Adding new features to WRF scripting and tutorial

models/wrf/shell_scripts/add_bank_perts.ncl

@@ -1,6 +1,6 @@
 ; Simple utility to add perturbation fields from the
 ; perturbation bank files to a provided wrfinput file.
-; consistent with use in the advance_model.csh script 
+; consistent with use in the advance_model.sh script 
 ; where we are substituting this call for running 
 ; da_wrfvar.exe, so the wrfinput file to perturb
 ; should be called 'wrfinput_next', and is the mean
@@ -16,15 +16,23 @@ begin
  ens_mem_num = ""+MEM_NUM
  pert_bank_path = PERTS_DIR
 
-; perturbation scaling:
+;perturbation scaling: default tutorial settings
  scale_T = 1.3
  scale_U = 1.3
  scale_V = 1.3
  scale_Q = 1.3
  scale_M = 0.6
 
+; Shuts off perturbations, only used for forecast mode
+;perturbation scaling:
+; scale_T = 0.0
+; scale_U = 0.0
+; scale_V = 0.0
+; scale_Q = 0.0
+; scale_M = 0.0
+
 ; for random pert, pick from larger bank
- bank_size = 100
+ bank_size = 60
  con  = bank_size / 32766.0      ; 32766.0 forces a 0.0 to 1.0 range
 ; get a seed
  wallClock = stringtoint(systemfunc("date -u +%N"))
@@ -45,7 +53,7 @@ begin
 
 
 ;For WRFv4 or later prognostic temp variable is THM
- pert_fields = (/"U", "V", "T", "QVAPOR","MU"/)
+ pert_fields = (/"U", "V", "THM", "QVAPOR","MU"/)
  wrf_fields  = (/"U", "V", "THM", "QVAPOR","MU"/)
  pert_scale  = (/scale_U,scale_V,scale_T,scale_Q,scale_M/)
  nperts = dimsizes(pert_fields)

models/wrf/shell_scripts/assim_advance.sh

@@ -0,0 +1,107 @@
+#!/usr/bin/env bash
+#
+# DART software - Copyright UCAR. This open source software is provided
+# by UCAR, "as is", without charge, subject to all terms of use at
+# http://www.image.ucar.edu/DAReS/DART/DART_download
+#
+# datea, emember, paramfile are command-line arguments - OR -
+# are set by a string editor (sed) command.
+
+set -uo pipefail
+
+datea="${1:-}"
+emember="${2:-}"
+paramfile="${3:-}"
+
+if [[ -z "${datea}" || -z "${emember}" || -z "${paramfile}" ]]; then
+  echo "usage: $0 <datea> <emember> <paramfile>"
+  exit 2
+fi
+
+source "${paramfile}"
+
+domains="${NUM_DOMAINS}"
+
+start_time="$(date +%s)"
+echo "host is $(hostname)"
+echo "assim_advance.sh is running in $(pwd)"
+
+cd "${RUN_DIR}"
+
+read -r -a gdate  < <(echo "${datea} 0 -g" | "${RUN_DIR}/advance_time")
+
+if (( ASSIM_INT_MINUTES <= 0 )); then
+  read -r -a gdatef < <(echo "${datea} ${ASSIM_INT_HOURS} -g" | "${RUN_DIR}/advance_time")
+else
+  read -r -a gdatef < <(echo "${datea} ${ASSIM_INT_MINUTES}m -g" | "${RUN_DIR}/advance_time")
+fi
+
+yyyy="${datea:0:4}"
+mm="${datea:4:2}"
+dd="${datea:6:2}"
+hh="${datea:8:2}"
+nn="00"
+ss="00"
+
+#  Copy files to appropriate location
+echo "${start_time}" > "${RUN_DIR}/start_member_${emember}"
+
+# Go into member directory and generate the needed wrf.info file
+cd "${RUN_DIR}/advance_temp${emember}"
+
+icnum="$(printf "%04d" "${emember}")"
+if [[ -e "${RUN_DIR}/advance_temp${emember}/wrf.info" ]]; then
+  ${REMOVE} "${RUN_DIR}/advance_temp${emember}/wrf.info"
+fi
+: > wrf.info
+
+if [[ "${SUPER_PLATFORM}" == "LSF queuing system" ]]; then
+
+  cat > "${RUN_DIR}/advance_temp${emember}/wrf.info" <<EOF
+${gdatef[1]}  ${gdatef[0]}
+${gdate[1]}   ${gdate[0]}
+$yyyy $mm $dd $hh $nn $ss
+          $domains
+mpirun.lsf ./wrf.exe
+EOF
+
+elif [[ "${SUPER_PLATFORM}" == "derecho" ]]; then
+
+  # module load openmpi
+  cat > "${RUN_DIR}/advance_temp${emember}/wrf.info" <<EOF
+${gdatef[1]}  ${gdatef[0]}
+${gdate[1]}   ${gdate[0]}
+$yyyy $mm $dd $hh $nn $ss
+           $domains
+ mpiexec -n 4 -ppn 4 ./wrf.exe
+EOF
+
+fi
+
+cd "${RUN_DIR}"
+
+# filter_control accounts for multiple domains
+# Appends input (filter_restart) and output (prior) in consecutive pairs
+# Should be consistent with filter_control setup for first_advance.sh
+# during intial forecast step
+
+echo "${emember}" > "${RUN_DIR}/filter_control${icnum}"
+
+dn=1
+while (( dn <= domains )); do
+  dchar="$(printf "%02d" "${dn}")"
+  echo "filter_restart_d${dchar}.${icnum}" >> "${RUN_DIR}/filter_control${icnum}"
+  echo "prior_d${dchar}.${icnum}"          >> "${RUN_DIR}/filter_control${icnum}"
+  (( dn++ ))
+done # loop through domains
+
+#  integrate the model forward in time
+"${RUN_DIR}/new_advance_model.sh" "${emember}" "${domains}" "filter_control${icnum}" "${paramfile}"
+${REMOVE} "${RUN_DIR}/filter_control${icnum}"
+
+end_time="$(date +%s)"
+length_time=$(( end_time - start_time ))
+echo "duration = ${length_time}"
+
+exit 0
+

