            '10', '11', '12',
            '13', '14', '15',
            '16', '17', '18',
            '19', '20', '21',
            '22', '23', '24',
            '25', '26', '27',
            '28', '29', '30',
            '31',
        ],
        'system_version': 'version_3_1',
    },
    'glofas_historical.grib')


Get the EFAS reforecast data 
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')


EFAS

Removal of subsetting for EFAS

An issue has been identified with the EFAS sub-region extraction tool, whereby it serves data that is not correctly located on the river network. The sub-region extraction tool has therefore been removed from the EFAS CDS entries, and any area specified in cdsapi requests will return the entire domain .

Data previously downloaded using this tool should be disregarded.

For more information please see EFAS-Known Issues

Coordinates precision

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.

Remote processing

Time series extraction:

Extract timeseries: EFAS medium-range historical 

 


Area cropping:

 

  Area cropping: EFAS medium-range historical
 

 

  
## === retrieve EFAS Seasonal Forecast ===
 
## === subset Switzerland region ===
 
import cdsapi
 
 
if __name__ == '__main__':
    c = cdsapi.Client()
 
    YEARS  = ['%d'%(y) for y in range(2020,2022)]
 
 
    MONTHS = ['%02d'%(m) for m in range(1,13)]
 
    LEADTIMES = ['%d'%(l) for l in range(24,5160,24)]
     
    for year in YEARS:
 
        for month in MONTHS:
             
            c.retrieve(
                'efas-seasonal',
                {  
                'variable': 'river_discharge_in_the_last_24_hours',
                'format': 'grib',
                'model_levels': 'surface_level',
                'year': year,
                'month': '12' if year == '2020' else month,
                'leadtime_hour': LEADTIMES,
                'area': [ 47.9, 5.8, 45.7, 10.6 ]
 
                },
                f'efas_seasonal_{year}_{month}.grib')

 


Local processing
Time series extraction:
Important - Download upstream area
EFAS 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 that contains the projected coordinates and replace them in EFAS, as described in the code block below.


 

  Extract timeseries: EFAS medium-range Reforecast
 

 

  
import xarray as xr
import pandas as pd
from pyproj import Transformer,CRS

parameter = "dis06"

ds = xr.open_dataset("efas_reforecast.grib", engine="cfgrib")
df = pd.read_csv("GRDC.csv")
uparea = xr.open_dataset("ec_uparea4.0.nc") # the upstream area

# replace x, y
ds["x"] = uparea["x"]
ds["y"] = uparea["y"]

# define reprojection parameters
laea_proj = CRS.from_proj4("+proj=laea +lat_0=52 +lon_0=10 +x_0=4321000 +y_0=3210000 +ellps=GRS80 +units=m +no_defs")
transformer = Transformer.from_crs('epsg:4326', laea_proj, always_xy=True)

total = len(df)

rows = []
count = 0
for lon, lat, id in zip(df.long, df.lat, df.grdc_no):
    x1, y1 = transformer.transform(lon, lat)
    extracted = ds.sel(x=x1, y=y1, number = 1, method="nearest")[parameter]
    df_temp = extracted.drop_vars(["surface", "number"]).to_dataframe().reset_index()
    df_temp["grdc"] = str(id)
    df_temp = df_temp.drop("step", axis=1)
    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")

 


Area cropping:

 

  Area cropping: EFAS medium-range Reforecast
 

 

  
import xarray as xr
import rioxarray as rio
from pyproj import Transformer, CRS
import numpy as np

# Rhine's basin bounding box coordinates in WGS84.
coords = [5.450796, 11.871059, 46.296530, 50.972204] # W,E,S,N

# source/target reference systems proj
EPSG_4326 = '+proj=longlat +datum=WGS84 +no_defs'
EPSG_3035 = '+proj=laea +lat_0=52 +lon_0=10 +x_0=4321000 +y_0=3210000 +ellps=GRS80 +units=m +no_defs' # EFAS 

