Heuzé et al. (2021) Spaceborne infrared imagery for early detection of Weddell Polynya opening

Extract from Table B1 of Heuzé et al. (2021): characteristics of the 30 polynyas. Latitude (lat) and longitude (lon) are in degrees N and E respectively; maximum area in km2; duraction in days.

C. Heuzé , L. Zhou, M. Mohrmann, and A. Lemos (2021) Spaceborne infrared imagery for early detection of Weddell Polynya openings, The Cryosphere, vol. 15, pp. 3401–3421, doi:10.5194/tc-15-3401-2021.

We use daily infrared satellite data since 1982 to investigate the Weddell Polynya, an opening in the Antarctic winter sea ice. We find:

  • That although the usual narrative is that the Weddell Polynya opened once over 1974-1976, and then did not re-open until 2016, there were in fact 30 polynyas in our dataset.
  • That our algorithm could detect up to 15 days in advance that an opening was imminent, returning no false positive.
  • By comparing infrared temperature to in-situ and reanalysis data, that variations in specific infrared properties can indicate whether the polynya opens in response to upwelling or a lead.

Infrared data however are strongly affected by clouds, which are very common in Antarctica in winter. We therefore suggest that infrared data be used after the opening, to obtain scientific information, but not for operational purposes (at least, not without extra data).

Download the full-text here.

Heuzé (2021) Antarctic Bottom Water and North Atlantic Deep Water in CMIP6 models

C. Heuzé (2021) Antarctic Bottom Water and North Atlantic Deep Water in CMIP6 models, Ocean Sci., doi:10.5194/os-17-59-2021, vol 17, pp 59-90.

CMIP5 models were rather biased when it came to their deep and bottom water properties, formation, and transports (see my own research [1],[2],[4],[10] and [17]). Are the new CMIP6 models better? I looked at 35 CMIP6 models, for the last 30 years of their historical run (1985-2014), both in the Southern Ocean and the North Atlantic. And as usual, the performance of the models depends on what you are interested in:

  • More models seem to successfully, realistically produce Antarctic Bottom Water (AABW) via shelf overflow.
  • Yet, most CMIP6 models still have unrealistically large areas with open ocean deep convection, both in the Southern Ocean and in the North Atlantic. “Too deep, too often, over too large an area” remains the best description for CMIP6 models.
  • Bottom property biases have notably decreased. The most biased model in particular, INM-CM5, is in particular much better than its predecessor INM-CM4. The most accurate models are those from the CESM2 family, which feature an overflow parameterisation.
  • The link between deep water formation and bottom properties depends on the region. In the Southern Ocean, more deep convection = more biased bottom waters; in the North Atlantic, less biased.
  • Most models have a warm bias, which may be a result of the reference we chose to compare them to rather (WOA2018).
  • Regarding their transport, the AMOC is within the observational range for most CMIP6 models, which is a notable improvement since CMIP5. Southern MOC observations still are too few, but most CMIP6 models now lie within their range.
  • In the Atlantic, the spread of the water masses in CMIP6 models is controlled by the strength of the MOCs: the stronger the AMOC, the further south we detect NADW and the least AABW can spread northward.
  • In the Indian and Pacific Ocean, the northward spread of AABW is not linked to the MOC but to the properties, in particular the salinity gradient in the Antarctic Circumpolar Current: the weaker the gradient to overcome, the further the AABW spread.

I encourage you to check the performances of your favourite model in the many tables of the paper!

The INM model still is way too salty, but no longer needs its own colour bar. Black contours indicate a maximum MLD deeper than 2000 m (from supp. fig. A2, Heuzé 2021)

Download the full-text here.

Research theme: Sea level and flooding

Project “Would the Northern European Enclosure Dam really protect Sweden from sea level rise? (NEEDS)”, funded by FORMAS grant 2020-00982
Ca 4 million SEK; I am the PI.
January 2021 – December 2024

The aim of the project is to determine what causes flooding in Sweden: remote effects that could be blocked out by distant sea walls, or local effects such as precipitation or tides. We’re using a combination of all data sources (in-situ, remote sensing, models) that we will analyse with help from Machine Learning.

Team:

  • Myself (mostly for supervision);
  • David Ek, Master’s student, April 2021 – April 2022;
  • Lea Poropat, Postdoc, starting June 2021;
  • Heather Reese;
  • Collaboration with Dan Jones and Scott Hosking from the British Antarctic Survey AI Lab.

