[RegCNET] 10 km run
Rupak Rajbhandari
rupak.rajbhandari at gmail.com
Fri Mar 16 09:08:20 CET 2007
Hi,
Last year I tried to downscale, from 50 km to 15/20 km resolution for
the Himalayan area (Nepal) in order to compare with the station data
(Gridded), but the machine crashed all the time. I was using
PentiumIV, 1.7 Ghz machine. I hope this time it works!!!
Rupak Rajbhandari
Department of Meteorology
Tribhuvan University
NEPAL
On 3/15/07, Moetasim <mashfaq at purdue.edu> wrote:
> Hi,
>
> 1) I agree with gao that a direct run at 10 km using 2.5 degree ICBC data is a
> bit harsh use of RegCM3. The better approach is to run RegCM at intermediate
> resolution (say 25/30 km) and then run a further nested run at 10 km. Also, it
> would be advisable to use better resolution ICBC data (e.g. a high resolution
> ERA40).
>
> 2) One reason of RegCM over estimation is your small domain. I expect model
> getting drier if you have a reasonbaly large domain. Small domains/high
> resolutions tend to give over estimation of precpitation. You have to select a
> reasonably large domain. A 50 x 50 domain at 50 km would be a minumum near me-
> this would require you to select a ~250 x 250 domain at 10 km at least.
>
> 3) As gao said, CRU data should not be used to compare high resolution run.
> Better choice is station data or high resolution gridded data.
>
> Moet
>
>
> The high precp at 10 km is bit surprising for me. I understand that the domain
> of your simulation is very small (smaller than what a RCM domain should be as a
> minimum).
>
> Quoting gaoxj at cma.gov.cn:
>
> > Hi Csaba,
> >
> > Seems you are doing the highest resolution simulation by RegCM3. I think we
> > do not have such experiences before. I have some suggestions, but do not know
> > whether they will work or not.
> >
> > 1. Generally, people say that the ratio for the driven resolution to RCM
> > resolution is better to be in the range of 3-5, 10 is the maximum. There
> > should be some references on this, but I do not have in hand. Maybe you can
> > try the double-nesting, from 2.5 degrees to say 50 km first, then from 50 to
> > 10?
> >
> > 2. CRU is in the resolution of 0.5 deg. It smoothes the precip peak.
> > Meanwhile, over mountain areas, it tends to under-estimate the real precip
> > because no weather stations there. What about try to evaluate the model by
> > some station data?
> >
> > 3. A critical issue is that RegCM3 is hydro-statistic. Some people say a
> > model like that can go to 10 km, some said not and you need a
> > non-hydro-statistic one. I have no idea about this. Maybe others have?
> >
> > Good luck.
> >
> > Regards,
> >
> > Gao
> >
> > ----- Original Message -----
> > From: "Torma Csaba" <delivitez at nimbus.elte.hu>
> > To: <regcnet at lists.ictp.it>
> > Sent: Wednesday, March 14, 2007 6:20 PM
> > Subject: [RegCNET] 10 km run
> >
> >
> > > Dear RegCM Users,
> > >
> > > We just started to use RegCM3 driven by ERA40 2.5 degree resolution ICBC
> > > data for a European subregion containing the Carpathian Basin and Hungary.
> > >
> > > Our grid resolution is 10 km.
> > >
> > > In the precipitation pattern we found big biases as compared to the CRU
> > > data, especially during summer (convective events cause too much
> > > precipitation). We tried to use both Emanuel and Grell convective schemes,
> > > but the result was quite the same: the precipition was overestimated by
> > > the model with a factor of about 2.
> > >
> > > We suspect that the 10 km resolution causes this overestimation. Using 25
> > > km grid resolution the result is much more realistic, but we would like to
> > > use 10 km grid spacing. Is there any possible solution for this
> > > overestimation? Any suggestion is welcome!
> > >
> > > Thanks in advance.
