Clarification on the accumulation period of ERA5 accumulated variables retrieved at 3-hourly intervals

Datasets and Usage
1

Dear ECMWF Support Team,

I am working with ERA5 reanalysis data and would like to clarify the temporal interpretation of the accumulated variables retrieved from the CDS.

I am using the dataset:

reanalysis-era5-single-levels

with the following request settings:

  • Product type: reanalysis
  • Data format: NetCDF
  • Time: 00:00, 03:00, 06:00, …, 21:00 UTC
  • Spatial resolution: 0.25° × 0.25°
  • Region: 25°N–10°S, 70°E–105°E

Among the variables, I downloaded the following accumulated parameters:

  • total precipitation (tp)
  • evaporation (e)
  • surface solar radiation downwards (ssrd)
  • surface thermal radiation downwards (strd)
  • surface net solar radiation (ssr)
  • surface net thermal radiation (str)
  • surface latent heat flux (slhf)
  • surface sensible heat flux (sshf)

The downloaded NetCDF files contain a valid_time coordinate at 3-hourly intervals. For example:

2016-11-01 00:00
2016-11-01 03:00
2016-11-01 06:00
2016-11-01 09:00
...

The accumulated variables are identified in the file metadata as:

GRIB_dataType = "fc"
GRIB_stepType = "accum"

My question is about the exact accumulation period represented by each value.

For example, for tp at:

2016-11-01 03:00 UTC

does this value represent:

(A) precipitation accumulated over the preceding 3 hours, i.e. approximately 00:00–03:00 UTC,

or

(B) precipitation accumulated over a 1-hour forecast step, with the 03:00 valid time corresponding to approximately 02:00–03:00 UTC?

In other words, does requesting:

"time": ["00:00", "03:00", "06:00", ...]

cause the CDS to return 3-hour accumulated values, or does it simply select the corresponding forecast-valid times of the ERA5 HRES accumulated fields, whose accumulation period may remain 1 hour?

This distinction is important for my application because I am using these data as atmospheric forcing for a ROMS ocean model. If the values are 1-hour accumulations, I need to construct 3-hour accumulated fluxes before converting them to mean fluxes (W m⁻² or kg m⁻² s⁻¹). If they are already 3-hour accumulations, I should not perform an additional temporal summation.

I would therefore greatly appreciate it if you could clarify:

  1. What is the exact accumulation period of an accumulated variable such as tp when retrieved from reanalysis-era5-single-levels at 3-hourly intervals?
  2. What are the corresponding time, step, and stepRange values internally used by the CDS for such a request?
  3. Does the temporal spacing specified in the CDS request (00, 03, 06, ...) affect the accumulation period, or does it only determine which valid times are returned?
  4. If I need a true 3-hour accumulation ending at 03:00 UTC (00:00–03:00 UTC), what is the recommended way to retrieve it from ERA5?
  5. For variables such as ssrd, strd, ssr, str, slhf, and sshf, should the same interpretation be applied?

If possible, could you provide a concrete example for tp at 03:00 UTC showing the exact forecast initialization time, step/stepRange, and accumulation interval represented by the value?

Thank you very much for your help.

Best regards,

Peng