Modification to assimilate radar reflectivity simultaneously with the conventional data and to tune weights of mixed ensemble in EnVar

[shoyokota]

DESCRIPTION OF CHANGES:

This PR is to activate "tuning weights of MixEn" ( NOAA-EMC/GSI#541 ), "tuning cross-scale/variable/time covariances in SDL/VDL/TDL" ( NOAA-EMC/GSI#542 ), and "simultaneous radar reflectivity and conventional DA without side effects" ( NOAA-EMC/GSI#543 ). The following options are added. Here, "r_ensloccov4{scl,var,tim}" (real) was used instead of previous "i_ensloccov4{scl,var,tim}" (integer). In addition, the restriction forcing pure EnVar in assimilating radar reflectivity was removed.

• weight_ens_gfs: weight of global ensemble (default: 1.0)
• weight_ens_fv3sar: weight of regional (FV3-LAM) ensemble (default: 1.0)
• r_ensloccov4scl: factor multiplying to cross-scale localization (default: 1.0) (used only in nsclgrp>1)
• r_ensloccov4var: factor multiplying to cross-variable localization (default: 1.0) (used only in ngvarloc>1)
• r_ensloccov4tim: factor multiplying to cross-temporal localization (default: 1.0) (used only in l_timloc_opt=.true.)
• if_use_w_vr: if false, vertical wind (w) is not used in the observation operator of radar radial wind even w is included in a state vector (default: .true.)

TESTS CONDUCTED:

I checked that the following tests were as expectation on Orion.

• tuning weights of MixEn ("l_both_fv3sar_gfs_ens=.true.", "weight_ens_gfs=0.5", "weight_ens_fv3sar=0.5")
• simultaneous radar reflectivity and conventional DA with hybrid EnVar, and VDL with tuning cross-variable covariance ( "ngvarloc=2", "r_ensloccov4var=0.05", "if_use_w_vr=.false", "beta1_inv=0.15")

DEPENDENCIES:

Following PRs should be merged to GSI used in this workflow

• GitHub Issue NOAA-EMC/GSI#538 Add options to tune weights of mixed ensemble for FV3-LAM EnVar NOAA-EMC/GSI#541
• GitHub Issue NOAA-EMC/GSI#539 Add options to tune cross-scale/variable/time covariances in EnVar NOAA-EMC/GSI#542
• GitHub Issue NOAA-EMC/GSI#540 Modify to assimilate radar reflectivity and conventional data simultaneously without side effects in EnVar NOAA-EMC/GSI#543

DOCUMENTATION:

TBD

ISSUE (optional):

This PR is related to MixEn and SDL/VDL/TDL options introduced in #451

[chunhuazhou]

@shoyokota Are you going to change the workflow template (ush/templates/FV3LAM_wflow.xml) to have the option to run only one GSI task (ANAL_GSI_TN), in other words, to have the option to skip envar radar analysis (HYBRID_RADAR_REF_TN) when assimilating conventional and reflectivity simultaneously? We need to make changes to the workflow in order to run SDL+VDL+HYB, I think. Something like DO_SDL_VDL_HYB = TRUE (or FALSE) in the configure file and also in the workflow and scripts would be helpful for determining which tasks to run and what kind of dependencies are needed. My two cents here.

[shoyokota]

@chunhuazhou Such an option has already been in the current workflow as "DO_ENVAR_RADAR_REF_ONCE" since #451. In activating VDL (ngvarloc=2), this option is set to TRUE in set_rrfs_config_SDL_VDL_MixEn.sh, and skips HYBRID_RADAR_REF_TN.