> > >
> > > Kind regards,
> > > Csaba TORMA
> > > _______________________________________________
> > > RegCNET mailing list
> > > RegCNET at lists.ictp.it
> > > https://lists.ictp.it/mailman/listinfo/regcnet
> > >
> > _______________________________________________
> > RegCNET mailing list
> > RegCNET at lists.ictp.it
> > https://lists.ictp.it/mailman/listinfo/regcnet
> >
> reasonablyk we
> > do not have such experiences before. I have some suggestions, but do not know
> > whether they will work or not.
> >
> > 1. Generally, people say that the ratio for the driven resolution to RCM
> > resolution is better to be in the range of 3-5, 10 is the maximum. There
> > should be some references on this, but I do not have in hand. Maybe you can
> > try the double-nesting, from 2.5 degrees to say 50 km first, then from 50 to
> > 10?
> >
> > 2. CRU is in the resolution of 0.5 deg. It smoothes the precip peak.
> > Meanwhile, over mountain areas, it tends to under-estimate the real precip
> > because no weather stations there. What about try to evaluate the model by
> > some station data?
> >
> > 3. A critical issue is that RegCM3 is hydro-statistic. Some people say a
> > model like that can go to 10 km, some said not and you need a
> > non-hydro-statistic one. I have no idea about this. Maybe others have?
> >
> > Good luck.
> >
> > Regards,
> >
> > Gao
> >
> > ----- Original Message -----
> > From: "Torma Csaba" <delivitez at nimbus.elte.hu>
> > To: <regcnet at lists.ictp.it>
> > Sent: Wednesday, March 14, 2007 6:20 PM
> > Subject: [RegCNET] 10 km run
> >
> >
> > > Dear RegCM Users,
> > >
> > > We just started to use RegCM3 driven by ERA40 2.5 degree resolution ICBC
> > > data for a European subregion containing the Carpathian Basin and Hungary.
> > >
> > > Our grid resolution is 10 km.
> > >
> > > In the precipitation pattern we found big biases as compared to the CRU
> > > data, especially during summer (convective events cause too much
> > > precipitation). We tried to use both Emanuel and Grell convective schemes,
> > > but the result was quite the same: the precipition was overestimated by
> > > the model with a factor of about 2.
> > >
> > > We suspect that the 10 km resolution causes this overestimation. Using 25
> > > km grid resolution the result is much more realistic, but we would like to
> > > use 10 km grid spacing. Is there any possible solution for this
> > > overestimation? Any suggestion is welcome!
> > >
> > > Thanks in advance.
> > >
> > > Kind regards,
> > > Csaba TORMA
> > > _______________________________________________
> > > RegCNET mailing list
> > > RegCNET at lists.ictp.it
> > > https://lists.ictp.it/mailman/listinfo/regcnet
> > >
> > _______________________________________________
> > RegCNET mailing list
> > RegCNET at lists.ictp.it
> > https://lists.ictp.it/mailman/listinfo/regcnet
> >
> precipitationwe
> > do not have such experiences before. I have some suggestions, but do not know
> > whether they will work or not.
> >
> > 1. Generally, people say that the ratio for the driven resolution to RCM
> > resolution is better to be in the range of 3-5, 10 is the maximum. There
> > should be some references on this, but I do not have in hand. Maybe you can
> > try the double-nesting, from 2.5 degrees to say 50 km first, then from 50 to
> > 10?
> >
> > 2. CRU is in the resolution of 0.5 deg. It smoothes the precip peak.
> > Meanwhile, over mountain areas, it tends to under-estimate the real precip
> > because no weather stations there. What about try to evaluate the model by
> > some station data?
> >
> > 3. A critical issue is that RegCM3 is hydro-statistic. Some people say a
> > model like that can go to 10 km, some said not and you need a
> > non-hydro-statistic one. I have no idea about this. Maybe others have?
> >
> > Good luck.
> >
> > Regards,
> >
> > Gao
> >
> > ----- Original Message -----
> > From: "Torma Csaba" <delivitez at nimbus.elte.hu>
> > To: <regcnet at lists.ictp.it>
> > Sent: Wednesday, March 14, 2007 6:20 PM
> > Subject: [RegCNET] 10 km run
> >
> >
> > > Dear RegCM Users,
> > >
> > > We just started to use RegCM3 driven by ERA40 2.5 degree resolution ICBC
> > > data for a European subregion containing the Carpathian Basin and Hungary.
