Versions Compared

Key

  • This line was added.
  • This line was removed.
  • Formatting was changed.
Why I cannot access the most recent data?

(lightbulb) Because of the agreed data policy with the data providers, the S2S realtime type of forecast can be accessed only with 3 weeks delay. There should not be any such restriction for the reforecasts.

...

What are reforecasts

...

?

(lightbulb)

...

Reforecasts are forecasts run using the same model version as

...

real-time forecast for a number of past dates

...

The way to produce reforecasts is not unique

...

. More details in A brief description of reforecasts

How to access S2S data efficiently?

(lightbulb) In order to retrieve data efficiently users should follow instructions about S2S reforecasts retrieval efficiency.

...

Do you have any examples of S2S data retrieval scripts?

(lightbulb) Find links here to efficient retrieval scripts for different S2S datasets via ECMWF Web API.

...

Do you provide analysis in S2S?

(lightbulb) The S2S database does not contain the analysis used to initialize the S2S models. However, for some models (e.g ECMWF) step 0 is available (see details below), but only for instantaneous variables. For some models (e.g. ECMWF),  step 0 of the control forecast corresponds to the unperturbed analysis. 

If you are interested in re-forecasts, most re-forecasts are initialized either from JRA, ERA or NCEP re-analysis which can be available from other public datasets. 

...

What is the starting step of S2S parameters?

(lightbulb)  The instantaneous variables are archived from step 0, except for some centres (e.g. NCEP or BoM) who start at step=24 because they don't archive step 0 (step 0 has the same value as the re-analysis). Because of that the data at step 0 and step 24 are different and step=0 corresponds to the initial conditions. 

For accumulated variables (e.g. total Precipitation), step 0 is usually not archived because its value is equal to 0 and step=24 corresponds to the accumulation during the first 24 hours.

For daily averaged variables, there is no step 0.

For MJO RMMS, the calculation for all models starts from step=24 (first common time step for all the models). It uses instantaneous winds at 850 and 200 hPa at step=24h and OLR accumulated over the first 24 hours 

...

Is there an issue with accumulated fields at day 16 in IFS (ECMWF) model?

(lightbulb) Read the explanation in Issues with data.

...

  • In this case reforecasts are produced at the same time as the real-time forecasts.
  • Each reforecast refers to a real-time forecast.

Main Concepts

Realtime Date

Info
  • It is the date of the real-time ensemble forecast to be calibrated.
  • If the purpose of the reforecasts is to create a model climatology, then the date of the real-time ensemble forecast is not relevant. 
  • In the MARS system this is expressed with the keyword "date".

Model Version Date

Info
  • It is the version of the model  used to produce the reforecasts.
  • Because the version of the model is time-stamped with a date, the term "Model Version Date" is used instead.
  • It is very important that the version of the model used to produce the reforecasts AND the version of the model used to produce the real-time forecast are the same. One CANNOT calibrate a real-time forecast with reforecast that has been produced by a different version of the model.  
  • In the MARS system, this is expressed with the keyword "date".

Hindcast Date

Info
  • It refers to a start date for any of the past dates of the reforecast.
  • In the MARS system, this is expressed with the keyword "hdate".

Examples

...

reforecast

configuration

...

realtime

date

...

model version

date

...

reforecast

length

...

reforecast

frequency

...

number of hindcast

dates available for a

model version date

...

hdate=19810101/19810106/19810111/19810116/19810121/19810126/../

20131201/20131206/20131211/20131216/20131221/20131226

...

...