[NARCCAP-discuss] Model evaluation and bias correction

[Shuang-Ye Wu]

Hi Micheal,

Thank you very much for your perspectives and the linked resources.  Really
appreciate it.

Shuang-Ye

On Mon, Jun 13, 2011 at 12:51 PM, Michael Wehner <mfwehner at lbl.gov> wrote:

> Shuang-Ye
> You raise a number of difficult questions.
>
> 1) is it appropriate to compare models directly with station data? Or
> should one compare to a gridded observational product? I am using the the
> data at http://www.esrl.noaa.gov/psd/data/gridded/data.unified.html
> 2) How does one evaluate different models in a multi-model context? See
> http://www.ipcc-wg2.gov/meetings/EMs/IPCC_EM_MME_GoodPracticeGuidancePaper.pdffor some useful perspectives.
> 3)  NCEP vs GCM runs for bias corrections. I don't think that there is a
> correct answer for this question. Given that there is only a single
> realization for each RCM, there is no way to access the statistical
> significance of the differences in bias corrections obtained by using the
> differently driven runs. For consistency in the projections, one might argue
> that you should use the GCM driven runs. On the other hand, it would also
> make sense to use both to make a small attempt to determine robustness.
>
> Hope that this helps.
> Michael
>
> On Jun 13, 2011, at 9:24 AM, Shuang-Ye Wu wrote:
>
> > Hi all,
> >
> > This question might be a bit lengthy, so please bear with me.
> >
> > I am trying to use NARCCAP models to project possible changes in daily
> > precipitation pattern for a particular area.  The first step is to
> compare
> > NARCCAP model retrospective runs with observed station data.  The next
> step
> > is to correct bias based on this comparison, using some kind of scaling
> > methods.  My question is: should I use NCEP-driven runs or GCM-driven
> runs
> > for the comparison and bias correction?  I thought I should use
> GCM-driven
> > retropsective runs to evaluate the total bias from both the global and
> > regional models, and correct the total bias for future projections.  But
> > some suggest I should use NCEP runs for model evaluation, because I'd
> then
> > compare observation with NARCCAP models driven by observation data.  If
> > that's the case, should I do bias correction separately using GCM-driven
> > runs?  Any ideas/suggestions?
> >
> > Thank you!
> >
> > Best,
> > Shuang-Ye
> >
> > --
> >
> >
> >
> > ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
> > Shuang-Ye Wu, PhD
> > Assistant Professor
> > Geology Department
> > University of Dayton
> > 300 College Park
> > Dayton, Ohio 45469-2364
> > Tel: 937-229-1720
> > Fax: 937-229-2889
> > _______________________________________________
> > narccap-discuss mailing list
> > narccap-discuss at mailman.ucar.edu
> > http://mailman.ucar.edu/mailman/listinfo/narccap-discuss
>
> _______________________________________________
> narccap-discuss mailing list
> narccap-discuss at mailman.ucar.edu
> http://mailman.ucar.edu/mailman/listinfo/narccap-discuss
>



-- 



~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
Shuang-Ye Wu, PhD
Assistant Professor
Geology Department
University of Dayton
300 College Park
Dayton, Ohio 45469-2364
Tel: 937-229-1720
Fax: 937-229-2889