[Met_help] [rt.rap.ucar.edu #97842] History for HIRA not Running

[John Halley Gotway via RT]

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

John, I am verifying our meoscale ensemble over a small domain in Africa (about 5 - 10 observations).  I have HIRA turned on in the config file (MET 8.1):

//
// HiRA verification method
//
hira = {
   flag       = TRUE;
   width      = [ 1, 5 ];
   vld_thresh = .8;
   cov_thresh = [ ==0.25 ];
   shape      = SQUARE;
}

But the data output does not show any HIRA neighborhoods in the output.  We are also running WRF over that domain and with the same HIRA settings I am getting neighborhood data in the output.  Looking at the meps listing I get with verbosity set to four, there is no HiRA information in the file.  Is HiRA allowed on probability data?  We did this in the past I believe.   I can send the data set if needed.

Bob
Thanks




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

Subject: HIRA not Running
From: John Halley Gotway
Time: Wed Dec 09 17:33:44 2020

Bob,

I see that you're not getting any HiRA output from Point-Stat, when
you
expected to. And you asked this question:
"Is HiRA allowed on probability data?"

Checking in the code, I see on this line of point_stat.cc that HiRA is
explicitly NOT applied when the input is a probability field:
https://github.com/dtcenter/MET/blob/17e1c9740224bf84a8bd0e84b5d254ccc5981bb8/met/src/tools/core/point_stat/point_stat.cc#L1237

For a deterministic forecast, the HiRA methodology processes the
forecast
values near the observation point in two ways... as an ensemble of
deterministic forecasts... and also by applying a threshold to derive
a
HiRA probability forecast.

If the input is itself a probability field, then HiRA would give us an
ensemble of probability forecasts. And MET does not include logic for
verifying an ensemble of probabilities. (Instead, we typically use an
ensemble of deterministic forecasts to derive probabilities.) In
addition,
we'd be thresholding those neighborhood probabilities to define a HiRA
probability!

So that's why you don't see HiRA output when evaluating probability
forecasts in Point-Stat. If, for some reason, it does make sense to
you to
process the forecast probabilities in this way, you could explicitly
tell
MET to not interpret the input as a probability forecast by setting:
   prob_as_scalar = TRUE;

Search for "prob_as_scalar" on this page for more details:
https://dtcenter.github.io/MET/latest/Users_Guide/data_io.html#configuration-
file-details

Hope that helps clarify.

Thanks,
John


On Wed, Dec 9, 2020 at 1:30 PM George McCabe via RT
<met_help at ucar.edu>
wrote:

>
> Wed Dec 09 13:29:06 2020: Request 97842 was acted upon.
> Transaction: Given to johnhg (John Halley Gotway) by mccabe
>        Queue: met_help
>      Subject: HIRA not Running
>        Owner: johnhg
>   Requestors: robert.craig.2 at us.af.mil
>       Status: new
>  Ticket <URL:
https://rt.rap.ucar.edu/rt/Ticket/Display.html?id=97842 >
>
>
> This transaction appears to have no content
>

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