[gradsusr] GFS Precipitation

Wed May 9 05:21:30 EDT 2012

Eric,

Thank you very much for so detailed explanation.

Yes, the product I am using (sfluxgrb files) keeps only  PRATE,
defined exactly as you mentioned: could be 6h mean or 3h mean
rate for the short-range forecasts until 192h and 12h mean for the
medium-range forecasts on the coarser grid.

For me using PRATE is quite convenient as it is a "flux",
so I estimate an amount of freshwater delivered to the sea surface
by multiplying PRATE on the ocean model integration time step.
Using an APCP would require additional efforts for convertion it to flux
as the integration period is not constant during the meteorological
forecast.

The sfluxgrb  GFS product I am using is defined on Gaussian grid
so has slightly higher spatial resolution compared to the pgrb,
and, also, keeps only selection of data on the single level,
basically surface or other special atmospheric levels like
the top of atmosphere. It is also convenient as I basically
do not need upper atmosphere data.

Sincerely yours,
Sergey Varlamov

Eric Altshuler wrote:
> Sergey,
>
> PRATE is actually the time average precipitation rate over the same periods that are used to accumulate APCP. In other words, PRATE is the average precip from T-6h to T for T=6,12,18,24,... and from T-3h to T for T=3,9,15,21,... The same "peculiarity" in APCP also applies to PRATE.
>
> PRATE and APCP are proportional to each other. For T=6,12,18,24... PRATE=APCP/21600 (6h=21600s). For T=3,9,15,21,... PRATE=APCP/10800 (3h=10800s).
>
> The information I found about PRATE and APCP can be obtained by running wgrib2 on the GFS files. The data we get at COLA are the 'pgrb2' files, which is probably the most commonly used GFS product. At present, the most recent available GFS run is for 2012050900.
>
> wgrib2 -V gfs.t00z.pgrb2f12 | grep PRATE
>
> gives the following result:
>
> 199:38000342:vt=2012050912:surface:6-12 hour ave fcst:PRATE Precipitation Rate [kg/m^2/s]:
>
> For APCP, the output is:
>
> 200:38124544:vt=2012050912:surface:6-12 hour acc fcst:APCP Total Precipitation [kg/m^2]:
>
> On the other hand, doing the same thing for the 9-h forecast file gives:
>
> 199:37371015:vt=2012050909:surface:6-9 hour ave fcst:PRATE Precipitation Rate [kg/m^2/s]:
> 200:37496621:vt=2012050909:surface:6-9 hour acc fcst:APCP Total Precipitation [kg/m^2]:
>
> After forecast hour 192, the GFS output is 12-hourly and all forecast hours have the same accumulation/averaging period, T-12h to T.
>
> The above applies to the 'pgrb2' type files. I'm not sure if it applies to other GFS products. You can run similar tests on your 'sfluxgrb' files to see if the accumulation/averaging periods behave the same way. The 'pgrb2' files include both PRATE and APCP, but I think the 'sfluxgrb' files only have PRATE.
>
> Best regards,
>
> Eric L. Altshuler
> Assistant Research Scientist
> Center for Ocean-Land-Atmosphere Studies
> 4041 Powder Mill Road, Suite 302
> Calverton, MD 20705-3106
> USA
>
> E-mail: ela at cola.iges.org
> Phone: (301) 902-1257
> Fax: (301) 595-9793
>
> ----- Original Message -----
> From: "Sergey Varlamov" <vsm at jamstec.go.jp>
> To: "GrADS Users Forum" <gradsusr at gradsusr.org>
> Sent: Tuesday, May 8, 2012 11:04:03 PM
> Subject: Re: [gradsusr] GFS Precipitation
>
> Eric,
>
> Thank you for these details as it could be important for our
> applications as well.
> I am sometimes using GFS data for forcing of regional ocean models.
> Normally I use data from the
> http://nomads.ncep.noaa.gov/pub/data/nccf/com/gfs/prod/gfs.{y}{m}{d}{h}/gfs.t{h}z.sfluxgrbf{f}.grib2
> with {X} replaced with numbers for year, month etc.
> In terms of wgrib2 I use for estimation of precipitation "PRATE:surface"
> parameter.
> I assume that, independently of the forecast lead time,
> it is precipitation rate estimated at the  model surface,
> not an accumulated precipitation that you mention.
>
> However, if I start to use other GFS products it will need to be aware
> of the peculiarity that you mention...
>
> In this relation, could you give a reference to the source of your
> information
> on the definition of GFS accumulated precipitation?
>
> By the way, Japan Meteorological Agency global product
> provides precipitation accumulated from the start of forecasting
> (over 6h, then over 12h etc. as forecast time is increasing).
> So precipitation is all time a pain for utilization.
>
> Sincerely yours,
>
> Sergey Varlamov
>
>
> Eric Altshuler wrote:
>   
>> Adrian,
>>
>> GFS total accumulated precip (apcpsfc) includes all forms of precipitation that reaches the surface during the accumulation period. For forecast hours evenly divisible by 6, the accumulation period is from T-6h to T, while for other forecast hours (divisible by 3 but not 6) it is from T-3h to T.
>>
>> Best regards,
>>
>> Eric L. Altshuler
>> Assistant Research Scientist
>> Center for Ocean-Land-Atmosphere Studies
>> 4041 Powder Mill Road, Suite 302
>> Calverton, MD 20705-3106
>> USA
>>
>> E-mail: ela at cola.iges.org
>> Phone: (301) 902-1257
>> Fax: (301) 595-9793
>>
>> ----- Original Message -----
>> From: "Adrian Scherzinger" <adrian.scherzinger at meteotest.ch>
>> To: gradsusr at gradsusr.org
>> Sent: Monday, May 7, 2012 7:19:57 AM
>> Subject: [gradsusr] GFS Precipitation
>>
>> Hi
>>
>> I have a question about total precipitation in GFS. I got cunfused when 
>> I was looking at precipitation forecast in high altitudes (Himalaya):
>>
>> Does the accumulated precipitation at surface (apcpsfc) include solid 
>> precipitation (e.g snow, graupel)? Or do I have to include other 
>> parameters like Accumulated Snow to get the real total precipitation?
>>
>> Any help is appreciated!
>>
>> Adrian
>>
>>   
>>     