[Wrf-users] Comparing U,V 10 meters to observations

Ligia Bernardet Ligia.Bernardet at noaa.gov
Wed Jan 27 08:12:44 MST 2010

Josh and Juan,

I ran into this issue while trying to compare 10-m model wind forecasts against hurricane maximum wind reports, which are also an average over a few minutes.  There was no good solution. 

Having WRF output time series of variables at a grid point or time series of statistical quantities (such as maximum winds) has been a standing request to developers, and may be addressed at some point. 

I have talked to several hurricane modelers who, for research purposes, output the model winds every time step to look at how much it varies. Different authors got to different conclusions. Some modelers noted that their model winds fluctuate a lot (and therefore averaging is really necessary before comparing to obs), other noted that their winds are pretty steady (and therefore direct comparison to observations is not a problem).

So, I think this is an outstanding issue. If others have experience with this, I would also like to know more.


On Jan 26, 2010, at 5:06 PM, Hacker, Joshua (Josh) (CIV) wrote:

> Apologies to those who know this better than I do, but this is an interesting question…
> Unless you do something fancy, the WRF outputs instantaneous grid-point values, regardless of your output interval.  Those values are filtered non-trivially by some function of implicit and explicit diffusion in the model (think physics, damping, time and space discretization, etc).  Therefore there is no one answer to your question. 
> In practice (I believe) most people just compare directly to 10-min averaged wind obs (WMO standard), or 2-min averaged wind obs (USA) but in my opinion 2 minutes is usually too short to be fair.
> One approach to getting at the averaging in the model is through data assimilation and statistical consideration of observation error levels when the error may be dominated by representativeness error (difference between scales represented in obs and model).  Others have looked in spectral space to get an idea of the averaging scales.
> The other thing that you might consider, if you have the data to do it, is to compare WRF forecasts with obs averaged over many different time lengths (or alternatively band-pass filtered).  Then you can get a more complete picture of what time scales it can or cannot predict with skill, and decide whether it is doing what you hope.
> That said, I’d be interested in hearing any other comments from the community and especially our verification gurus – is there any “conventional” wisdom or rules of thumb that folks are using or that have recently shown promise, empirically?  Have I missed something?
> Sorry that there is no simple answer.
> Josh
> From: wrf-users-bounces at ucar.edu [mailto:wrf-users-bounces at ucar.edu] On Behalf Of Juan Gonzalez
> Sent: Tuesday, January 26, 2010 12:33 PM
> To: wrf-users at ucar.edu
> Subject: [Wrf-users] Comparing U,V 10 meters to observations
> Hi everyone,
> I am interested on validating the WRF 10m wind forecast with observational data. I would like to know what is the corresponding average scale for the WRF 10m winds, that is, do the correspond to 1-minute, 10-minute average winds or what kind of average, if any?
> My WRF configuration outputs wind every 3 hours.
> Thanks for your help, 
> Juan Gonzalez 
> -- 
> Juan O. Gonzalez
> Research Assistant - Caribbean Integrated Coastal Ocean Observing System
> Graduate Student - Physical Oceanography
> Marine Sciences Department
> University of Puerto Rico - Mayaguez Campus
> _______________________________________________
> Wrf-users mailing list
> Wrf-users at ucar.edu
> http://mailman.ucar.edu/mailman/listinfo/wrf-users

-------------- next part --------------
An HTML attachment was scrubbed...
URL: http://mailman.ucar.edu/pipermail/wrf-users/attachments/20100127/7f92caa6/attachment.html 

More information about the Wrf-users mailing list