-------------------------------------------------------------------------------- DRAFT ---------------------------------------------------------------------------------

This article applies to the ERA5 sub-daily datasets:

• ERA5 high resolution (HRES), containing
  
  • sub-daily atmospheric data (stream oper)
  • sub-daily ocean wave data (stream wave)
• ERA5 ensemble (EDA), containing
  
  • sub-daily atmospheric data (stream enda)
  • sub-daily ocean wave data (stream ewda)

This article does not apply to ERA5 monthly data. What streams are available? (Section 'Data organisation').

Before continuing with this article you might want to read:

• What is ERA5
• What is Ensemble forecasting
• How to download ERA5 data via the ECMWF Web API

Analysis and forecast

In the ERA5 data archive two types of data are available, 'analysis' (an) and 'forecast' (fc):

• An analysis, of the atmospheric conditions, is a blend of observations with a previous forecast. An analysis can only provide instantaneous parameters (parameters valid at a specific time, e.g temperature at 12:00), but not accumulated parameters, mean rates or min/max parameters.
• A forecast starts with an analysis at a specific time (the 'initialization time'), and a model computes the atmospheric conditions for a number of 'forecast steps', at increasing 'validity times', into the future. A forecast can provide instantaneous parameters, accumulated parameters, mean rates, and min/max parameters.

To see which parameters are available as analysis (an) and/or forecasts (fc)  see the ERA5 documentation, section 'Parameter listings'

Instantaneous, accumulated, mean rate and min/max parameters

Each parameter is classed as either 'instantaneous', 'accumulated', 'mean rate' or 'min/max', depending on the temporal properties of the parameter:

• Instantaneous parameters refer to a specific point in time , for example temperature at 12:00. For a list of available surface and single level instantaneous parameters see the ERA5 documentation, Table 2.
• Accumulated parameters are accumulations during a particular time period, for example precipitation between 17:00 and 18:00. For a list of available accumulated parameters see the ERA5 documentation, Table 3.
• Mean rate parameters are accumulated vales averaged over a particular time period, for example mean snowfall rate between 17:00 and 18:00, per second. For a list of available mean rate parameters see the ERA5 documentation, Table 4.
  
• Min/max parameters are the minimum or maximum 'instantaneous' value within a particular time period, for example minimum temperature between 17:00 and 18:00. For a list of available min/max parameters see the ERA5 documentation, Table  5.

Time and Step

'time' in analyses

Each analysis has a validity time, i.e. the time the data values refer to (not the time when the analysis was computed).

All validity times are in hours UTC.

Depending on the selected stream, ERA5 daily analysis data is available hourly (i.e. with validity time 00:00, 01:00, 02:00, ... , 23:00) or 3-hourly (i.e. with validity time 00:00, 03:00, 06:00,  ... , 21:00). See also ERA5 data documentation, 'Temporal resolution' and the ERA5 Catalogue, streams.

The concept of 'step' does not apply to analyses.

'time' in forecasts

ERA5 is largely a series of weather forecasts. Each forecast starts with the atmospheric conditions at a specific 'initialization time'. In ERA5 a new forecast is computed twice a day, with initialization time 06:00 and 18:00 UTC.

In the ERA5 data archive, for forecasts, 'time' specifies the initialization time.

'step' in forecasts

Each forecast computes the future atmospheric conditions, and at certain points during this computation a snapshot of the computed data is taken, post-processed, and stored in the ERA5 data archive. These snapshots are called 'steps'. In ERA5 there is a step every 1 hour or 3 hours, depending on the selected stream.

Steps are referenced in hours from the forecast initialization time. This is regardless of the step interval. For example, for time=06:00, step 3 is always at 09:00 (06:00+3h).