> > >
> > > Our grid resolution is 10 km.
> > >
> > > In the precipitation pattern we found big biases as compared to the CRU
> > > data, especially during summer (convective events cause too much
> > > precipitation). We tried to use both Emanuel and Grell convective schemes,
> > > but the result was quite the same: the precipition was overestimated by
> > > the model with a factor of about 2.
> > >
> > > We suspect that the 10 km resolution causes this overestimation. Using 25
> > > km grid resolution the result is much more realistic, but we would like to
> > > use 10 km grid spacing. Is there any possible solution for this
> > > overestimation? Any suggestion is welcome!
> > >
> > > Thanks in advance.
> > >
> > > Kind regards,
> > > Csaba TORMA
> > > _______________________________________________
> > > RegCNET mailing list
> > > RegCNET at lists.ictp.it
> > > https://lists.ictp.it/mailman/listinfo/regcnet
> > >
> > _______________________________________________
> > RegCNET mailing list
> > RegCNET at lists.ictp.it
> > https://lists.ictp.it/mailman/listinfo/regcnet
> >
> reasonablyk we
> > do not have such experiences before. I have some suggestions, but do not know
> > whether they will work or not.
> >
> > 1. Generally, people say that the ratio for the driven resolution to RCM
> > resolution is better to be in the range of 3-5, 10 is the maximum. There
> > should be some references on this, but I do not have in hand. Maybe you can
> > try the double-nesting, from 2.5 degrees to say 50 km first, then from 50 to
> > 10?
> >
> > 2. CRU is in the resolution of 0.5 deg. It smoothes the precip peak.
> > Meanwhile, over mountain areas, it tends to under-estimate the real precip
> > because no weather stations there. What about try to evaluate the model by
> > some station data?
> >
> > 3. A critical issue is that RegCM3 is hydro-statistic. Some people say a
> > model like that can go to 10 km, some said not and you need a
> > non-hydro-statistic one. I have no idea about this. Maybe others have?
> >
> > Good luck.
> >
> > Regards,
> >
> > Gao
> >
> > ----- Original Message -----
> > From: "Torma Csaba" <delivitez at nimbus.elte.hu>
> > To: <regcnet at lists.ictp.it>
> > Sent: Wednesday, March 14, 2007 6:20 PM
> > Subject: [RegCNET] 10 km run
> >
> >
> > > Dear RegCM Users,
> > >
> > > We just started to use RegCM3 driven by ERA40 2.5 degree resolution ICBC
> > > data for a European subregion containing the Carpathian Basin and Hungary.
> > >
> > > Our grid resolution is 10 km.
> > >
> > > In the precipitation pattern we found big biases as compared to the CRU
> > > data, especially during summer (convective events cause too much
> > > precipitation). We tried to use both Emanuel and Grell convective schemes,
> > > but the result was quite the same: the precipition was overestimated by
> > > the model with a factor of about 2.
> > >
> > > We suspect that the 10 km resolution causes this overestimation. Using 25
> > > km grid resolution the result is much more realistic, but we would like to
> > > use 10 km grid spacing. Is there any possible solution for this
> > > overestimation? Any suggestion is welcome!
> > >
> > > Thanks in advance.
> > >
> > > Kind regards,
> > > Csaba TORMA
> > > _______________________________________________
> > > RegCNET mailing list
> > > RegCNET at lists.ictp.it
> > > https://lists.ictp.it/mailman/listinfo/regcnet
> > >
> > _______________________________________________
> > RegCNET mailing list
> > RegCNET at lists.ictp.it
> > https://lists.ictp.it/mailman/listinfo/regcnet
> >
> _______________________________________________
> RegCNET mailing list
> RegCNET at lists.ictp.it
> https://lists.ictp.it/mailman/listinfo/regcnet
>
More information about the RegCNET
mailing list