Systematic errors in model representation of stratosphere- troposphere coupling Ted Shepherd Grantham Chair in Climate Science Department of Meteorology University of Reading

In both hemispheres, stratospheric polar vortex variability is connected to the troposphere: in NH, the effect is strongest over the North Atlantic Southern and Northern Hemisphere “annular modes” (SAM and NAM), based on hemispheric EOFs Thompson & Wallace (2000 J. Clim.) Zonal wind Surface geopotential

There is an apparent downward propagation of annular mode anomalies: stratosphere-troposphere coupling – A warmer polar stratosphere (weaker vortex) leads to an equatorward shift in the midlatitude tropospheric jet Composites of Northern Annular Mode (NAM) indices Baldwin & Dunkerton (2001 Science)

Stratosphere-resolving climate models generally predict less of a poleward shift in the wintertime North Atlantic storm track – Attributed to weakening of Arctic stratospheric polar vortex – Figure shows percentage change in frequency of extreme wintertime rainfall from 4xCO 2 : right is effect of stratosphere Scaife et al. (2012 Clim. Dyn.)

Stratosphere-troposphere coupling has a strong seasonal dependence, which is quite different in the two hemispheres – Figure shows interannual std dev of monthly mean polar T Kuroda & Kodera (2001 JGR) NH SH

The dominant form of stratospheric polar vortex variability in the NH occurs through Stratospheric Sudden Warmings (SSWs) – About half of all SSWs are short-lived, as in , while half have extended recovery periods, as in – Figures show Aura-MLS polar-cap average temperatures Hitchcock, Shepherd & Manney (2013 J. Clim.)

The extended recoveries from SSWs (right) are associated with a strong suppression of planetary-wave fluxes (colour) from the troposphere (contours show zonal winds). Also seen in models. Vertical EP flux anomalies Hitchcock et al. (2013 J. Clim.); see also Hitchcock et al. (2013 JAS) So strat-trop response is complex; mechanism is not well understood

Models can produce quite realistic simulations of Arctic polar vortex variability Simulations suggest considerable multi-decadal variability, even for three-member ensembles Hitchcock et al. (2013 J. Clim.)

In general, stratosphere-resolving climate models simulate SSWs fairly well However the models need to be tuned carefully to achieve this (gravity-wave drag) Butchart et al. (2011 JGR) McLandress & Shepherd (2009 J. Clim.) CMAM (CCMVal-1) CMAM (CCMVal-2)

Stratosphere-resolving models can correctly predict the surface response to SSWs when initialized at the time of the SSW – Figure shows response averaged over days after the SSW, for 20 SSWs from (model: ensemble of 10) Sigmond, Scinocca, Kharin & Shepherd (2013 Nature Geosci.)

Morgenstern et al. (2010 JGR) 850 hPa NAM index Stratosphere-resolving climate models do not provide a robust prediction of how the surface NAM will respond to climate change How much of this spread is due to orographic GWD?

In CMAM, the Arctic wintertime mean sea level response to doubled CO 2 changed dramatically between two different (but plausible) parameter settings in the orographic GWD scheme Difference consistent with Scaife et al. (2012): weakened stratospheric vortex / weaker poleward shift in tropospheric jet Sigmond & Scinocca (2010 J. Clim.) (DRAG)

The difference was not due to the different orographic GWD response to doubled CO 2 – The orographic GWD response (colours) is a vertical dipole, reflecting momentum conservation (Shepherd & Shaw 2004 JAS), so has a negligible effect on surface pressure DJF zonal wind and OGWD response to doubled CO 2 in T63 dynamical CMAM Sigmond & Scinocca (J. Clim., in press) Sigmond & Scinocca (2010 J. Clim.) Contours show zonal wind response

Rather, whether the CMAM Arctic vortex strengthened or weakened under doubled CO 2 depended on the mean state – So the sensitivity to orographic GWD is via its effect on the climatological winds, which affect the planetary-wave response (shown below) to doubled CO 2 DJF zonal wind and OGWD response to doubled CO 2 in T63 dynamical CMAM Sigmond & Scinocca (J. Clim., in press) Sigmond & Scinocca (2010 J. Clim.)