models/wrf/shell_scripts/assimilate.sh

@@ -0,0 +1,59 @@
+#!/bin/bash
+#
+# DART software - Copyright UCAR. This open source software is provided
+# by UCAR, "as is", without charge, subject to all terms of use at
+# http://www.image.ucar.edu/DAReS/DART/DART_download
+
+# datea and paramfile are command-line arguments - OR -
+# are set by a string editor (sed) command.
+
+set -uo pipefail
+
+datea="$1"
+paramfile="$2"
+
+# shellcheck disable=SC1090
+source "$paramfile"
+
+start_time="$(date +%s)"
+echo "host is $(hostname)"
+
+cd "${RUN_DIR}"
+echo "${start_time}" > "${RUN_DIR}/filter_started"
+
+# Make sure the previous results are not hanging around
+if [[ -e "${RUN_DIR}/obs_seq.final" ]]; then
+  ${REMOVE} "${RUN_DIR}/obs_seq.final"
+fi
+if [[ -e "${RUN_DIR}/filter_done" ]]; then
+  ${REMOVE} "${RUN_DIR}/filter_done"
+fi
+
+#  run data assimilation system
+if [[ "${SUPER_PLATFORM}" == "LSF queuing system" ]]; then
+
+  export TARGET_CPU_LIST=-1
+  export FORT_BUFFERED=true
+  mpirun.lsf ./filter || exit 1
+
+elif [[ "${SUPER_PLATFORM}" == "derecho" ]]; then
+
+  export MPI_SHEPHERD=FALSE
+  export TMPDIR=/dev/shm
+
+  ulimit -s unlimited
+
+  mpiexec -n 256 -ppn 128 ./filter || exit 1
+
+fi
+
+if [[ -e "${RUN_DIR}/obs_seq.final" ]]; then
+  touch "${RUN_DIR}/filter_done"
+fi
+
+end_time="$(date +%s)"
+length_time=$(( end_time - start_time ))
+echo "duration = $length_time"
+
+exit 0
+

