Data used in "Southern Ocean summer warming is regulated by storm-driven mixing"

The data included in this repository was used to generate the figures in the submitted manuscript "Storms regulate Southern Ocean summer warming" by du Plessis and co-authors.

Abstract: "Sea surface temperature (SST) in the Southern Ocean (SO) is the fingerprint of ocean heat uptake and critical for air-sea interactions. However, SO SST is biased warm in climate models, reflecting our limited understanding of the mechanisms that set its magnitude and variability. An important factor driving SST variability is synoptic-scale weather systems, such as storms, yet their impacts are difficult to directly observe. Using in-situ observations from underwater and surface robotic vehicles in the subpolar SO, we show evidence that storms regulate the summer evolution of SST through altering the mixed layer effective heat capacity and entraining colder water from below. Through these mechanisms, we determine that interannual variations in SO SST reflect changes in storm intensity and prevalence, which, in turn, are driven by the Southern Annular Mode. Our results demonstrate a causal link between storm forcing and lower frequency SST variability, which has implications for addressing SST biases in climate models."

Datasets

The observations in this study were made as a part of the SOSCEx-STORM experiment, which fits into the larger observational programme the Southern Ocean Seasonal Cycle Experiment (Swart et al. 2012). SOSCEx-STORM undertook a twinned deployment of a Wave Glider and a profiling Slocum glider which were piloted in conjunction with each other. The platforms were deployed and retrieved from the R/V Agulhas II at 54°S, 0°E, south of the Polar Front, and sampled together between 20 December 2018 and 8 March 2019. 

Slocum glider data
The glider was equipped with a continuously pumped Seabird Slocum Glider CTD, which was processed with the GEOMAR MATLAB toolbox and vertically gridded to 1 m depth intervals. 

Relevant data name: slocum_grid_processed.nc

Slocum glider Microstructure data:
The Webb Teledyne G2 Slocum glider was equipped with a Rockland Scientific Microstructure Profiler (MicroRider). The MicroRider was equipped with two piezo-electric accelerometers and two air-foil shear probes oriented orthogonally. Microstructure data was only collected during the glider climbs to prolong battery life and obtain dissipation estimates as close to the surface as possible. See Nicholson et al. (2022) for details of the MicroRider processing. The mixing layer depth (XLD) was estimated as in Brainnerd and Gregg et al. (1995).

Disspitation data name: slocum_eps.nc
Mixing layer depth data name: slocum_xld.nc

Slocum glider SST data: Initial data processing removed temperature data from the upper 2 m during the glider climb phase, and so to obtain an SST value from the Slocum glider temperature profiles, we calculated the median value between 0.5 m and 10 m depth for each dive.  

Slocum SST data name: slocum_sst_median_10m.nc

Wave Glider data
The Liquid Robotics SV3 Wave Glider was fitted with an Airmar WX-200 Ultrasonic Weather Station mounted on a mast at 0.7 m above sea level, providing wind speed measurements at a rate of 1 Hz, averaged into 1-hour bins. The wind measurements were corrected to a height of 10 m above sea level. Note that the Airmar WX-200 weather station of the Wave Glider was faulty and the wind speed, wind direction and wind stress data was replaced by hourly ERA5 data. 

Wave Glider data name: WG_era5_1h_processed_28Aug2022.nc

NOAA OI SST and sea ice
Monthly SST data was obtained from the NOAA optimum interpolation (OI) SST V2 product, which uses both in-situ and satellite data from November 1981 to January 202329. Data is provided by the National Centers for Environmental Prediction and made available on a 1◦ grid. All SST data where co-located sea ice concentration was above 0 has been removed from this analysis. NOAA OI SST and sea ice were obtained from https://psl.noaa.gov/data/gridded/data.noaa.

Datasets: sst.mnmean.nc, icec.mnmean.nc, lsmask.nc

Storm tracking dataset
To track storm trajectories, we used storm tracks contained in monthly files for the Southern Ocean identified and used in the JGR-Oceans publication:

Lodise, J., Merrifield, S. T., Collins, C., Rogowski, P., Behrens, & J., Terrill,E, (In Review). Global Climatology of Extratropical Cyclones From a New Tracking Approach and Associated Wave Heights from Satellite Radar Altimeter. Journal of Geophysical Research: Oceans. https://doi.org/10.1029/2022JC018925

Data can be accessed at https://github.com/jlodise/JGR2022_ExtratropicalCycloneTracker 

All Southern Ocean storm locations can be found at: ec_centers_1981_2020.nc

Storm radius datasets

ERA5 data of air-sea heat flux, 10 m wind speed for all hourly instances where a storm center was within 1000 km of the gliders (Figs. 2 and 3, Extended Data Figs. 2 and 4) 

Datasets of combined_storms_{variable}_no_ice.nc are data of air-sea heat flux and 10 m wind speed for all instances for 1000 km x 1000 km box around each storm center during summer months from 1981-2020 (> 570,000) used in Figs. 4 and 5. These data are considerably large (combined total >40 GB). Please contact me at marcel.du.plessis@gu.se to find a suitable way to share the data.

The data was processed as follows:

1. Download storm centers from https://github.com/jlodise/JGR2022_ExtratropicalCycloneTracker/tree/main

2. Run process-lodise-storm-centers.ipynb to save all the cyclone center data as 'ec_centers_1981_2020.nc'

3. Run filter-cyclone-centers.ipynb:
    - cut all cyclone centers south of 40S
    - only choose cyclone centers in DJF
    - calculates minimum distance to land for each storm
    - saves a dataset called 'ec_centers_1981_2020_with_min_dist_to_land.nc'  
        - contains the variable distance to land for all filtered cyclones
        - we do this step because it takes about an hour to run
    - remove cyclones within 500 km from land
    - remove cyclones less than 24 hours
    - we are left with 11005 storms
    - data saved as 'ec_centers_1981_2020_500km_from_land_filtered_24hours'

4. Run storm_processing_cutouts.ipynb (this took several days)
    - loads ec_centers_1981_2020_500km_from_land_filtered_24hours.nc
    - runs through each summer, processes storm_localization.py 
    - saves data as storms_{variable}_{year}.nc
        - e.g. storms_winds_1981.nc - winds for all 1000 km radius cyclones in DJF 1981/82

5. combine_storm_years.ipynb 
    - creates datasets for each variable that has storms for all year called combined_storms_{variable}.nc

6. remove_sea_ice_from_storms.ipnyb 
    - makes data nan where sea ice is present in each cyclone
    - creates datasets called combined_storms_{variable}_no_ice.nc

7. seasonal-means-storms.ipynb
    - calculates the mean for all storms for each year 
    - saves them one dataset: combined_storms_{variable}_seasonal_means.nc'

EN4 mixed layer depths
We use the EN4 database of quality controlled temperature and salinity profiles from 2004 to 2022 to produce our MLD for the interannual analysis (Good et al. 2013). We use the profiles that contain the Cheng et al. (2014) XBT corrections and Gouretski and Cheng (2020) MBT corrections. We limit the data intake to 2004 as this marks the beginning of the Argo period. All under-ice profiles are removed. We calculate the MLD for each individual profile using the density threshold of de Boyer Montegut et al. (2004) where the density value first exceeds the 10 m reference value by 0.03 kg m-3. We then determine the median MLD value for each month within 3 x 3 degree grid cells, then obtain a mean value for each DJF season per 3 x 3 degree grid cell. 

Relevant data name: en4_monthly_mixed_layer_depth_median.nc

Southern Ocean Fronts
Position of the Subantarctic Front and Polar Front are from: