[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]

20031126: SKIP and GDCNTR

Greg,

comments interspersed below:


>From: Greg Quina <address@hidden>
>Organization: UCAR/Unidata
>Keywords: 200311261637.hAQGbGEH002540

>Hello,
>         I have grib files that are projected on the 12 km Arakawa-E 
>grid.  The actual dimensions of our grid are 606x1067 on a similar grid to 
>that of the original ETA grid 96 in the grdnav.tbl, and there is data for 
>every point on this 606x1067 grid.  The Data Representation Type is 203.  I 
>have been using the NAGRIB routine to decode the data to gempak grid files.
>
>Nagrib seems to work fine (using gdlist to verify the data) when 
>transforming the data to the staggered grid on the following projection:
>!ID     #  PRJ  A1  A2 A3  LLT      LLN      ULT     ULN   KX    KY DELN  X
>ET12 96 CED 50 -111 0 -3.441 -148.799 50.026 -3.230 1211 1067 0.5     0
>However, I can not get the skip command to work, except by skipping both 
>row and column (i.e., skip = 1 works fine).  Skip_plot_x nor skip_plot_y 
>will allow me to skip wind points.  No matter what I use for skip_plot_x or 
>skip_plot_y (for example, skip = /1;-1), I get the same result.  I am not 
>sure why this feature will not work.
>
>I have commented out the portion of the code that staggers the data if data 
>type is 203 and re-compiled to get a 606x1067 GEMPAK grid on the following 
>projection:
>!ID     #   PRJ  A1  A2 A3   LLT      LLN      ULT     ULN   KX    KY DELN   X
>ET12 96 CED  50 -111 0  -3.441 -148.799 50.026 -3.230  606 1067  1.0     0
>This non-staggered data (to avoid skip) is plotted well, but I am not sure 
>if the lat/lon plotted coordinates here would match that of the mass points 
>on the staggered grid.

GEMPAK can only work with rectangular grids, eg all data has to exist at
each grid point for calculations, so you would want the 1211x1057 grid.
The staggered grid has to be interpolated so that U,V will be at the same grid 
points as H,T,RH, etc. In the E-grid, the u,v will be offset. The interpolation
in nagrib will interpolate all quantities to each point.

The number of grid points allowed by default in calculations is
set to 400,000 for LLMXGD in MCHPRM.xxxx and gemprm.h. The Maximum storage
of a grid is set to 750,000 for LLMXTG in MCHPRM.xxxxx. So, your
1211x1067 grid is too big for the defaults in the distribution.

If you have compiled the distribution locally, you need LLMXTG to
be large enough to store your 1211x1067 grid.
On calculations in gdcntr, gdplot2, etc, you can use IJSKIP to
load a subset of the data (either aerial or subsampled) for
calculations. The 400,000 points would be the default, but you could increase 
that as well. The IJSKIP allows you to use a subset of points in the
larger data file, whereas the SKIP parameter in plotting is just to
alternate the plotting of the values after calculations. If your
data file has exceeded the number of grid points compiled into the arrays, that
may have caused your problem with SKIP= /1;-1.

>
>In addition, I am trying to plot vegetation type, soil type, and land/sea 
>masks gridded data sets.  When I use gdcntr I continuously get an 
>interpolated data set.  That would be fine for most of my fields, except 
>for the above because they consist of integer class values and should not 
>be interpolated.  I want to fill the entire grid box with its value, 
>instead of doing any interpolation from grid box to grid box.  I thought 
>that setting contur = 0 (or 0/0 or blank) would achieve this, but 
>interpolation between grid values still exists when mapped by gdcntr.


I provided a TYPE of "Z" in gdplot2 for just filling in a grid box rather than
contouring which would interpolate between your values. This essentially
makes the output an image rather than a contour plot.


