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

Re: More Fly Cat Stuff

Mike,
The GEMPAK distribution you are downloading from our site is unmodified
in regard to the XVFR code as provided by NCEP.

The problem you are seeing with the 11Z observation from GBD:
KGBD 131055Z AUTO 25007KT 5SM CLR 09/07 A3005 RMK AO1

is related to roundoff error. The VSBY value of 5 miles above produces
an XVFR of 2.0 in the testpr.f program which calls the routine pr_xvfr:

Select a subroutine or type EXIT:PR_XVFR
Enter the ceiling in hundreds of feet
-9999.00
  Enter the visibility in miles
5.0
  xvfr =    2.000000

However, when using the data stored in the decoded metar file. produces
XVFR=3.:
>sfparm = dset;xvfr;text
 GEMPAK-SFLIST>r
 PARM = PMSL;ALTI;TMPC;DWPC;SKNT;DRCT;GUST;WNUM;CHC1;CHC2;CHC3;VSBY;P03D;P03I;  
 
        MSUN;SNOW;WEQS;P24I;TDXC;TDNC;P03C;CTYL;CTYM;CTYH;P06I;T6XC;T6NC;CEIL;  
 
        P01I;XVFR                                                               
 

    STN    YYMMDD/HHMM      PMSL     ALTI     TMPC     DWPC     SKNT     DRCT
                            GUST     WNUM     CHC1     CHC2     CHC3     VSBY
                            P03D     P03I     MSUN     SNOW     WEQS     P24I
                            TDXC     TDNC     P03C     CTYL     CTYM     CTYH
                            P06I     T6XC     T6NC     CEIL     P01I     XVFR
    GBD    070313/1100  -9999.00    30.05     9.00     7.00     7.00   250.00
                        -9999.00 -9999.00     1.00 -9999.00 -9999.00     5.00
                        -9999.00 -9999.00 -9999.00 -9999.00 -9999.00 -9999.00
                        -9999.00 -9999.00 -9999.00 -9999.00 -9999.00 -9999.00
                        -9999.00 -9999.00 -9999.00 -9999.00 -9999.00     3.00


KGBD 131055Z AUTO 25007KT 5SM CLR 09/07 A3005 RMK AO1


The cause of this is that the VSBY being passed to the PR_XVFR routine 
is showing a value of vsby = 5.00000047683
and when compared by (vsby .le. 5.0) is failing that test which you
would have
expected to match.

The metar.pack file uses a precision of .01 for the VSBY value, however
this roundoff error in the representation of the number in the retrieved
data is less than that.
and the logical comparison is exact. The 'C' code library provides a
G_DIFF function for this type of fuziness- but this routine is Fortran.

As a workaround, you may want to modify the pr_xvfr.f routine ".le."
comparisons as:
ELSE IF ( ceil .le. 30.001 ) THEN

and

ELSE IF ( vsby .le. 5.001 ) THEN


 I will send this information to the NCEP developers  for their input
with AWC.

Steve Chiswell
Unidata User Support



On Tue, 2007-03-13 at 07:42 -0500, Mike Bender wrote:
> Steve,
> 
> We did some checking of the AWC web site using their metar java tool.  
> GBD is showing as MVFR with the 5 mile vis and clr skies.  However, 
> using the Unidata "out of the box" version we continue to get VFR for 
> GBD.  Looks like they may have modified the code at AWC.
> 
> Mike
> 
> 
-- 
Steve Chiswell <address@hidden>
Unidata