Download CDS ERA5 data using R

Kevin_Marsh · 1 April 2021 10:33

Users can download CDS data using the ecmwfr package in R with “cds” as the "service" although it is not supported by C3S. In the examples below, "KEY" and "UID" are your CDS API key and UID respectively.

PLEASE NOTE THAT SINCE THE MOVE TO THE NEW CDS IN OCTOBER 2024, "UID" IS NO LONGER REQUIRED BY THE CDS API.
-YOU SHOULD THEREFORE MAKE SURE THAT YOUR R PACKAGE TAKES THIS CHANGE INTO ACCOUNT AND IS UP TO DATE

Please have a look at the following example to download ERA5 2m temperature for a selected area in netCDF format:

library(ecmwfr)

cds.key <- "Insert_your_CDS_API_KEY_here"
wf_set_key(user = "Insert_your_CDS_UID_here", key = cds.key, service = "cds")

request <- list(
dataset_short_name = "reanalysis-era5-single-levels",
product_type   = "reanalysis",
format = "netcdf",
variable = "2m_temperature",
year = "2016",
month = "08",
day = "16",
time = c("00:00", "01:00", "02:00", "03:00", "04:00", "05:00", "06:00", "07:00", "08:00", "09:00", "10:00", "11:00", "12:00", "13:00", "14:00", "15:00", "16:00", "17:00", "18:00", "19:00", "20:00", "21:00", "22:00", "23:00"),
# area is specified as N, W, S, E
area = c(50, -20, 30, 20),
target = "download_e5_single.nc"
)

file <- wf_request(user = "Insert_your_CDS_UID_here",
                     request = request,
                     transfer = TRUE,
                     path = "~",
                     verbose = TRUE)

This would return the data to a file called download_e5_single.nc.

The code below retrieves ERA5pressure level temperature data at 850hpa for a selected area in netCDF format:

library(ecmwfr)

cds.key <- "Insert_your_CDS_API_KEY_here"
wf_set_key(user = "Insert_your_CDS_UID_here", key = cds.key, service = "cds")

request <- list(
dataset_short_name = "reanalysis-era5-pressure-levels",
product_type   = "reanalysis",
format = "netcdf",
variable = "temperature",
pressure_level = "850"
year = "2016",
month = "08",
day = "16",
time = c("00:00", "01:00", "02:00", "03:00", "04:00", "05:00", "06:00", "07:00", "08:00", "09:00", "10:00", "11:00", "12:00", "13:00", "14:00", "15:00", "16:00", "17:00", "18:00", "19:00", "20:00", "21:00", "22:00", "23:00"),
# area is specified as N, W, S, E
area = c(50, -20, 30, 20),
target = "download_e5_pressure.nc"
)

file <- wf_request(user = "Insert_your_CDS_UID_here",
                     request = request,
                     transfer = TRUE,
                     path = "~",
                     verbose = TRUE)

Malcolm_Mistry · 25 August 2021 16:12

Thanks. This really helps. Just wondering, is there a way to pass the 'area' argument (to subset the data by bounding box) for those datasets that are not on a regular lon lat grid. Example UERRA dataset which is on a Lambert Conformal Conic grid. For such datasets, the 'Geographical area' option does not appear when selecting/filtering the variable. I see there is a related question here (though this question in fact is for another dataset that is already on a regular lon lat grid, yet not available to subset the bounding box). I am not sure why it would be difficult to transform the user defined regular lon lat coordinates to the dataset's native projection on the back-end, i.e. the Show API request should be able to do this in the background and display the 'area' coordinates for subsetting.

Kevin_Marsh · 31 August 2021 21:21

Hi Malcolm,

No, at this time I don't think it is possible to perform an area subset in the way you describe using the CDS API, but it is a good suggestion for a future development,

Thanks,

Kevin

Malcolm_Mistry · 1 September 2021 03:37

Hi Kevin,

Thanks again!

Rgds

Malcolm

Amaury_Alloyer · 26 April 2024 13:59

Hi Kevin, thank you, that helped a lot. Do you know if we ca use “ecmwfr” R package to download cmip6 projections ?
Best,

Kevin_Marsh · 6 June 2024 14:48

Hi, ‘ecmwfr’ is a 3rd party package, and I’m afraid I don’t know if it would work with CDS cmip6 projection data.
Kevin

Charles_Ramos · 29 July 2024 07:13

Hello Kevin. Since a new CDS has been launched, new credentials are already released. Should I use the 32-character ID for the user field instead of the usual 6-digit ID?

I’ve tried it but there is an error.

Kevin_Marsh · 30 July 2024 10:28

Hi Charles,

if you login to CDS-beta and go to
https://cds-beta.climate.copernicus.eu/how-to-api
you should see your CDS-beta login credentials (note there is no longer a ‘UID’).

I do not know if the ecmwfr package will handle these details correctly.

Koen_Hufkens · 21 August 2024 09:31

Hi Charles, an update to the package is in the works. For now this will only use the old CDS/ADS setup. I’m roughly a week away from a formal release.

Koen_Hufkens · 15 December 2024 09:08

To update this thread, a new version of {ecmwfr} for R is available and inline with the current API here:

Charles_Ramos · 18 December 2024 02:34

Thanks for this, Koen!

Alexander_Kaune · 12 December 2025 12:55

Thanks Koen, this new version of {ecmwfr} for R is still valid? For the following code in R I’m getting “ Test failed: Data identifier reanalysis-era5-land is not found in Web API, CDS or ADS datasets.\n Or your login credentials do not match your request.”

Minimal test script for CDS connection

library(ecmwfr)

Set key with validation disabled (if available)

tryCatch({

Try to set key with current .cdsapirc contents

config_path ← “C:/Users/alexk/.cdsapirc”
if(file.exists(config_path)) {
lines ← readLines(config_path, warn = FALSE)
key_line ← grep(“^key:”, lines, value = TRUE)
if(length(key_line) > 0) {
api_key ← trimws(sub(“^key:\s*”, “”, key_line))

  # Try without validate parameter first
  wf_set_key(key = api_key, service = "cds")
  print("✅ Key set successfully")
}

}
}, error = function(e) {
print(paste(“Setting key failed:”, e$message))
})

Test request

test_request ← list(
dataset_short_name = “reanalysis-era5-land”,
product_type = “reanalysis”,
variable = “2m_temperature”,
year = “2020”,
month = “01”,
day = “01”,
time = “00:00”,
area = c(10, -86, 9, -85), # Small area in Costa Rica
format = “netcdf”,
target = “test.nc”
)

print(“Testing download…”)
tryCatch({
result ← wf_request(request = test_request, time_out = 300, path = temp_dir)
print(“ Test download successful!”)
}, error = function(e) {
print(paste(“ Test failed:”, e$message))
})

Koen_Hufkens · 14 December 2025 18:38

Your request passes on my end, but required formatting the quotation marks around entries correctly. Always use the rstudio plugin to convert the python API requests to the R version, manually editting things often leads to hard to trace errors.