Steve Chiswell




>
>Would you have any suggestions on the following:
>1.  How to plot the staggered data using skip or another feature (ijskip)?
>2.  Would plotting the 606 x 1067 grid match the staggered 1211x1067 mass 
>points?
>3.  How can I fill entire grid boxes with their grid point values, using no 
>interpolation?
>
>Please let me know if you need any further clarification on the data or 
>what I am looking for in the output.
>
>Thanks,
>Greg Quina
>
>---------------------------------------------------------
>Greg Quina
>NASA Goddard Space Flight Center --- SAIC
>Hydrological Sciences Branch
>NASA-GSFC Mail Code 974.1
>Greenbelt, MD 20771
>Office: Building 22, Room 008
>
>Email: address@hidden
>Phone: (301) 286-8272
>---------------------------------------------------------
>
>From address@hidden Fri Nov 21 07:55:54 2003
>Received: from conan.unidata.ucar.edu (conan.unidata.ucar.edu [128.117.140.64]
> )
>       by unidata.ucar.edu (UCAR/Unidata) with ESMTP id hALEtsEH010333
>       for <address@hidden>; Fri, 21 Nov 2003 07:55:5
> 4 -0700 (MST)
>Received: (from www@localhost)
>       by conan.unidata.ucar.edu (UCAR/Unidata) id hALEtsM6018393;
>       Fri, 21 Nov 2003 07:55:54 -0700 (MST)
>Date: Fri, 21 Nov 2003 07:55:54 -0700 (MST)
>Organization: UCAR/Unidata
>Keywords: 200311211455.hALEtsM6018393
>To: address@hidden
>Subject: Gempak - SGI IRIX64 - ETA Projection 96, nagrib
>MIME-Version: 1.0
>From: "Greg Quina" <address@hidden>
>Content-Type: text/plain; charset="ISO-8859-1"
>Content-Transfer-Encoding: 7bit
>Message-ID: <address@hidden>
>X-Spam-Status: No, hits=0.4 required=5.0
>       tests=AWL,NOSPAM_INC,SPAM_PHRASE_01_02,SUPERLONG_LINE
>       version=2.43
>X-Spam-Level: 
>
>Institution: Goddard Space Flight Center
>Package Version: 5.6.j
>Operating System: SGI IRIX64
>Hardware Information: 1280Mb Memory
>Inquiry: Hello,
>I am working with land surface models (LSMs), and we are attempting to couple 
> our land model with the 12 km ETA model.  The output of the LSMs and ETA mode
> l will contain grib files that are projected on the current 12 km Arakawa-E g
> rid.  However, in contrast to the current operational setup of the 12 km grid
> , all output data are on a common grid (meaning no staggering is involved).  
> The actual dimensions of our grid are 606x1067 on a similar grid to that of t
> he original ETA grid 96 in the grdnav.tbl, and there is data for every point 
> on this 606x1067 grid.  I have been using the NAGRIB routine to decode the da
> ta to gempak grid files.
>
>
>In the grdnav table, if I set projection 96 to the following:
>!ID       #   PRJ  A1   A2   A3     LLT       LLN       ULT       ULN   KX   K
> Y  DELN  X
>ET12  96  CED  50  -111   0   -3.441  -148.799  50.026   -3.230  606  1067  1.
> 0    0
>.....  when I use gdcntr to view the data, I get a strange representation of t
> he data that does not match the map background.
>
>
>If I set the grdnav.tbl file to:
>!ID       #   PRJ  A1   A2   A3     LLT       LLN       ULT       ULN   KX   K
> Y  DELN  X
>ET12  96  CED  50  -111   0    -3.441  -148.799  50.026   -3.230 1211 1067   0
> .5    0
>
>
>I can produce an image in gdcntr that looks reasonably good.  However, the LSM
>  gives undefined values (-9999) over the oceans and lakes.  These undefined v
> alues are being used to interpolate data along the coastlines (where there is
>  no wind data)..... this yields fields such as temperature with huge negative
>  values, but not -9999.  I have tried using the skip parameter in gdcntr to s
> kip every other point, and it only seems to help when I skip both row and col
> umn (SKIP = 1).  When I try to skip only columns of the data (SKIP = /1), the
>  problem remains.
>
>Would you have any suggestions on how to either plot the data directly using K
> X=606 (in the grdnav.tbl) OR by adding columns and by  using KX=1211 and elim
> inate the interpolated values from being shown?
>
>
>Please let me know if you need any further clarification on the data or what I
>  am looking for in the output.
>
>
>Thanks,
>Greg Quina
>
>
>