In October I headed to Denmark for a ten-day workshop on the results from the latest generation of complex climate models. It was organised by Ruth Mottram and Celine Heuzé, and was specifically focussed on the Arctic. Around twenty-five of us attended, from all disciplines of oceanography, atmospheric science, glaciology, modelling, observations and remote sensing.
The beautiful Søminestationen (sea mine station in Danish), located by the side of a fjord outside of Copenhagen.
Our work on the CMIP6 data (see below) was interspersed with lectures on the Arctic climate system. These included the physics of the atmospheric boundary layer, deep circulation of the Arctic Ocean, and the mass balance of terrestrial ice sheets. We also had talks on more methods-based topics, such as regional climate models, the work of climate services agencies, and the organisation of the CMIP programme.
On top of these longer form talks, all participants described their work and research interests in 10-15 minute presentations. These were super interesting, spanning stratospheric dynamics to paleo-circulation of the ocean.
What is CMIP6?
The Intergovernmental Panel on Climate Change (the IPCC) is tasked by governments around the world with telling them what will happen to the world if we follow different carbon emission pathways. To do this, the IPCC calls upon the results of a suite of complex models of Earth’s atmosphere, ocean and ice sheets.
To systematically and fairly examine the outputs of models, it’s important to know how they relate to each other. Individual models are designed differently, so give different “answers” to what will happen if we put a certain amount of carbon into the atmosphere. They have variously complex sub-models for the oceans, for sea ice, and for atmospheric dynamics, which results in various strengths of warming. Various reductions in sea ice. Various shutdowns to the Atlantic circulation. So the IPCC consults models, and presents a spread of answers for each question. Data from the Coupled Model Intercomparison Project (CMIP) contains these answers - we are now on round six of this project.
Schematic of the Coupled Model Intercomparison Project. The models are intercompared with a variety of metrics and experiments, and their outputs are used to understand variability, predictability, and future climate.
A Science Question
The Arctic is warming at around three to four times the global average rate. As well as lifting the global average rate due to its outsized warming, this amplification also stresses physical and biological systems in the region. But the complex models used by the IPCC do not agree on the future magnitude of this amplification. What might cause this spread?
Recent Arctic amplification in he ERA5 atmospheric reanalysis data set. In these data the Arctic is warming at 3.6x the global average rate.
One reason that models simulate different magnitudes of surface Arctic warming is because they differentially trap energy near the surface rather than radiating it out to space or further up in the atmosphere. In winter the snow surface radiatively cools itself to below the temperature of the air above, creating a temperature inversion. In this scenario, air temperature increases with height in the atmosphere. Inversions create “stable” atmospheric conditions which reduces the upward movement of sensible heat, trapping it at the surface. So… do models with stronger atmospheric inversions have larger Arctic amplification?
Winter variability of air temperature with height under clear (red) and cloudy (blue) skies at the SHEBA campaign. Figure from Pithan et al., 2014. Temperature initially increases with height - this is called a temperature inversion.
As well as answering the above question regarding Arctic amplification, we might also ask the following: how strong should a model’s temperature version be under a given set of meteorological conditions? To address this, we corralled weather balloon data from three field campaigns: SHEBA, MOSAiC and the Soviet NP stations.We then characterised the strengths of atmospheric inversions that were observed, and compared theses strengths to other, coincident meteorological variables. These observed relationships between inversion strength and surface air temperature, wind speed and cloud cover can then be compared with the CMIP6 models.
Results?
Actually doing the analysis was difficult, as it required the analysis of 6-hourly data from the models, which is both computationally and memory intensive. Furthermore, not all model runs that contribute to CMIP6 output the 6-hourly data that we wanted to analyse. I’m going to keep our results under my hat for now, because we might press on and write them up fully. But suffice to say that our results were interesting, and hopefully you’ll see them in your favourite journal one day.
I’d like to thank Ruth Mottram and Celine Heuzé for organising the workshop, and Tina Odaka for organising all the compute-power and teaching us how to use it! My thanks also go to Felix Pithan and Jonny Day, who acted as mentors for our group as we worked, guiding us towards the most interesting and realistically solvable questions.
Group photo on a (presumably decommissioned) torpedo outside the research station, which takes its name from the history of naval weapons development.