[Met_help] [rt.rap.ucar.edu #68042] History for query reg the calculation of FSS

[John Halley Gotway via RT]

----------------------------------------------------------------
  Initial Request
----------------------------------------------------------------

dear met help. 
this is a query reg the calculation of FSS. 
 
Firstly, I want to confirm if this would be the procedure. 
1. I need the gridded (same resolution and mapping) observed and model simulated rainfall. 
2. I have to run the grid stat tool for a particular configuration file (which will define the RF threshold and the radius within which the data is to be accepted). 
 
3. Now my question is, how will define the centre point around with the radius is defined (say 200km). Actually, I want to evaluate model's performance for a thunderstorm event at my station. Now how will I do that??? 
 
FSS is one of the O/p of the grid stat. Can U help me????
 
Your ideas reg this would be of great help. 

geeta
 		 	   		  

----------------------------------------------------------------
  Complete Ticket History
----------------------------------------------------------------

Subject: query reg the calculation of FSS
From: John Halley Gotway
Time: Fri Jul 04 11:12:35 2014

Geeta,

Yes, you're correct - Grid-Stat is the tool in MET that computes the
Fractions Skill Score.  It's contained in the NBRCNT (for neighborhood
continuous statistics) line type.  Fractions Skill Score is computed
in the
following way:

(1) Run Grid-Stat by passing it a forecast and observation field that
are
on exactly the same grid.
(2) For each threshold specified in the "cat_thresh" setting in the
config
file, apply the threshold to the forecast and observation fields to
define
0/1 bitmap fields.
(3) Next, for each entry in the "nbrhd.width" setting in the config
file,
let's call it n, do the following:
 - For each grid point in the forecast field, look in the nxn box
around
the grid point and count up the number of grid points with a 1 (i.e.
meaning the event occurred at that point).  Then compute what we call
a
"coverage" value for that grid point at the number of events divided
by
nxn.  For example, if 5 grid points in a 3x3 box have the event, the
coverage value is 5/9.
 - Do this for every grid point in the forecast field to define a
forecast
coverage field.
 - Apply the same logic to the observation field to define an
observation
coverage field.
(4) The fractions brier score (FBS) and fractions skill score (FSS)
are
computed over these fractional coverage fields.
(5) For each masking region defined in the "mask" section of the
config
file, compute FBS and FSS over the grid points in that region.

When you run Grid-Stat, you should see an output NBRCNT line for each
combination of field, categorical threshold, neighborhood width, and
masking region.

Hope that helps clarify.

Thanks,
John Halley Gotway
met_help at ucar.edu


On Wed, Jul 2, 2014 at 11:54 PM, Geeta Geeta via RT
<met_help at ucar.edu>
wrote:

>
> Wed Jul 02 23:54:55 2014: Request 68042 was acted upon.
> Transaction: Ticket created by geeta124 at hotmail.com
>        Queue: met_help
>      Subject: query reg the calculation of FSS
>        Owner: Nobody
>   Requestors: geeta124 at hotmail.com
>       Status: new
>  Ticket <URL:
https://rt.rap.ucar.edu/rt/Ticket/Display.html?id=68042 >
>
>
> dear met help.
> this is a query reg the calculation of FSS.
>
> Firstly, I want to confirm if this would be the procedure.
> 1. I need the gridded (same resolution and mapping) observed and
model
> simulated rainfall.
> 2. I have to run the grid stat tool for a particular configuration
file
> (which will define the RF threshold and the radius within which the
data is
> to be accepted).
>
> 3. Now my question is, how will define the centre point around with
the
> radius is defined (say 200km). Actually, I want to evaluate model's
> performance for a thunderstorm event at my station. Now how will I
do
> that???
>
> FSS is one of the O/p of the grid stat. Can U help me????
>
> Your ideas reg this would be of great help.
>
> geeta
>
>

------------------------------------------------