Dear Mrs Owlad. to have better understanding about ECMWF reanalysis which is totally different from analysis one, please kindly download the file that I attached.BeST WISHES.

Both, analysis and reanalysis products are based on a model. In the analysis, the model changes with time, as the model is improving (hopefully  ;-) ). In case of scientific need, sometimes a long period is re-run again, based on only one model and therefore the quality of the model output is the same, differences in the output are due to different observations (and to some part also changing quality in observations).

Resolution is also different for these products, the analysis data are higher resolved, as these are used to start / initialise the forecast.

Cheers

Many thanks Tim for usefull information

If you are looking for "analysis" data for a wide area and extended period of time, you may have problems finding them.

Keeping that in mind, I believe that the reanalysis data is an acceptable "substitution".

Of course, that is just my opinion.

Greetings

Hi Athanasios, no, I think you are wrong. Well known re-analysis products have some limited period like ERA40 (40 years) or ERAinterim or the old ERA15 (15 years) ... analysis data cover the whole period of the centre (e.g. ECMWF), but are based on varying model versions.

Re-Analysis data need the same input to be produced, as analysis data.

So which data do you prefer to work with? Analysis or Reanalysis? Which one is closer to observation and more reliable?

Hi Tim.

Maybe I wasn't very clear. Reanalysis or interim are not analysis. No question there. But if you cannot find the required analysis data, you could use reanalysis. Of course you have to be clear and state that you are working with reanalysis data.

If you still believe that I’m wrong, I would be really glad if you could elaborate on the matter. It would be beneficial for me and ms Elaheh Owlad too.

Dear Elaheh, dear Athanasios, 

as I tried to explain above: analysis data are produced on the fly each day. These 4dim data are model output: initialised by a 3dim atmosphere from the day before and running a model simulation including all information of the weather observations of the current day.

This model (to get the analysis data) is always under development and a frequent update of the underlying model to get this 4dim data is done. Therefore the quality of the time series, e.g. of the last 50 years, varies with time, as we can assume, the model got significantly improvements over the last 50 years. Obvious changes are  e.g. the higher horizontal and vertical resolution with time.

Re-Analysis data are produced the same way with a frozen model - so there is no change in time, the resolution is not varying with time, and mainly the resolution is lower than the resolution of the analysis data produced at the time when the re-analyis data are produced. So, if ECMWF decides to have a perfect new model and people will be interested in a 50 time series of homogenous data set (reanalysis), they would take there best (what ever this means right now) model and re-do the processing of analysis data - and these data are called re-analysis. Therefore, the time period  of re-analysis data is always shorter (focussing some period, not the full period of having observations) and limited to lower resolution (due to computer costs) and partly the number of output variables is reduced.

Therefore it is not possible to take re-analysis data, if analysis do not exist, it would be the other way around: re-analysis data are limited.

Which data to use?

It depends on the research ... is your tempered covered by both: analysis and re-analysis? Are all variables available? Which resolution is necessary?

If you would compare e.g. a climate model simulation for present day with observations, you would try to compare two 30 year time periods, therefore you would try to get some re-analysis data. If you investigate some synoptic events, you would preferably go for analysis data, as you would benefit from higher resolution.

Does this help?

Cheers,  Tim

Dear Tim.

That was really enlightening. Thank you for your time.

One problem I’ve encountered sometime ago was that when I looked for analysis data (eg surface wind every 6 hours) from the ECMWF, I found out that they were not freely distributed. Reanalysis data (in general) on the other hand, were available after a simple registration.

Do you know if I missed something? Is there any way to get analysis datasets from ECMWF?

No, you have not missed up something. Re-analysis data are (to some degree) free, analysis data could  be freely  available via your member state participation of your local weather service. Political (and financial) details I am not that familiar with. I would redirect you to your weather service.

That is was my info too.

It is a pity they are not available to the public.

Thank you again!

In my opinion, re-analysis data are regenerated by using data assimilation models and adding more observations from several non-realtime dataset. By contrast, analysis data are always done by operational data assimilation system which could be performed  to provide an initial condition for a subsequent forecasting.

Hi Elaheh Owlad

from the perspective of coastal engineering

There are two types of ECMWF data, first is operational data which has future wether forecasts for 5 days later with resolution of 0.5*0.5, because of the uniformity lack in precision due to model updating ,another project was defined as Reanalysis which the model was run again with uniform specifications in specific durations,which is famous as ERA-40 with resolution of 1.25* 1.25. for wave modeling near Iran's south coastal regions it is better to use operational because reanalysis estimates the wind velocity less than it's actual amount .