Vertical ocean heat transport near Antarctic ice shelves: data and processing code

Antarctic ice shelves, or the floating extension of terrestrial glaciers, help limit Antarctica’s contribution to global sea level rise by slowing the export of grounded ice into the Southern Ocean. Identifying the distribution of meltwater from the ocean-driven melt of these ice shelves helps determine where and at what rate they are thinning, providing insight into how quickly their buttressing effect is diminishing. In regions with rapid ice shelf melt, the resulting freshwater flux into the coastal ocean contains enough buoyancy forcing to reshape coastal currents and alter ocean stratification up to thousands of kilometers from the meltwater’s source. Global circulation models are employed to help understand how ice shelf meltwater modifies the Southern Ocean and are extended forward in time to attempt to predict the consequences of a future with increased ice melt. However, because they are unable to efficiently resolve small scale turbulent mixing, these models make assumptions regarding the mechanisms that drive the mixing and use parameterizations to simplify the effects of turbulence, such as the diffusion rates heat and salinity. While common turbulence parameterizations have been found to work well throughout much of the Earth’s oceans, the lack of in situ turbulence measurements near Antarctica have left parameterizations comparatively uniformed and untested near melting ice shelves. This dissertation works to improve the understanding of the downstream effects of ice shelf meltwater via the collection of rare in situ oceanographic measurements near Antarctic ice shelves.

Here, we present the identification and analysis of meltwater layers within the Southern Ocean near two Antarctic ice shelves. At Nansen Ice Shelf, traces of meltwater in the form of frigid Ice Shelf Water were observed at the opening of the ice shelf cavity, and vertical heat flux estimates derived from direct measurements of microstructure turbulence show approximately 10 W m-2 of heat transport into the meltwater from both above and below. The position of this layer within the water column was heavily modified by the presence of a coastal, submesoscale eddy, which lifted the meltwater toward the ice shelf-ocean interface, potentially resulting local variations in basal melt rate. On the other hand, at Dotson Ice Shelf, ice shelf melt in the form of warmer Glacial Meltwater was observed across two layers in front of the ice shelf cavity at quantities not observed for over a decade. These meltwater layers were determined to be largely derived from different sources, with the deeper Subglacial Meltwater layer originating from beneath the western corner of Dotson Ice Shelf, and the shallower Ventilated Meltwater layer originating from the previous melt of other upstream ice shelves. Here, additional but more limited measurements of turbulence were used to validate modeled energy dissipation rates, which then again indicate approximately 10 W m-2 of heat transport through the meltwater: largely upwards through the deeper layer and downwards through the shallower layer. Our results show that the meltwater layers resulted in the formation of step-like changes temperature stratification, which support particularly efficient mixing. Together, the observations from these two unique ice shelf systems show that melt from ice shelves is intrinsically tied to local oceanography, and that our ability to predict how future changes in the Antarctic cryosphere will affect global thermohaline circulation relies on the accuracy of well-informed ocean mixing parameterizations.

Methods
In Chapter 2, ship-based conductivity-temperature-depth (CTD) and lowered acoustic Doppler current profiler (LADCP) data was collected via in situ profiling aboard the Research Vessel Ice Breaker (RVIB) Araon. Seafloor bathymetry data was as produced by multi-beam sonar aboard the research cruise. Glider CTD and MicroRider (MR) data was collected in situ across two missions. Ice shelf bathymetry data is from Dow et al. (2018). Sea ice data was derived from Moderate Resolution Imaging Spectroradiometer (MODIS) imagery via the National Aeronautics and Space Administration (NASA) Worldview platform. Wind data were recorded by Automatic Weather Station (AWS) Manuela, part of the University of Wisconsin-Madison AWS Program.

In Chapter 3 and Chapter 4, ship-based vertical microstructure profiler data (VMP), again along with CTD and LADCP data, were collected via in situ profiling aboard the RVIB Araon. Glider-based CTD data were collected in situ across a single mission. Historical CTD and LADCP data were sourced from Jenkins et al. (2018). Landsat imagery was provided courtesy of the U.S. Geological Survey. Sea ice data was derived from Nimbus-7 Passive Microwave Data via the NASA Snow and Ice Data Center Distributed Active Archive Center (Cavalieri et al., 1996). Wind data was derived from ERA5 reanalysis data (Hersbach et al. 2023).

In Appendix C, multibeam bathymetry, CTD, LADCP, and vehicle-mounted acoustic Doppler current profiler (VM-ADCP) data were all collected via surveying on the Research Vessel Frantz. Further details regarding the CTD and VM-ADCP datafiles are provided in the CTD_metadata.xlsx and ADCP_metadata.xlsx spreadsheets. Lake Shasta elevations and Shasta Dam flowrates were sourced from the California Data Exchange Center.

All data processing for the main chapters, and the majority of processing for Appendix C, took place using MATLAB (version 2024b). In Chapter 2, ship-based CTD and LADCP data was processed using the process_ship.m script, which combines all of the raw ship data files into the ship.mat datafile for this chapter. MR data was processed using the process_MR.m script, which processes all of the MR data files in order, resulting in the MR.mat datafile. Finally, glider data was processed with process_glider.m, which combines data from the two glider missions into a single glider.mat data file for this chapter.

In Chapter 3 and Chapter 4, VMP data was processed using the process_VMP.m script, which only processes one raw VMP datafile at a time and thus requires multiple runs with manual user edits (change input filename on line 6) between each run. Ship-based CTD and LADCP data was processed using another process_ship.m script, which combines all of the raw ship data files into the ship.mat datafile for these chapters. Ship-based data was reprocessed to match the format of Jenkins et al. (2018) data with the process_2022_for_Jenkins.m script, the outputs of which are saved as Dotson2022.mat. The process_Jenkins_data.m script then calculates meltwater transport for each study year (variables data_2000, data_2007, data_2009, etc.) which are together saved as all_data.mat. Glider data was processed using another process_glider.m script, which produces the glider.mat datafile for this chapter. Finally, ship- and glider-based turbulence model results were produced with the process_ship_model.m and process_glider_model.m scripts, respectively.

In Appendix C, VM-ADCP data were processed using process_transect.m. This script also only processes one data transect or data station at a time, and for each transect/station, requires multiple iterations with manual user edits each time. A first iteration requires the filename to be chosen (line 8). The indices of bad data (line 10), indices of Head Tower interference (line 12), and cutoff depth (line 16) should start as empty arrays. The padding for dam interference should start as 0. After a first run, the output figure is examined and bad indices (line 10) are chosen from identifying major outliers within the velocity data (u, v, w, and speed subplot columns). Head Tower interference indices (line 12) are chosen from examining the amplitude and correlation subplots. A buffer for dam interference (line 14) is chosen if high velocities near the dam are still visible along the deepest velocity data after Step 6 (and carried through remaining steps). A depth cutoff (line 16) is chosen to remove the high velocity spikes at the bottom of the average velocity profile (far right subplot column). If any of these lines (10, 12, 14, 16) were changed, the script is rerun before the output (variable ADCP) can be saved (e.g., as A24_01_T01.mat, A24_01_T02.mat, etc.). All processed data from a single data group were saved together (e.g., as A24_01_processed.mat).

See the readme for more information.