Context

Over the last decades, the Arctic sea ice has experienced a drastic decline which is expected to continue in the near-term future. Arctic changes are thought to impact the mid-latitude atmospheric circulation to an extent and through mechanisms which are still highly debated. Heat exchange between sea ice and atmosphere plays a crucial role on the rate of Arctic sea ice melting (Rothrock et al, 1999 ; Screen and Simmonds, 2010) as well as on the teleconnections between polar and non-polar regions (Bader et al 2011 ; Vihma, 2014 ; Overland et al 2015). Lower troposphere heating through turbulent heat fluxes (latent and sensible) dictate the thermal structure of the lower atmosphere and plays a key role in the development of atmospheric circulation anomalies (Overland, 2015).

The representation of surface turbulent fluxes of sensible and latent heat at the ice-atmosphere interface is relatively crude in most climate models. They are typically estimated using bulk formulas (e.g. Roy et al, 2015), based on Monin-Obukhov similarity theory (e.g. Andreas 2002 ; Andreas et al, 2010a,b). Most of the existing formulations, however, have been developed exploiting observational campaigns which took place in the tropics or the mid-latitudes (e.g Belamari and Pirani 2007) and are not necessarily suitable for the particular conditions found in polar regions (e.g. high stability of the atmospheric profile, surface roughness). Formulations developed specifically for polar conditions rely on sparse observational data and still lack an accurate formulation of the stability functions appropriately accounting for high stability and surface roughness (e.g. Andreas 2002 ; Andreas et al, 2010a,b ; Lupkes et al, 2012, 2013). The transfer coefficients are therefore often instead assumed to be constant over sea ice (Parkinson and Washington 1979 ; Maykut, 1982) in state-of-the-art climate models. The lack of available observations is the main reason for the lack of more advanced or suitable parameterisations, together with the insufficiently detailed simulated surface roughness.

The Year of Polar Prediction (YOPP) was an extended period (2017-2020) of coordinated intensive observational and modeling activities aiming at improving polar prediction capabilities. YOPP was a WCRP/WWRP (World Climate Research Programme / World Weather Research Programme) initiative as part of the Polar Prediction Project and provided an opportunity to gather a wide variety of new observations in the polar regions. The ASET project offers to improve the realism of modelled Arctic climate changes and linkages between polar and mid-latitude regions through the development of novel formulations of turbulent heat exchanges between the atmosphere and sea ice, exploiting the wealth of YOPP new observations.

Objectives

1. The first objective is to build a database documenting as extensively as possible the atmosphere-ice interface and heat exchanges from the YOPP-endorsed and previous campaigns.

2. The second is to develop new parameterisations for the latent and sensible heat fluxes at the ice surface exploiting the YOPP data.

3. The third is to improve our understanding and refine estimates of the Arctic climate change and its impact on the midlatitudes, exploiting model improvements from ASET.

Team

Virginie Guemas
Donald Cummins
Théo Brivoal
Laurent Bessières
Sébastien Blein
Youcef Amar

Results

1 - WP1 Observational database

An observational database has been built gathering the data from the following campaigns :
– MOSAiC (Multidisciplinary Drifting Observatory for the Study of Arctic Climate) : The Polarstern icebreaker was trapped in the ice north of Siberia and drifting toward Fram Strait during a 1-year long campaign (2019-2020). Three meteorological stations have been deployed around the Polarstern within a distance of 10 to 30km. All provided measures of wind, temperature and humidity together with estimates of surface turbulent fluxes.
– SHEBA (Surface Heat Budget of the Arctic Ocean) : The SHEBA ice camp drifted approximately 2700 km in the Beaufort Gyre between 2 October 1997 and 11 October 1998. Four remote sites that ranged in distance from 0.25 to 30 km from the main camp were also instrumented. All provided measures of meteorological variables as well as surface turbulent fluxes.
– ASCOS (Arctic Summer Cloud Ocean Study) : The Oden icebreaker was deployed from 1st August 2008 to 9 September 2008, drifting between the North Pole and Fram Strait. Two different masts over an ice floe were estimating turbulent heat fluxes together with wind, temperature and humidity.
– ACCACIA (Aerosol-Cloud Coupling and Climate Interactions in the Arctic) : Two aircrafts conducted eight flights from 21 to 31 March 2013 over Fram Strait and in the Barents Sea, measuring turbulent fluxes and meteorological parameters in the atmospheric boundary layer, as well as surface meteorological parameters through radar.
– ACSE (Arctic Cloud in Summer Experiment) : The Oden icebreaker left Tromsö, Norway, on 5 July 2014 and followed the Siberian Shelf, until reaching Barrow, Alaska, on 18 August. A second leg left Barrow on 21 August following a similar route back, albeit farther north and returned on 5 October in Tromsö. A mast at the bow of the ship was used to obtain turbulent surface fluxes while a weather station on the seventh deck at about 25 m measured temperature, humidity and wind.
– Arctic Ocean 2016 : The Oden icebreaker departed from Longyearbyen, Svalbard on 8 August, and operated in the Arctic Ocean, mainly in the Amundsen Basin and in areas around the underwater mountain ranges, Lomonosov Ridge and Alpha Ridge until 19 September 2016. Similar instrumentation as in the ACSE expedition was employed.
A library to load those data into python as autodocumented variables and datasets is publicly available under git@github.com:virginieguemas/TurbObsLibrary.git.

2 - WP2 Representation of air-ice latent and sensible heat fluxes

In models, surface turbulent fluxes are calculated using bulk formulas, which are functions of meteorological parameters and so-called bulk exchange coefficients. From campaign data, those bulk exchange coefficients are estimated for each co-localized measurement of meteorological parameters and turbulent fluxes. Developing a bulk parameterization of surface turbulent fluxes essentially consists in proposing a formulation of these exchange coefficients as a function of the meteorological conditions.