The step interval in ERA5 is:

• Every 1 hour in the HRES atmospheric (stream=oper), HRES wave (stream=wave) and EDA wave (stream=ewda) forecasts. Hence data is available with validity times 06:00 (step 0, initialization), 07:00 (step 1, i.e. initialization + 1h), 08:00 (step 2, i.e. initialization + 2h), and so on, and equivalent for the 18:00 initialization.
• Every 3 hours in the EDA atmospheric (stream=enda) forecasts. Hence data is available with validity times 06:00 (step 0, initialization), 09:00 (step 3, i.e. initialization +3h), 12:00 (step 6, i.e. initialization + 6h), and so on, and equivalent for the 18:00 initialization.

'step' and instantaneous, accumulated and min/max parameters

The interpretation of 'step' also depends on the parameter:

• Instantaneous parameters are valid at the time indicated by time+step. For example, temperature from the forecast at time=06:00, step=3, represents the temperature at 06:00 + 3h, i.e. at 09:00.

• Accumulated parameters are aggregated up to a Step Y, starting at the previous Step X:

• Mean rate parameters are similar to accumulated parameters, except that the quantities are averaged, instead of accumulated, up to a Step Y , from the previous Step X, so the units include "per second". For example:
  • In HRES atmospheric (hourly steps), mean rate precipitation at time=06:00, step=3 represents average precipitation up to [06:00 + 3h], starting from the previous step (Step 2), i.e. average precipitation in the 1-hour period 08:00 to 09:00, per second.
  • In EDA atmospheric (3-hourly steps), mean rate precipitation at time=06:00, step=3 represents average precipitation up to [06:00 + 3h], starting from the previous step (Step 0), i.e. average precipitation in the 3-hour period 06:00 to 09:00, per second.

• Min/max parameters  (parameters named 'Minimum/Maximum ... since previous post-processing' ) are similar to accumulated parameters, except that instead of accumulating, only the min/max value of the period from Step X to Step Y is archived.

Summary

The following table summarizes the different parameter types available in ERA5 from analysis and forecast:

Examples

See also How to download ERA5 data via the ECMWF Web API

Example 1:  hourly data, accumulated; e.g. total precipitation and evaporation

The resulting WebAPI Python script to download data:

#!/usr/bin/env python
from ecmwfapi import ECMWFDataServer
server = ECMWFDataServer()
server.retrieve({
    "class": "ea",                         # do not change
    "dataset": "era5",                     # do not change
    "expver": "1",                         # do not change
    "stream": "oper",                      # do not change
    "date": "2015-01-01/to/2015-01-31",
    "param": "228.128/182.128",            # parameter IDs (the .128 is optional)
    "levtype": "sfc",                      # tp and e are 2-dimensional fields ('surface field')    
    "type": "fc",                          # tp and e are accumulated parameters, hence available from forecast (fc), not analysis (an)
    "time": "06:00/18:00",                 # There are 2 forecasts per day, starting at 06:00 and 12:00
    "step": "1/2/3/4/5/6/7/8/9/10/11/12",  # From each forecast we retrieve steps 1(h) to 12(h). 
    "target": "output",                    # change this to your desired output file name
})

These specifications give you hourly 'total precipitation' data from 2015-01-01, 06:00 to 2015-02-01, 06:00. 

Discard the last 6 hours (2015-02-01, 00:01 to 06:00). 

The specification above result in data starting at 06:00 on the first day, so you are missing the first 6 hours (00:00 to 06:00). Retrieve the missing 6 hours from the last forecast of the previous day ('date':'2014-12-31', 'time':'18:00', 'step':'7/8/9/10/11/12')

Merge the two downloaded files

Example 2:  daily average, minimum and maximum of 2m temperature

In ERA5 there is no daily data, so you have to download sub-daily  data and aggregate to full days yourself.