Publications:

Nothing yet, the project has not even officially started!

Research theme: AABW in CMIP5 models

NERC Open CASE, number 1093171, awarded to Karen Heywood.
= My PhD “Antarctic Bottom Water in CMIP5 models: characteristics, formation, evolution”
October 2011 to March 2015.
Supervisors: Karen Heywood and David Stevens (UEA), and Jeff Ridley (UK MetOffice)

Outcomes of my PhD:

  • Quantified biases in Antarctic Bottom Water temperature and salinity in CMIP5 models [1][2];
  • Determined the causes for these biases: impossibility to export shelf water due to the mixing scheme, and open ocean deep convection in the Weddell Polynya[1][PhD thesis];
  • Quantified the consequence of these biases on climate change projections [3];
  • Found a numerical solution to suppress the unrealistic formation process [4].

Research theme: Greenland and the North Atlantic

Project “Is Greenland meltwater going to stop the Atlantic overturning circulation?”, funded by VINNMER-Marie Curie, dnr 2015-01487.
Ca 2.5 million SEK; I was the PI.
Started in July 2015, finished in June 2018.

Team:

  • Myself, as postdoc researcher;
  • Anna Wåhlin, Gothenburg University, and Helen Johnson, Oxford University, as mentors / hosts;
  • Master student Lovisa-Waldrop Bergman.

Outcomes and publications:

  • Collected hydrographic data and produced the first map the  path of meltwaters from Petermann Glacier out of its fjord into Nares Strait [8];
  • Modelled the oceanic circulation of Nares Strait and determined its sensitivity to initial conditions (Master’s student’s work);
  • Quantified biases in full depth water properties and deep water formation processes in the North Atlantic in CMIP5 models [10];
  • Determination of ocean currents from infrared remote sensing – collaboration with Chalmers University of Technology to try and increase the North Atlantic data coverage [6][9][11].

Research theme: Antarctic polynyas

Project “Warm oceanic Inflows for Near-real time Detection Of Weddell polynya from Space (WINDOWS)”, funded by Rymdstyrelsen grant 164/18
Ca 4.5 million SEK; I am the PI.
Started in January 2019, finishes in December 2022.

The aim is to create an “early warning system” to detect that the winter sea ice is going to open, by combining passive and active satellite remote sensing. We are using the Weddell Polynya in the Southern Ocean as test subject.

Team:

  • Myself until February 2021;
  • Postdoc Lu Zhou, April 2021 – April 2023;
  • Postdoc Adriano Lemos, January-August 2020;
  • PhD student Martin Mohrmann, co-supervised with S. Swart and H. Ploug from the Marine Sciences Department, Gothenburg University, January 2018 – June 2022.

Publications:

  • Representation of Antarctic Polynyas in CMIP6 models: [25] (accepted)
  • Detection of upcoming sea ice opening several days ahead from passive remote sensing: [23] and [13];
  • On how exceptional the Weddell Polynya re-opening of 2016-2017 has been: [14][17].

Research theme: Arctic deep waters

Project “Why is the deep Arctic Ocean Warming? (WAOW)”, funded by Vetenskapsrådet grant 2018-03859
Ca 4 million SEK; I am the PI (single applicant here)
Started in January 2019, finishes in December 2022

The aim of this project is to finally determine the path and variability of the deep waters of the Arctic Ocean, from 2000 m to the sea floor, using notably data that we collected during the international MOSAiC expedition (2020) and will collect as part of the Synoptic Arctic Survey (2021).

Team:

  • Myself;
  • PhD student Salar Karam, of which I am the main supervisor, October 2019 to September 2023;
  • Maren Walters from Uni Bremen, for expertise with CFC analysis;
  • Collaboration with the wider MOSAiC Team Ocean and the Swedish Synoptic Arctic Survey consortium.

Publications:

Nothing yet; we are currently analysing the MOSAiC data.

Heuzé et al. (2020) Global decline of deep water formation with increasing atmospheric CO2

C. Heuzé, M. Mohrmann, E. Andersson and E. Crafoord (2020). Global decline of deep water formation with increasing atmospheric CO2. EarthArXiv. doi:10.31223/X56K6D.