In the Southern Hemisphere, the ozone hole has been the primary driver of recent summertime trends in the Southern Annular Mode (Arblaster & Meehl 2006 J. Clim.) A stronger late- spring stratospheric vortex leads to a poleward shift in the summertime tropospheric jet Linear trends up to 2000 Gillett & Thompson (2003 Science)

Similarly, ozone recovery needs to be accounted for in projections of summertime climate change (cf. Son et al. 2009) CMAM McLandress, Shepherd, et al. (2011 J. Clim.) SAM trends also have implications for Southern Ocean heat and carbon uptake, and potentially for ice- sheet stability Effect of ozone loss Effect of ozone recovery

However, climate models tend to have a systematic bias towards a too-late Antarctic vortex breakup – To what extent does this compromise projections of summertime SH high-latitude climate? Butchart et al. (2011 JGR)

Consistent with this mean bias, the models also tend to have too persistent SAM variability, especially in the summer (left) Leads to much too much apparent “predictability” of the surface (850 hPa) SAM in the models (right) – Shows fraction of day variance predicted by persistence Gerber et al. (2010 JGR)

The SH jet has a maximum around 60°S – At this latitude band, the surface is represented entirely as ocean in the models, hence no orographic GWD! McLandress, Shepherd, Polavarapu & Beagley (2012 JAS) OGWD in CMAM

When CMAM is run in data assimilation mode, increments imply missing drag around these latitudes, which descends from the upper stratosphere as the zero wind line descends (left) There is other evidence for the role of oro GWD at these latitudes An ad hoc inclusion of extra oro GWD in this latitude belt substantially reduces the zonal-wind bias in CMAM (right) Zonal wind increments from data assimilation McLandress, Shepherd, Polavarapu & Beagley (2012 JAS)

Stratospheric variability acts to prolong tropospheric SAM variability in late-spring/summer – Yet even with this influence removed, CMAM SAM timescales are much too long in the summer – Could have implications for SAM response to forcing Simpson, Hitchcock, Shepherd & Scinocca (2011 GRL) CMAM CMAM with stratospheric AM variability suppressed Obs

Models tend to locate the tropospheric eddy-driven jet too far equatorward, in both hemispheres (black are obs) – Reflected here in the location of the node of annular-mode variability – Biases are similar when observed SSTs are imposed, implying the errors arise from atmospheric processes Gerber et al. (2010 JGR)

Bias-correcting the climatological tropospheric jet in CMAM does not reduce the bias in SAM timescale Simpson, Hitchcock, Shepherd & Scinocca (J. Clim., in press) Contradicts Kidston & Gerber (2010 J. Clim.) claim that the SAM timescale bias results from jet latitude bias Lesson: cannot rely on correlations; need to break feedback loop between eddies and mean flow to identify biases CMAM Obs

In CMAM, the summertime SAM timescale bias arises from lack of damping of the SAM by planetary wave k=3 – Positive forcing of SAM by synoptic-scale eddies is OK This bias is evident only in the summer season The same bias is evident in all the CMIP5 models Simpson, Shepherd, Hitchcock & Scinocca (J. Clim., in press)

Strat-trop coupling: stronger/weaker strat polar vortex leads to poleward/equatorward shift in trop jet, on all timescales In the NH, stratosphere-resolving climate models are able to realistically represent vortex variability and strat-trop coupling – Models need to be carefully tuned, usually with oro GWD – Spread in modelled response to climate change may reflect spread in model climatologies; e.g. sensitivity to oro GWD In the SH, models exhibit a systematic bias towards a too-late vortex breakdown, which leads to much too persistent tropospheric SAM variability in SH summer – Bias may result from missing oro GWD around 60°S – Models are also deficient in damping of SAM by k=3 – Common equatorward bias in jet location seems to be a symptom rather than a cause of the SAM timescale bias Summary

Yoden, Taguchi & Naito (2002 JMSJ) Polar temperatures at 30 hPa (approx 25 km) Interannual variability in the NH may not be well characterized by the historical record (which is too short)

There has been considerable interest in annular-mode timescales, motivated by the fluctuation-dissipation theorem In the NH, the long-timescale variability in models seems to occur too late in the season, also the “predictability” of 850 hPa NAM – i.e. fraction of day surface variance predicted by persistence Gerber et al. (2010 JGR)

However ensembles of simulations suggest that the seasonality of the stratospheric NAM timescale is not well characterized by the half-century observational record Hitchcock, Shepherd & Manney (2013 J. Clim.)

Over Europe, the surface effects of stratospheric variability are at least as important as those from ENSO (e.g. Jan-Feb 2013) Provides a mechanism for effect of stratospheric forcing on high-latitude surface climate (e.g. QBO, solar variability) – Figure shows wintertime surface air temperature differences (in K) between circulation regimes Thompson, Baldwin & Wallace (2002 J. Clim.)