Impact of sub-seasonal atmosphere-ocean interactions on extreme event statistics

Modified frequencies and magnitudes of extreme events due to climate change can have large impacts on societies and are therefore a key area of current research. Large model ensembles are required to quantify and attribute changes to extreme events. Until now, the large ensembles used for such studies are commonly atmosphere-only models forced with time-varying sea surface temperatures (SST) and sea ice. 

This approach is very powerful but presents problems of internal physical consistency. In an SST-forced model, the ocean acts as having an infinite heat capacity whereas in the real-world SST emerges dynamically from the interaction of atmospheric and oceanic processes (Dong et al., 2020). This is particularly relevant for the North Atlantic where ocean processes, in particular meridional heat transport, are key drivers of air-sea coupling. A long-standing challenge, however, is the computational cost of spinning up fully coupled atmosphere-ocean models that hinders their application to large-ensemble, high-resolution simulations required to quantify changing hazard frequencies of low-probability events.  

In this work we combine the HadAM4 (Webb et al., 2001) atmospheric model at N144 resolution with a Slab ocean (Hewitt & Mitchell, 1997; Williams et al., 2003), which includes a simple sea ice model, to yield the atmosphere-Slab Ocean model HadSM4. The Slab Ocean is forced with diagnosed heat convergence (Q-Flux) and surface currents for sea ice advection (a useful model-development finding for this kind of experiment is that including sea ice velocity information from reanalyses in the surface current field yields a substantially improved spatial pattern of sea ice). We are therefore able to directly compare SST-forced atmosphere-only runs with Q-Flux-forced runs where SST is an emergent property of the model, specifically accounting for the passive response of SSTs in the North Atlantic. Using the distributed infrastructure of climateprediction.net (Guillod et al., 2017; Massey et al., 2015) we run large ensembles to compare extreme statistics and quantify the importance of fast ocean-atmosphere coupling for extreme event statistics.

We further use this large ensemble setup to investigate the dynamics that drive extreme events from the ocean through air-sea interaction to atmospheric processes. We address is whether and how the slope of a return-time plot (related to the scale parameter of a GEV distribution) is affected by atmosphere-ocean interactions, since this statistic plays a central role in determining relative-risk estimates in event attribution studies. We then investigate how a perturbation to the Q-Flux, representing a change in ocean heat transport, propagates through the system and alters the statistics of extreme events.

Dong et al., 2020, Climate Dynamics, 55(5–6), 1225–1245. 

Guillod et al., 2017, Geoscientific Model Development, 10(5), 1849–1872. 

Hewitt & Mitchell, 1997, Climate Dynamics, 13(11), 821–834.

Massey et al., 2015, Quarterly Journal of the Royal Meteorological Society, 141(690), 1528–1545. 

Webb et al., 2001, Climate Dynamics, 17(12), 905–922. 

Williams et al., 2003, Climate Dynamics, 20(7–8), 705–721.