We analysed the 1% CO2 idealised run of 30 CMIP6 models and found:

  • Globally, open ocean deep convection ceases around 600 ppm. We instead enter a new regime, with large areas of mixed layers not exceeding 500 m;
From Fig 1: Hotspots of deep water formation in the North Atlantic and their decline with rising CO2 concentration (inserts). Black line is the multimodel mean (30 models); dark shading = 75% of the models; light shading = 90%.
  • Unlike what was found in studies based on individual models, deep convection does not start in the Arctic Ocean;
  • Consequently, the flows of North Atlantic Deep Water and Antarctic Bottom Water (AMOC and Southern MOC) fall at half their pre-industrial values;
From supp. Fig. 1: The Atlantic Meridional Overturning Circulation (AMOC) and Southern MOC at 30S are halved by the end of the run. Black line is the multimodel mean (30 models); dark shading = 75% of the models; light shading = 90%.
  • The main reason is a global large increase in stratification, which is caused by rising upper ocean temperatures and/or surface freshening. Trends in wind are insignificant; the wind cannot break the stratification.

We submitted this to Nature Climate Change. Twice. The first time in October 2019, the manuscript featured “only” 13 models. We were asked to resubmit once we had more models. The second version, with 30 models and more robust methods, was rejected in October 2020 for not being dramatic enough (my interpretation).
None of us is paid to work on this project. We all need to concentrate on more exciting things (e.g. the MOSAiC data), and, simply, it stopped being fun a long time ago. So sure, we could have played the peer-review game with another journal and forced ourselves through another year of selling this paper. Instead, following advice received from the community, we put the manuscript on EarthArXiv, in hope that it can be useful in its present state.

Click here for the full-text

Aldenhoff et al. (2020) First-year and Multiyear Sea Ice Incidence Angle Normalization of Dual-polarized Sentinel-1 SAR Images in the Beaufort Sea

W. Aldenhoff, L.E.B. Eriksson, Y. Ye and C. Heuzé (2020), First-year and Multiyear Sea Ice Incidence Angle Normalization of Dual-polarized Sentinel-1 SAR Images in the Beaufort Sea, IEEE Journal of Selected Topics in Applied Earth Observations and Remote Sensing. vol 13, pp 1540-1550, doi: 10.1109/JSTARS.2020.2977506

Obs: W. Aldenhoff was then a PhD student under my supervision (graduated 2020).

Last paper of Wiebke’s PhD to be published (see also [12], [18] and [19]), although this work was among the first executed. The reason? It is common practice to normalise the backscatter by the incidence angle, but we wondered:

  • does this normalisation really improve on the results?
  • and what is the best way to perform the normalisation?

Using some rare near-coincidental ascending and descending images from Sentinel-1, we could retrieve the slopes backscatter / incidence angle for sea ice, for HH and HV. And the short answer is: yes, normalisation does make a big difference for automatic ice classification and for mosaicing. This is especially true for multi year ice, whose backscatter depends on its age.

One issue remains: first year ice backscatter still is too close to the noise level, to the point that the first subswath cannot be used.

Fig 10 of Aldenhoff et al. (2020) showing the effect of the normalisation when mosaicing.

Download the full-text here.

Wåhlin et al. (2020) Ice front blocking of ocean heat transport to an Antarctic ice shelf

A. Wåhlin, N. Steiger, E. Darelius, K.M. Assmann, M.S. Glessmer, H.K. Ha, L. Herraiz-Borreguero, C. Heuzé, A. Jenkins, T.W. Kim, A.K. Mazur, J. Sommeria, and S. Viboud (2020), Ice front blocking of ocean heat transport to an Antarctic ice shelf. Nature, 578, 568–571, doi:10.1038/s41586-020-2014-5.

In autumn 2017, we spent a month at the Coriolis rotating platform in Grenoble (France) and had way too much fun in their 13-m diameter tank playing with lasers and cameras and building a plexiglass ice shelf. Back to the office, a comparison of our experiment results with measurements from moorings by the Getz ice shelf revealed that:

  • the steep change in water depth at the ice shelf front blocks the barotropic component of the flow;
  • the baroclinic component, which contains very little heat, is the only one that can intrude under the ice shelf;
  • the thinner the ice shelf, the more the heat-rich barotropic component can intrude and further melt the ice shelf from below.

That is, we discovered yet another worrying self-reinforcing climate change feedback: as the ice shelf melts in response to warming ocean and atmosphere, the change in water depth becomes less steep, and the warmer barotropic flow can start contributing to the melting and accelerate it. For more information, check the full text.

Download the full-text here.