FYI, [email protected], useful info on sub region.


-----Original Message-----
From: <Whitehall>, Kim Whitehall <[email protected]>
Date: Sunday, April 19, 2015 at 12:00 PM
To: "Loikith, Paul C (329C-Caltech)" <[email protected]>
Cc: rcmes-dev <[email protected]>
Subject: Re: sub-region testing

>
>
>
>On TuesdayApr 14, at 12:23 PM, Loikith, Paul C (329C-Caltech)
><[email protected]> wrote:
>
>Hi Kim,
>
>I did some playing around and testing of the subregion example that makes
>a portrait diagram and was able to successfully get it to run by changing
>the input files (the precipitation files for CORDEX Africa that are in
>the example code aren't in the local directory but there are two tasmax
>ones so I just switched them out).  It is hard to check whether the
>output values are correct easily, but I will cross validate the results
>with other methods as well to make sure the code is doing what we hope it
>is doing once I understand a few more of the code elements.
>
>To make sure I understand what is going on in the example I thought it
>would be useful to document a few questions I had going through the
>script as a first timer.  Perhaps you could clarify more inline.  Forgive
>me if these are insanely naïve, but some of code is unclear to me so this
>is a good opportunity to double check everything and to also find places
>where an OCW user may get confused.
>
>1)  Lines 76-80:
>
>Resampling Datasets:  I was a little confused about what this means.
>"Resampling" is probably not the best word to use to document this
>because that conjures up some sort of bootstrapping statistical test to a
>climate scientist…I don't think there is really any resampling happening
>here.  
>
>It looks like dsp.water_flux_unit_conversion is converting the
>precipitation units to some standard unit?  Is this correct?  If so, what
>is the protocol here?  Are they all being converted to mm/sec /day just
>monthly accumulated?
>
>Similarly temporal rebin is a little confusing.  I know we've used this
>terminology for awhile, but it isn't really conventional.  Maybe temporal
>regridding would be a better choice of terminology.  In the case here,
>everything is monthly so there shouldn't be any temporal adjustments at
>all, unless I'm mistaken.  It appears that it is making each month into
>30 days?  
>
>2)  Lines 98-101:  Here it looks like the climatology of the observations
>and models are being computed presumably resulting in a 12 x lat x lon
>array with one mean value for each month.  Afterwards it is unclear to me
>why the climatology would be computed.  Are anomalies computed somewhere
>else before the pattern correlation is performed?  Maybe I'm missing
>something here.
>
>3)  Line 135:  What is the pattern correlation being computed between?
>My naïve impression would be that the climatologies of the observations
>and the models are being compared, but it isn't clear averaged over what
>months.  A pattern correlation can only be computed between two 2D
>arrays, so there is no time dimension, but if it were the two
>climatologies I'm not sure which of the 12 months would be compared.  I'm
>probably just missing something here, but it should be clear what the
>pattern correlation is correlating either way.
>
>Thanks!
>
>Paul  
>
>
>
>
>
>
>
>
>
>
>

Reply via email to