[Fwd: Re: [Met_help] Question, comments]

Wed Oct 1 11:51:54 MDT 2008

MET Folks,

A user has requested that we add a new column of output to the MODE output stats file relating to area.  Currently there are columns in the output named:
AREA = the number of points within an object.
AREA_FILTER = the number of points within an object containing non-zero data in the RAW field.

This user would like us to add a third column named:
AREA_THRESH = the number of points within an object that meet the threshold criteria set when defining the object in the RAW field.

He thinks this measure would tell him more about the "density" of the raw data within the object.

For example, suppose you're running MODE on a precip field, and you've defined 1 object with AREA = 1000 by convolving and thresholding at >= 5mm.
There may be some points in the raw field with zero data that got "convolved" into the object.  So AREA_FILTER (= non-zero points in the object) may be = 800.
Then we'd check to see which points in the raw field are >= 5mm.  And we may get AREA_THRESH = 500.

For precip, I'm pretty sure the following relationship would hold: AREA >= AREA_FILTER >= AREA_THRESH.

It seems like a reasonable addition to me.  Any objections?

John

-------- Original Message --------
Subject: Re: [Met_help] Question, comments
Date: Wed, 01 Oct 2008 10:42:07 -0700
From: Correia, James <james.correia at pnl.gov>
To: John Halley Gotway <johnhg at rap.ucar.edu>

John-
Thanks for the help. I didn't know there was a tool to choose which pages of
a ps file to keep. This is what I needed. I intend on keeping all the plots
but having all 4 in one file (with 1464 times per year) was a bit large.

For the last item regarding the area and area filter, I think it would be
useful for both. That is keep area, and area_filter, but add area_thresh
(here you defined it as the number of points at or above the threshold).
That would be a good addition.

jimmyc

On 9/29/08 4:02 PM, "John Halley Gotway" <johnhg at rap.ucar.edu> wrote:

> James,
> 
> My responses are inserted inline below:
> 
> Correia, James wrote:
>> Hi
>> I have been using MODE on 3 hourly rainfall from a regional climate model
>> (i.e. Been staring at a lot of data).
>> 
>> Question:
>> Is there a way to turn off 1 or 2 of the plots produced by MODE in the
>> output postscript file?
> 
> Regarding turning off some of the plots that MODE produces... here are your
> options:
> 
> (1) On the command line, you can turn OFF the postscript plots using the
> "-plot" command line option.
> (2) If you're scripting your calls to MODE, you could have MODE produce them,
> and then use something like the "psselect" command to pull out the pages you'd
> like to keep.  And then delete the original
> PostScript file from MODE.
> (3) You could edit the file "mode.cc" to suppress the creation of some pages,
> and then recompile MET.
> 
> If you need help doing (2) or (3), just let me know and I'd be glad to help.
> Which page(s) would you like to retain?
> 
>> 
>> I have been working on the assumption that only 1 time per input file is
>> required. Is this correct or can MODE handle more than 1 time per file?
>> 
> 
> As for whether or not MODE can handle multiple times in a file, I believe that
> the answer is no.  But let me explain.
> 
> If the MODE configuration file, you use the "fcst_grib_code" and
> "obs_grib_code" to specify the name of the field and accumulation interval in
> which you're interested.  For example:
> fcst_grib_code = "APCP/A3";
> obs_grib_code  = "APCP/A3";
> 
> MODE will open the input GRIB files (specified on the command line) and look
> for the first GRIB record that contains (in this case) precipitation
> accumulated over 3 hours.  If your GRIB file contains
> multiple records that meet this criteria (maybe valid at different times),
> it'll just grab the first one it encounters.
> 
>> 
>> Comments:
>> I generally like the way code operates. However, with 3 hourly output or
>> longer there is a dependency that large rainfall rate grid boxes, once
>> convolved, effectively increase the area of the object. In one instance I
>> have a precipitation field that is relatively broken (mixed low and high
>> precipitation rates and zero's). The resulting object is quite large
>> stretching from corner of the domain to the southern end of the domain. If
>> you would like I can send an image.
> 
> As you mention below, I'd suggest continuing to play around with the
> configuration settings to get MODE to define objects the way they want them to
> look.  By decreasing the convolution radius you'll
> get objects that are smaller and less smooth.  It can be set as low as 0 (= no
> convolution), but the big downside is that with too many objects, MODE will
> run pretty slowly and there will be too much
> output to deal with.
> 
> Ultimately, it's up to you to decide what scale you're interested in.  We
> often run MODE here multiple times using different configuration to look at
> performance at different scales.
> 
>> 
>> Is it possible that an additional measure be output that includes the
>> percent of the object area with restored grid point precipitation values
>> which fall at or below the 1st and/or second threshold value? I guess this
>> would be a confidence measure indicating if this object is uniformly filled
>> or chaotically filled.
> 
> The "AREA_FILTER" column in the current MODE output contains a count of the
> number of points inside an object that contain non-zero values.  So for
> precipitation, looking at AREA_FILTER/AREA would
> give you can indication of how "dense" the raw data is in the object.
> However, perhaps it'd be better to redefine this column to indicate the number
> of grid points in the object that meet the
> threshold criteria, rather than simply being non-zero.  That's actually
> probably a better approach anyway since it'd apply well to continuous fields
> too - rather than just precip.
> 
> Would that meet your needs?
> 
> If so, I could probably provide you with an updated file and have you
> recompile MET.
> 
> Hope that helps,
> John Halley-Gotway
> johnhg at ucar.edu
> 
>> 
>> Generally I have been playing with both threshold values, and the convolved
>> radius. I am only working with model data so I am not merging objects.
>> 
>> Thanks
>> jimmyc
>> 
>> 
>> James Correia Jr., PhD
>> Climate Physics Group
>> Post. Doc.
>> Pacific Northwest National Lab
>> 
>> "Wisdom. Strength. Courage. Generosity. Each of us are born with one of
>> these. It is up to us to find the other three inside of us."
>> -Into the West
>> 
>> 
>> 
>> 
>> 
>> _______________________________________________
>> Met_help mailing list
>> Met_help at mailman.ucar.edu
>> http://mailman.ucar.edu/mailman/listinfo/met_help