Spatial scale dependency of air-sea coupling

There's a strong spatio-temporal scale dependency of air-sea coupling, and a myraid of methods to separate those scales. Here, we take a few examples of air-sea coupling coefficients and produce global maps of them for full fields, smoothed fields and high-pass filtered fields.

Coupling coefficients

Coupling coefficients based on regression are used for measuring strength of air-sea coupling (Renault et al., 2016), particularly for thermal feedback (TFB) and current feedback (CFB). Additionally, other quantities can illustrate air-sea coupling and associated mesoscale features/implications.

TFB:

1. SST vs wind/stress magnitude
2. downwind SST gradient (winds) vs wind divergence
3. downwind SST gradient (stress) vs stress divergence
4. crosswind SST gradient (winds) vs wind curl
5. crosswind SST gradient (stress) vs stress curl

CFB:

1. surface current vorticity vs wind curl
2. surface current vorticity vs stress curl
   => wind curl vs stress curl

Other quantities:

• wind work
• surface KE
• surface geostrophic KE and vorticity
• eddy-induced Ekman pumping (Wek): curl, vortgrad, Stern, classical)

Model and technical details

• We use 7-years of ICON outputs from EERIE run (erc1011)
• Spatial filtering uses cdo operator smooth, a simple filter with diminishing weights from 1 to 0.2 for a specified filtering length (say 3deg)

Methodology

1. Extract and compute daily quantities
2. Remap onto regular 0.25deg grid
3. Perform 30-day running mean on 0.25deg grid files
4. Spatially high-pass 30-day running mean on 0.25deg grid files
5. Compute coupling coefficients

1) Extract and compute daily quantities

• Stress magnitude
  • taumag_erc1011.job
• Wind divergence and curl
  • create_daily_winddivcurl_erc1011.py
• Downwind SST gradients (stress), crosswind SST gradients (stress), stress divergence, stress curl
  • create_daily_SSTgrad_TAUdivcurl_erc1011.py
  • create_daily_taucurl_erc1011.py
  • submit_create_dm_SSTgradTAUdivcurl.job
• Surface vorticity
  • create_daily_sfcvort_erc1011.py
• Surface KE
  • sfcKE_erc1011.job
• Surface geostrophic KE and vorticity
  • create_daily_sfcgeoKE_erc1011.py
  • create_daily_sfcgeovort_erc1011.py
  • exec_create_dm_geostrophic.job
• Wind work
  • windwork_erc1011.job
  • eddywindwork_erc1011.sh
• Eddy-induced Ekman pumping (Wek): curl, vortgrad, Stern, classical)
  • create_daily_EkmanPumping_erc1011.py
  • exec_create_dm_EkmanPumping.job

Get daily SST, wind speed, turbulent fluxes

Extract SST, wind speed at 10m, LHF and SHF using intake_dailyr2b9_annual.job and submit_intake_dailyr2b9_annual.sh

Script includes time shift needed for these variables due to ICON's way of time-stamping output data.

Wind stress magnitude

Compute daily wind stress magnitude using taumag_erc1011.job

for yr in $(seq 2002 2008); do for mth in $(seq 1 12); do sbatch taumag_erc1011.job ${yr} ${mth}; done; done

Surface wind divergence and curl

uas and vas are outputted on atm grid, not ocean grid. Since ocean has fewer points than atm, we can regrid and compute divergence and curl on ocean grid. Then interpolate onto 0.25deg grid, then get 30-day running mean, then spatially filter.