# read EFAS reforecast and the upstream area
ds = xr.open_dataset("efas_reforecast.grib", engine="cfgrib")
uparea = xr.open_dataset("ec_uparea4.0.nc")

# replace x, y coordinates
ds["x"] = uparea["x"]
ds["y"] = uparea["y"]

# add reference system to EFAS dataset
ds = ds.rio.write_crs(EPSG_3035)


# Function to convert 4 coordinates into a Polygon
def bbox_from_wesn(coords, s_srs, t_srs):

    w, e, s, n = coords

    transformer = Transformer.from_crs(s_srs, t_srs, always_xy=True)

    # topleft
    topleft =  transformer.transform(w,n)
    #bottomleft
    bottomleft = transformer.transform(w, s)
    #topright
    topright = transformer.transform(e,n)
    # bottomright
    bottomright = transformer.transform(e,s)

    bbox = [
        {
            'type': 'Polygon',
            'coordinates': [[
                topleft, bottomleft, bottomright, topright, topleft
            ]]
        }
    ]

    return bbox

bbox = bbox_from_wesn(coords, s_srs=EPSG_4326, t_srs=EPSG_3035)

ds_clipped = ds.rio.clip(bbox)

ds_clipped.to_netcdf("efas_reforecast_rhyne.nc")


 


GloFAS
Remote processing
Time series extraction:

 

  Extract timeseries: GloFAS Medium-range Reforecast
 

 

  
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')

 


Area cropping:

 

  Area cropping: GloFAS Medium-range reforecast 
 

 

  
## === retrieve GloFAS Medium-Range Reforecast ===

## === subset India, Pakistan, Nepal and Bangladesh region === 


import cdsapi
from datetime import datetime, timedelta


def get_monthsdays():

    start, end = datetime(2019, 1, 1), datetime(2019, 12, 31)
    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

MONTHSDAYS = get_monthsdays()

if __name__ == '__main__':
    c = cdsapi.Client()
    
    # user inputs
	BBOX = [40.05 ,59.95, 4.95, 95.05] # North West South East
    YEARS  = ['%d'%(y) for y in range(1999,2019)]
    LEADTIMES = ['%d'%(l) for l in range(24,1128,24)]
    
    # submit request
    for md in MONTHSDAYS:

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

        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',
				'area': BBOX,# < - subset
                'hyear': YEARS,
                'hmonth': month ,
                'hday': day ,
                'leadtime_hour': LEADTIMES,
            },
            f'glofas_reforecast_{month}_{day}.grib')

 



 

  Area cropping: GloFAS Seasonal Forecast
 

 

  
## === retrieve GloFAS Seasonal Forecast ===

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

import cdsapi


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

    YEARS  = ['%d'%(y) for y in range(2020,2022)]


    MONTHS = ['%02d'%(m) for m in range(1,13)]

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

        for month in MONTHS:
            
            c.retrieve(
                'cems-glofas-seasonal',
                {   
                'variable': 'river_discharge_in_the_last_24_hours',
                'format': 'grib',
                'year': year,
                'month': '12' if year == '2020' else month,
                'leadtime_hour': LEADTIMES,
				'area': [ 10.95, -90.95, -30.95, -29.95 ]

                },
                f'glofas_seasonal_{year}_{month}.grib')

 



 

  Area cropping: GloFAS Seasonal Reforecast
 

 

  
## === 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')

 


Local processing
Time series extraction:

 

  Extract timeseries: GloFAS historical
 

 

  
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")

 


Area cropping:

 

  Area cropping: GloFAS historical
 

 

  
import xarray as xr
 
# Rhine's basin bounding box
bbox = [50.972204, 46.296530, 5.450796, 11.871059]  # N,S,W,E
 
ds = xr.open_dataset("glofas_historical.grib", engine="cfgrib")
 
ds_cropped = ds.sel(
    longitude=slice(bbox[2], bbox[3]), latitude=slice(bbox[0], bbox[1])
)
 
ds_cropped.to_netcdf("glofas_historical_cropped.nc")