models/wrf/shell_scripts/diagnostics_obs.sh

@@ -0,0 +1,138 @@
+#!/bin/bash
+#
+# DART software - Copyright UCAR. This open source software is provided
+# by UCAR, "as is", without charge, subject to all terms of use at
+# http://www.image.ucar.edu/DAReS/DART/DART_download
+
+# diagnostics_obs.sh - shell script that computes observation
+#                       specific diagnostics.
+#
+# $1 - analysis date
+# $2 - parameter file
+#
+
+set -uo pipefail
+
+datea="${1:?usage: diagnostics_obs.sh YYYYMMDDHH paramfile}"
+paramfile="${2:?usage: diagnostics_obs.sh YYYYMMDDHH paramfile}"
+
+source "${paramfile}"
+
+cd "${OBS_DIAG_DIR}"
+${COPY} "${RUN_DIR}/input.nml" input.nml
+
+read -r -a gdate < <(echo "${datea} 0 -g" | "${DART_DIR}/models/wrf/work/advance_time")
+yyyy2="${datea:0:4}"
+mm2="${datea:4:2}"
+dd2="${datea:6:2}"
+hh2="${datea:8:2}"
+
+#  Determine appropriate dates for observation diagnostics
+datef="$(echo "${datea} -${ASSIM_INT_HOURS}" | "${DART_DIR}/models/wrf/work/advance_time")"
+#  Forcing the obs_diag_output.nc diagnostic to be the last analysis time only (not cumulative)
+yyyy1="${datea:0:4}"
+mm1="${datea:4:2}"
+dd1="${datea:6:2}"
+hh1="${datea:8:2}"
+
+half_bin=$(( ASSIM_INT_HOURS / 2 ))
+datefbs="$(echo "${datef} -${half_bin}" | "${DART_DIR}/models/wrf/work/advance_time")"
+fbs_yyyy1="${datefbs:0:4}"
+fbs_mm1="${datefbs:4:2}"
+fbs_dd1="${datefbs:6:2}"
+fbs_hh1="${datefbs:8:2}"
+
+datefbe="$(echo "${datef} ${half_bin}" | "${DART_DIR}/models/wrf/work/advance_time")"
+fbe_yyyy1="${datefbe:0:4}"
+fbe_mm1="${datefbe:4:2}"
+fbe_dd1="${datefbe:6:2}"
+fbe_hh1="${datefbe:8:2}"
+
+datelbe="$(echo "${datea} ${half_bin}" | "${DART_DIR}/models/wrf/work/advance_time")"
+lbe_yyyy1="${datelbe:0:4}"
+lbe_mm1="${datelbe:4:2}"
+lbe_dd1="${datelbe:6:2}"
+lbe_hh1="${datelbe:8:2}"
+
+while [[ "${datef}" -le "${datea}" ]]; do
+  if [[ -e "${OUTPUT_DIR}/${datef}/obs_seq.final" ]]; then
+    ${LINK} "${OUTPUT_DIR}/${datef}/obs_seq.final" "obs_seq.final_${datef}"
+  fi
+  datef="$(echo "${datef} ${ASSIM_INT_HOURS}" | "${DART_DIR}/models/wrf/work/advance_time")"
+done
+
+# Create flist (absolute paths)
+readlink -f obs_seq.final_* > flist
+
+cat > script.sed <<EOF
+  /obs_sequence_name/c\\
+  obs_sequence_name = '',
+  /obs_sequence_list/c\\
+  obs_sequence_list = 'flist',
+  /first_bin_center/c\\
+  first_bin_center =  ${yyyy1}, ${mm1}, ${dd1}, ${hh1}, 0, 0,
+  /last_bin_center/c\\
+  last_bin_center  =  ${yyyy2}, ${mm2}, ${dd2}, ${hh2}, 0, 0,
+  /filename_seq /c\\
+  filename_seq = 'obs_seq.final',
+  /filename_seq_list/c\\
+  filename_seq_list = '',
+  /filename_out/c\\
+  filename_out = 'obs_seq.final_reduced',
+  /first_obs_days/c\\
+  first_obs_days = -1,
+  /first_obs_seconds/c\\
+  first_obs_seconds = -1,
+  /last_obs_days/c\\
+  last_obs_days = -1,
+  /last_obs_seconds/c\\
+  last_obs_seconds = -1,
+  /edit_copies/c\\
+  edit_copies        = .true.,
+  /new_copy_index/c\\
+  new_copy_index     = 1, 2, 3, 4, 5,
+  /first_bin_start/c\\
+  first_bin_start    = ${fbs_yyyy1}, ${fbs_mm1}, ${fbs_dd1}, ${fbs_hh1}, 0, 0,
+  /first_bin_end/c\\
+  first_bin_end      = ${fbe_yyyy1}, ${fbe_mm1}, ${fbe_dd1}, ${fbe_hh1}, 0, 0,
+  /last_bin_end/c\\
+  last_bin_end       = ${lbe_yyyy1}, ${lbe_mm1}, ${lbe_dd1}, ${lbe_hh1}, 0, 0,
+EOF
+
+sed -f script.sed "${RUN_DIR}/input.nml" > input.nml
+
+# Create the state-space diagnostic summary
+"${DART_DIR}/models/wrf/work/obs_diag"
+${MOVE} obs_diag_output.nc "${OUTPUT_DIR}/${datea}/."
+${MOVE} "$(ls -1 observation_locations.*.dat | tail -1)" "${OUTPUT_DIR}/${datea}/observation_locations.dat"
+
+# Create a netCDF file with the original observation data (may not have some of the unusual metadata)
+"${DART_DIR}/models/wrf/work/obs_seq_to_netcdf"
+${MOVE} obs_epoch* "${OUTPUT_DIR}/${datea}/"
+${REMOVE} ./*.txt obs_seq.final_* flist observation_locations.*.dat
+
+# Prune the obs_seq.final and store result  keeps first 5 copies? why not set num_output_obs = 0
+# is it the time subsetting that is of interest?
+${LINK} "${OUTPUT_DIR}/${datea}/obs_seq.final" .
+"${DART_DIR}/models/wrf/work/obs_sequence_tool"
+${MOVE} obs_seq.final_reduced "${OUTPUT_DIR}/${datea}/."
+${REMOVE} obs_seq.final
+
+# Process the mean analysis increment
+cd "${OUTPUT_DIR}/${datea}"
+${COPY} "${SHELL_SCRIPTS_DIR}/mean_increment.ncl" .
+dn=1
+while (( dn <= ${NUM_DOMAINS} )); do
+   dchar="$(echo "${dn} + 100" | bc | cut -b2-3)"
+   analysis_in="analysis_increment_d${dchar}.nc"
+   mean_out="mean_increments_d${dchar}.nc"
+
+   ncl "fname=\"${analysis_in}\"" "fout=\"${mean_out}\"" "${OUTPUT_DIR}/${datea}/mean_increment.ncl" > nclrun_d${dchar}.out
+
+   (( dn++ ))
+done # loop through domains
+
+touch "${OUTPUT_DIR}/${datea}/obs_diags_done"
+
+exit 0
+