allobs Verification

Introduction

It is possible to use Screening (NCONF=002) to calculate observation - forecast (O-F) statistics using forecasts of any length as the model first-guess. The ...

Screening adjustments

The screening of observations is switched off by adjusting the NAMCT0 and L_SCREEN_CALL

&NAMCT0
  L_SCREEN_CALL=.FALSE.,
/

Inputs

The inputs for the allobs data extraction are the same as for a regular DA Screening task - observations (ECMA) and a first-guess (short-forecast files - ICMSHHARM+hhhh and ICMSHHARM+hhhh.sfx).

Note

The forecasts being verified will need both model state (ICMSHHARM+hhhh) and the "full" SURFEX file (ICMSHHARM+hhhh.sfx) available for the Scextr task. You may need to adjust the VERITIMES and SWRITUPTIMES settings in your ecf/config_exp.h file.

The following settings are important:

`config_exp.h` setting
`SCREXTR`	Use Screening (`NCONF=002`) to produce O-F data
`SCREXTR_TASKS`	Number of parallel tasks for O-F extraction
`FGREFEXP`	Experiment name for FirstGuess. If set to `undef` it will use own forecasts
`OBREFEXP`	Experiment name for ODBs. If set to `undef` it will use own ODBs

Running

This extraction can be executed as part of a running experiment (with SCREXTER=yes) or using a standalone suite (PLAYFILE=allobsver).

Output

The output from the Scrextr task is a CCMA ODB with O-F statistics. This ODB is archived in $HM_DATA/archive/extract/obsver/odb_ver_${FGDTG}_${FCLENSTR}/ where FGDTG is the forecast cycle DTG and FCLENSTR is the forecast length verified. The ODB data is then converted to ODB-2 and sqlite files for use in Harp and other downstream applications using odbcon tools.