To remap from atm to oce grid, use remap_dm_yrmth_r2b8G_r2b9O.job

for yr in $(seq 2002 2008); do for mth in $(seq 1 12); do sbatch remap_dm_yrmth_r2b8G_r2b9O.job uas ${yr} ${mth}; done; done
for yr in $(seq 2002 2008); do for mth in $(seq 1 12); do sbatch remap_dm_yrmth_r2b8G_r2b9O.job vas ${yr} ${mth}; done; done

Compute daily wind divergence and curl using create_daily_winddivcurl_erc1011.py and submit_create_dm_winddivcurl.job

for yr in $(seq 2002 2008); do for mth in $(seq 1 12); do sbatch submit_create_dm_winddivcurl.job ${yr} ${mth}; done; done

Downwind and crosswind SST gradients from stress plus stress divergence and curl

Compute these quantities using create_daily_SSTgrad_TAUdivcurl_erc1011.py , create_daily_taucurl_erc1011.py and submit_create_dm_SSTgradTAUdivcurl.job

for yr in $(seq 2002 2008); do for mth in $(seq 1 12); do sbatch submit_create_dm_SSTgradTAUdivcurl.job ${yr} ${mth}; done; done

Surface vorticity

Use create_daily_sfcvort_erc1011.py and exec_create_dm_sfcvort.job

for yr in $(seq 2002 2008); do for mth in $(seq 1 12); do sbatch exec_create_dm_sfcvort.job ${yr} ${mth}; done; done

Surface KE

Use sfcKE_erc1011.job

for yr in $(seq 2002 2008); do for mth in $(seq 1 12); do sbatch sfcKE_erc1011.job ${yr} ${mth}; done; done

Surface geostrophic KE and vorticity

Use create_daily_sfcgeoKE_erc1011.py, create_daily_sfcgeovort_erc1011.py and exec_create_dm_geostrophic.job

for yr in $(seq 2002 2008); do for mth in $(seq 1 12); do sbatch exec_create_dm_geostrophic.job ${yr} ${mth}; done; done

Wind work

Use windwork_erc1011.job

for yr in $(seq 2002 2008); do for mth in $(seq 1 12); do sbatch windwork_erc1011.job ${yr} ${mth}; done; done

Eddy-induced Ekman pumping

Ekman pumping (Wek) has classical vs Stern (1965), where Stern = curl + vortgrad

Use create_daily_EkmanPumping_erc1011.py and exec_create_dm_EkmanPumping.job

for yr in $(seq 2002 2008); do for mth in $(seq 1 12); do sbatch exec_create_dm_EkmanPumping.job ${yr} ${mth}; done; done

2) Remap onto regular 0.25deg grid

Use remap_dm_yrmth_r2b9O_IFS25.job and submit_remapIFS25.sh

./submit_remapIFS25.sh

If files were created with remap_dm_yrmth_r2b9O_IFS25.job, then they need to be stitched together with stitchfiles.sh

./stitchfiles.sh

3) Perform 30-day running mean on 0.25deg grid files

Use 30dayrunmean_erc1011_IFS25.job

declare -a varnamearr=("to" "Wind_Speed_10m")
declare -a varnamearr=("taumag" "sfcvort" "windwork")
declare -a varnamearr=("winddiv" "windcurl")
declare -a varnamearr=("downSSTgrad" "taudiv" "crossSSTgrad" "taucurl")
declare -a varnamearr=("Wekcurl" "Wekvortgrad" "Wekstern" "Wekclassic")

for varname in "${varnamearr[@]}"
do
	sbatch 30dayrunmean_erc1011_IFS25.job ${varname}
done

4) Spatially high-pass 30-day running mean

Use sm_30dayrunmean_dailyIFS25_yrmth.job and submit_sm30dayrunmean.sh

./submit_sm30dayrunmean.sh

5) Compute coupling coefficients

Global maps of coupling coefficients are evaluated via temporal regressions for a given season (JJA, DJF). Standard errors are also computed to provide 95% confidence bounds to the coupling coefficients. We took a conservative estimate of 40 for effective degrees of freedom.

Please see regress_stderror.sh for details.

Intermodel Comparison

• ICON
• IFS/FESOM
• HadGEM = UM/NEMO
• IFS/NEMO

Perhaps each model component needs to be done one at a time.

Verification with observations/reanalysis

???