Open sebvi opened 6 months ago
https://github.com/wmo-im/CCT/wiki/Teleconference.16.May.2024 notes: Sebastien introduced the proposal; Sibylle commented that there should always be two bytes;
https://github.com/wmo-im/CCT/wiki/Teleconference.16.May.2024 notes: available for discussion probably for next fast-track;
we have an ecCodes branch if someone is interested to test our proposed implementation and give feedback on this: feature/ft2024-2_255
Hi @wmo-im/tt-tdcf ,
we haven't received comments or suggestions in relation to our prosposal for encoding verification scores. We are planning to go ahead with this during this fast track.
Really sorry for the delay.
I have a comment from one of our CCMEP experts:
In the table 4.120, it seems to be taken for granted that the contingency tables are 2x2 (hits, false alarms, misses and correct negatives). This seems restrictive to them. Shouldn't we consider swapping NxN tables whose 2x2 case is easy to obtain if need be by adding rows and columns so as to merge categories?
Thanks.
Dear @edouardsandrine ,
thank you for your comments. I passed it on to our verification experts. Do you have examples of contingency tables of higher dimensions than 2x2? The new Code Table 4.120 would be able to cater for several contigency tables if it was needed.
I will try to find such examples and come back to you. Thanks for your comment!
Please, find below some examples from a reference for the protocol for exchanging performance metrics between centers that provides examples of 3x3 and 4x4 contingency tables:
https://confluence.ecmwf.int/display/WLD/Data+format+of+reports+of+station-based+scores
https://github.com/wmo-im/tt-tdcf/wiki/2024.11.12.tt.tdcf notes: Sebastien believes that the proposal is ready for this FT; Jeff shared it with NCEP folks with no response so far; ECMWF uses these but currently in a different format
Initial request
ECMWF is interested in storing verification scores in GRIB2 in a near future.
With the existing GRIB2 templates, it is not possible to properly encode verification scores. The following templates are a proposal of how verification scores could be encoded in GRIB2.
We are not aiming to get this validated for FT 2024-2, we would like to start the discussion and eventually get it as experimental so that we can refine the proposal.
Amendment details
Note: The entries 7,8,9,10,11 and 12 need a threshold. The number of additional keys is equal to the number of arguments specified for the code number set regarding table 4.122 in the key ‘Kind of additional arguments for verification score’. For entries 11 and 12 and ensemble template must be used to define the total number of members and n.
Note: For #1: NA =1 or 2 with one scale factor / scaled value for a fixed threshold value if ‘type of limit for verification score’ includes only one value otherwise 2 for lower / upper limit
For #2: NA=1 or 2 1st tuple of scale factor / scaled value defined the total number of quantiles / 2nd tuple the quantile value if ‘type of limit for verification score’ includes only one value otherwise 4 for lower quantile / upper quantile
For #3: NA=1or 2 with scale factor / scaled value specifying the factor of standard deviation of the climatology if ‘type of limit for verification score’ includes only one value otherwise 2 for lower / upper limit
Comments
No response
Requestor(s)
Sebastien Villaume (ECMWF) Robert Osinski (ECMWF)
Stakeholder(s)
ECMWF
Publication(s)
Manual on Code: new templates for section 4 Manual on Code: new tables and new entries in code Table 4.2
Expected impact of change
None
Collaborators
No response
References
No response
Validation
No response