The resulting WebAPI Python script to download data:

#!/usr/bin/env python
from ecmwfapi import ECMWFDataServer
server = ECMWFDataServer()
server.retrieve({
    "class": "ea",                         # do not change
    "dataset": "era5",                     # do not change
    "expver": "1",                         # do not change
    "stream": "oper",                      # do not change
    "date": "2015-01-01/to/2015-01-31",
    "param": "167.128/201.128/202.128",    # parameter IDs (the .128 is optional)
    "levtype": "sfc",                      # tp and e are 2-dimensional fields ('surface field')    
    "type": "fc",                          # tp and e are accumulated parameters, hence available from forecast (fc), not analysis (an)
    "time": "06:00/18:00",                 # There are 2 forecasts per day, starting at 06:00 and 12:00
    "step": "1/2/3/4/5/6/7/8/9/10/11/12",  # From each forecast we retrieve steps 1(h) to 12(h). 
    "target": "output",                    # change this to your desired output file name
})
#!/usr/bin/env python
from ecmwfapi import ECMWFDataServer
server = ECMWFDataServer()
server.retrieve({
    "class": "ea",                         # do not change
    "dataset": "era5",                     # do not change
    "expver": "1",                         # do not change
    "stream": "oper",                      # do not change
    "date": "2015-01-01/to/2015-01-31",
    "param": "167.128/201.128/202.128",   # '2 metre temperature' (2t)

'Maximum temperature at 2 metres since previous post-processing'

'Minimum temperature at 2 metres since previous post-processing' 


    "levtype": "sfc",
    "type": "fc",
    "time": "06:00:00/18:00:00",
    "step": "1/2/3/4/5/6/7/8/9/10/11/12",
    "target": "output",
})

ERA5 analysis is available at each full hour, so you need:

• type : an
• date : 2015-01-01/to/2015-01-31
• 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

This results in hourly data for 2t. Then calculate the daily average.

The resulting WebAPI Python script to download data:

You could obtain the hourly 2m temperature (see Example 3), and then identify the daily maximum and minimum of the hourly values.

Alternatively you can use the parameters Maximum temperature at 2 metres since previous post-processing' (mx2t) and Minimum temperature at 2 metres since previous post-processing (mn2t):

The parameters named 'min/max since previous post-processing' are available from forecasts.

Is this a valid use case? Why not simply use analysis? One could, but an gives 2t at each full hour, while the min/max could be at a model step in between.

ERA5 forecasts are initialized at 06:00 and 18:00. Each forecast step is one hour, so you need from each 06:00 forecast and from each 18:00 forecast, mx2t and mn2t for first the first 12 forecast steps:

• type : fc
• date : 2015-01-01/to/2015-01-31
• time : 06:00/18:00
• step : 1/2/3/4/5/6/7/8/9/10/11/12

This results in hourly data from January 01, 06:00 to February 01, 06:00 (discard the last 6 hours).

You still need the first six hours of the first day, get these from the last forecast of the previous day:

• type : fc
• date : 2014-12-31
• time : 18:00
• step : 7/8/9/10/11/12

Sample Python script for the ECMWF WebAPI:

#!/usr/bin/env python
from ecmwfapi import ECMWFDataServer
server = ECMWFDataServer()
server.retrieve({
    "class": "ea",
    "dataset": "era5",
    "stream": "oper",
    "expver": "1",
    "date": "2015-01-01/to/2015-01-31",
    "type": "fc",
    "levtype": "sfc",
    "param": "201.128/202.128",             # 'Maximum temperature at 2 metres since previous post-processing' and 'Minimum temperature at 2 metres since previous post-processing'
    "time": "06:00:00/18:00:00",            # 2 forecasts per day, starting at 06:00 and 12:00
    "step": "1/2/3/4/5/6/7/8/9/10/11/12",   # For each forecast 18 one-hour steps are available. Here we use only steps 1 to 12.
    "grid": "0.30/0.30",
    "area": "60/-10/20/50",
    "target": "output",                     # change this to your output file name
})

The above script retrieves Min and Max 2m temperature since previous post-processing, from 2015-01-01 06:00 to 2015-02-01 06:00, as one-hour periods. To identify the daily max/min, find the the max/min of these one-hour maxima/minima.

See also

What is ERA5

ERA5 data documentation

How to download ERA5 data via the ECMWF Web API

Related articles