How to Download Faster

I am downloading a lot of data from the ERA5 dataset. I am downloading 20 years of temperature/precipitation data for a few thousand small regions. How do I do this quickly? It seems that the download time for each region is around 5-10 minutes, so if I have a few thousand regions, this will never finish. So are there any tips on how to use the API to download the data in a more efficient way?


Hi Joseph,

Is there any chance you group those small regions into some bigger ones so that you can download the data in less requests and then you extract the data from the downloaded files?



Thanks for the reply. So requesting the maximum amount data that my computer can handle reasonably well will speed up the entire process? Or are there any specifications that will dramatically slow down the process? For example if I request multiple variables (temp and radiation) in one query, or request multiple years of data will it actually be worse than requesting one variable at a time and one year at a time?



Hi Joe,

The recommendation is:

  • For daily data, make one request per month
  • For monthly data, make one request per year

If you are asking for too much, you may have to split the period to smaller ones. For each request, request as many variables as possible.



Hi Xiabo,

What is the recommendation for hourly data?


Hi Niels,

Hourly data falls in the daily data category.



Thanks. I tried that and found out it takes ca. 15 minutes to download one month of hourly data for 4 variables for one location. As this may be of help to others, here is my code:

import cdsapi
c = cdsapi.Client()

for month in range(1,13):
if (month<10):
month2 = '0' + str(month)
month2 = str(month)

'variable': [
'2m_dewpoint_temperature', '2m_temperature', 'surface_pressure',
'surface_solar_radiation_downwards', 'total_precipitation',
'area': '50.8/5.2/50.8/5.2',
'year': '2011',
'month': month2,
'day': [
'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',
'time': [
'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',
'format' : 'netcdf',
'/Users/.../data/weather/my-file' + month2 + '.nc'

I am using the 'area' variable to request only one location – an area collapsed to one point at 50.8 lat and 5.2 lon.

Hi Niels,

Performance is related to how busy the CDS is. gives you information about the current queue.

I hope this helps.


I have a similar problem with very slow downloading speed. Basically, the size of data downloaded via CDS API is 2.1MB, but it takes 45 minutes. This is unusual because  I can download data of more than 2GB when using the ecmwfapi for that amount of time. Additionally, I found that if the data size beyond 2.1MB or 20MB.  For example, I increased the area to for the area size by increasing longitude 9 degrees. Error always occur:



Exception: the request you have submitted is not valid. One or more variable sizes violate format constraints.

But with small data size, I can download data. 

Below is my code for retrieving ERA5 data:

code start

import calendar
import cdsapi
server = cdsapi.Client()

def retrieve_era5():
A function to demonstrate how to iterate efficiently over several years and months etc
for a particular era5_request.
Change the variables below to adapt the iteration to your needs.
You can use the variable 'target' to organise the requested data in files as you wish.
In the example below the data are organised in files per month. (eg "era5_daily_201510.grb")

yearStart = 1998
yearEnd = 1998
monthStart = 1
monthEnd = 1
for year in range(yearStart, yearEnd + 1):
    Year = str(year)
    for month in range(monthStart, monthEnd + 1):
        Month = str(month)
        # startDate = '%04d-%02d-%02d' % (year, month, 1)
        numberOfDays = calendar.monthrange(year, month)[1]
        Days = [str(x) for x in list(range(1, numberOfDays + 1))]
        # lastDate = '%04d-%02d-%02d' % (year, month, numberOfDays)
        target = "" % (year, month)
        # requestDates = (startDate + "/" + lastDate)
        era5_request(Year, Month, Days, target)

def era5_request(Year, Month, Days, target):
An ERA era5 request for analysis pressure level data.
Change the keywords below to adapt it to your needs.
(eg to add or to remove levels, parameters, times etc)
{'product_type': 'reanalysis',
'format': 'netcdf',
'variable': ['specific_humidity', 'u_component_of_wind', 'v_component_of_wind'],
'year': Year,
'month': Month,
'day': Days,
'pressure_level': ['300', '350', '400','450', '500', '550', '600', '650', '700','750', '775', '800','825', '850', '875','900', '925', '950','975', '1000'],
'time': ['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': [0, 100, -1, 101],},

if name == 'main':

code end

This code is just to do things small at first, try to download specific_humidity, u_component_of_wind, v_component_of_wind from 1998-1-1 to 1998-1-31, temporal resolution: 1 hour; spatial resolution: 0.25° x 0.25°; pressure levels: 300 hpa to 1000 hpa. Area :1°S to 0, 100°E to 101°E.

Below is the picture showing the results of running this code:

Below is the picture showing that downloading data by ecmwfapi, basically, 23minutes retrieving 2.18GB data:


When changing the 'area': [0, 100, -1, 101] to 'area': [0, 100, -70, 101] it works fine. But when changing the 'area': [0, 100, -1, 101] to  'area': [0, 100, -70, 130]. error occurred. 

KeyboardInterrupt or Exception: the request you have submitted is not valid. One or more variable sizes violate format constraints.

I thought this might be due to limit, I did the same through the website, it shows the data size is under the limit. 

So I do not know what is going on?  

Hi Ted,

Could you share the script which did not run?

Thank you,


Hi Xiaobo

Problem solved, it turns out that I reached out the limit., which I guess is 10GB. So now I just reterive that data day by day.

Thank for letting me know Ted.

Hi Xiaobo,

I am conducting probabilistic yield forecasting for rooftop PV systems. But I found it extremely time-consuming to download the ensemble forecasting of ssrd for my target PV site. It takes 1.5 hours to download the 50 ensemble of hourly ssrd forecasting for the target site in one day (around 350kb) . I have to download the ensemble forecasting for several years. Could you please give me some suggestions? My request is shown below:

server = ECMWFService("mars",    url="",

    "class": "od",
    "date": "20190101/to/20190102",
    "expver": "1",
    "levtype": "sfc",
    "number": "1/to/50",
    "param": "169.128",
    "area": "51.7/5.2/51.2/5.7",
    "step": "0/1/2/3/4/5/6/7/8/9/10/11/12/13/14/15/16/17/18/19/20/21/22/23",
    "stream": "enfo",
    "time": "0000",
    "type": "pf",

Hi Bin,

Since you are not accessing data on the CDS (Climate Data Store), could you raise a ticket at our support portal

Thank you,