The ASET observational database allowed us to dissect the uncertainties on the exchange coefficients estimated from field campaigns. Such a quantification of the uncertainty on the exchange coefficients had never been addressed before in the literature. We published an article (Blein et al 2024) documenting the contributions of the uncertainties arising from the sensors random errors and from the post-processing necessary to calculate the exchange coefficients (choice of stability functions, random errors in the stability function estimates, various approximations typically used in the literature). An optimal strategy for minimizing these uncertainties and estimating exchange coefficients from campaigns is proposed. Recommendations are provided regarding criteria to select data to be used when developing a bulk parameterization and how to use their uncertainties. The next article will focus on the recalibration of the parameters involved in the parameterization of Lupkes and Gryanik (2015) of exchange coefficients above a mixed ice-water surface. This recalibration will exploit the updated evaluations of the exchange coefficients from the ASET observational database (Blein et al, 2024) and will introduce weights on the data according to their uncertainty.

Although the Monin-Obukhov Similarity Theory which lays the ground for the bulk parameterizations is well established, its application through bulk algorithms combines several parameterizations and functions, each coming with its own estimation uncertainties in a time-consuming iterative algorithm. To bypass those challenges, machine-learning-based algorithms which relate the turbulent fluxes directly to the observable meteorological quantities they physically depend on have been evaluated on the ASET database. A neural-network based algorithm has been proposed to estimate surface momentum, sensible heat and latent heat fluxes which perform respectively marginally better than (for momentum), substantially better than (for sensible heat) and comparatively bad with (for latent heat) the more traditional bulk-based parameterizations on the pre-MOSAiC data. An article in Boundary-Layer Meteorology (Cummins et al, 2024) describes the algorithm and its performance. This algorithm was then evaluated on the MOSAiC data for which the error is reduced by up to 70% compared to the bulk-based approach as described in an article we published in Geophysical Research Letters (Cummins et al, 2023). A fortran-based code has been released publicly for inclusion and testing in coupled climate models.

All the tools which allowed to carry out these analyses are publicly available here :
– git@github.com:virginieguemas/CDlib.git
– git@github.com:virginieguemas/meteolib.git
– git@github.com:sebastienblein/ParamObsCDN.git
– git@github.com:donaldcummins/PolarFlux.git

3 - WP3 Impact on climate sensitivity

The CNRM-CM6 coupled climate model which participated in the CMIP6 exercise uses bulk formulas for calculating surface turbulent fluxes that do not account for polar specificities. The following modifications have been implemented :

– a large set of options of stability correction functions for surface turbulent fluxes including the most recently proposed targeted to stable polar conditions, i.e. the SHEBA functions (Paulson, 1970 ; Lettau, 1979 ; Holtslag and de Bruin, 1988 ; Beljaars and Holtslag, 1991 ; Dyer, 1974 ; Fairall et al, 1996 ; Grachev et al, 2000 ; Grachev et al, 2007).
– ice-specific formulations of aerodynamic roughness (Andreas et al, 2004, 2005, 2010) and scalar roughness (Andreas, 1987).
– the parameterization of Lupkes and Gryanik (2015) for more realistic surface turbulent fluxes above mixed water-ice surfaces, as well as an option to test recalibrations of this parameterization proposed in the more recent literature (i.e. Elvidge et al 2016).

The following modifications are being implemented :
– the neural-network based parameterization of surface turbulent fluxes in polar regions developed by Donald Cummins (see WP2) and described in Cummins et al (2024).

Sensitivity experiments to these modifications will be carried out by Laurent Bessières in order to guide strategic choices for the future version of the CNRM-CM coupled model.

4 - Complementary activities :

The ASET team contributes to the H2020 CRiceS project which started in September 2021. Through comprehensive analysis of new and emerging in-situ and satellite observations, CRiceS will advance model descriptions of sea ice dynamics and energy exchange. The objectives of the CRiceS project fully encompass those of ASET and the ASET activities have been proposed as a contribution to CRiceS. Complementary activities toward an improved representation of the snow cover on top of ice have also been proposed for CRiceS. The snow cover above ice affects heat exchanges not only through its impact on solar absorption but also through an impact on the surface roughness and therefore turbulent heat fluxes. Theo Brivoal has implemented and tested the advanced ISBA-ES snow scheme in the thermodynamic part of the ocean-sea ice component of CNRM-CM6. ISBA-ES includes a representation of snow density changes, grain size and a multi-band albedo scheme and accounts for processes such as blowing‐snow sublimation, snow melt and metamorphism. An article evaluating its performance over sea ice in ocean-sea ice and thermodynamic-only mode is under preparation (Brivoal et al, 2024). The adaptation of the dynamic will be pursued during the coming year.

Publications

[3] Blein S, Guemas V, Brooks IM, Elvidge AD, Renfrew IA, 2024, Uncertainties of Drag Coefficient Estimates Above Sea Ice from Field Data. Boundary-Layer Meteorol 190, 11. doi : 10.1007/s10546-023-00851-9

[2] Cummins DP, Guemas V, Blein S, Brooks IM, Renfrew IA, Elvidge AD, Prytherch J, 2024, Reducing Parametrization Errors for Polar Surface Turbulent Fluxes Using Machine Learning. Boundary-Layer Meteorol 190, 13. doi:10.1007/s10546-023-00852-8.

[1] Cummins DP, Guemas V, Cox CJ, Gallaher MR, Shupe MD, 2023, Surface turbulent fluxes from the MOSAiC campaign predicted by machine learning, Geophysical Research Letters, 50, e2023GL105698. doi:10.1029/2023GL105698.