Understanding climate model biases in Southern Hemisphere midlatitude

Understanding climate model biases in Southern Hemisphere mid-latitude variability Isla Simpson 1 Ted Shepherd 2, Peter Hitchcock 3, John Scinocca 4 (1) LDEO, Columbia University, USA (2) Dept of Meteorology, University of Reading, UK (3) DAMTP, University of Cambridge, UK (4) CCCma, Environment Canada, Canada

The Southern Annular Mode Dominant mode of variability in SH extra-tropical circulation Climatology First EOF ERA-Interim re-analysis

The SAM timescale Calculate the autocorrelation function =7 days Calculate the e-folding timescale.

The SAM timescale bias – CMIP 3 § Climate models exhibit much too persistent SAM anomalies in the summer season. Obs IPCC models Gerber et al (2008)

The SAM timescale bias – CCMVal 2 § Climate models exhibit much too persistent SAM anomalies in the summer season. Obs CCMVal models Gerber et al (2010)

The SAM timescale bias – CCMVal 2 § Climate Models exhibit a SAM that is much too persistent in the summer season. Gerber et al (2010)

The SAM timescale bias – CMIP 5 § Climate Models exhibit a SAM that is much too persistent in the summer season.

Many climate forcings produce a mid-latitude circulation response that projects onto the SAM. Ozone depletion Son et al (2010), JGR

Why is this potentially of concern for simulating forced responses? May indicate that we’re getting an important process wrong in the simulation of the SH extra-tropical circulation.

Why is this potentially of concern for simulating forced responses? § Eddy Feedbacks (Lorenz and Hartmann 2001, 2003), Robinson 2000) § Intraseasonal Forcing e. g. forcing from the stratosphere (Keeley et al 2009) § Dissipative processes e. g. surface friction

Can we isolate the role for “internal” tropospheric dynamics on the SAM timescale bias from the influence of stratospheric variability as an intraseasonal forcing on the SAM?

stratospheric influence onlate SAMintimescales? The. ASH vortex breaks down too GCMs, maybe this is resulting in enhanced stratospheric variability in the summer and contributing to the SAM timescale bias? Thought to be stratospheric variability that gives rise to this maximum…variability in the timing of the vortex breakdown (Baldwin et al 2003)

The Canadian Middle Atmosphere Model § Comprehensive stratosphere resolving GCM § T 63 L 71, lid=0. 0006 h. Pa § Without interactive chemistry § Prescribed SSTs § No QBO § Constant GHG’s (1990’s concentrations)

Model Experiments § 100 year free running control simulation (FREE) § 100 year nudged simulation (NUDGED) In NUDGED, the zonal mean vorticity, divergence and temperature in the stratosphere are nudged toward the zonal mean, seasonally varying climatology of FREE. We eliminate zonal mean stratospheric variability but keep the climatology the same.

The Nudging Process In spectral space Only acting on the zonal mean K

The Nudging Process In spectral space Only acting on the zonal mean climatology time

FREE and NUDGED have the same climatologies, but FREE has stratospheric variability, NUDGED does not. Vortex Breakdown Dates FREE NUDGED

ERA-Interim FREE

FREE NUDGED

Contribution from stratospheric variability Stratospheric variability enhances the SAM timescales in the SH spring.

ERA-Interim NUDGED

There does seem to be a problem in the “internal” dynamics of the tropospheric circulation. Is this caused by climatological circulation biases?

Relationship between climatological jet bias and SAM timescales If we improve the jet position, do we improve the timescale of SAM variability? Kidston and Gerber (2010)

Bias Correcting Experiments Obtain the mean tendency that is required to bring the model toward the ERA climatology (Kharin and Scinocca, 2012, GRL) applying that constant seasonally varying tendency to the model. Model Climatology Observed Climatology Time Different from nudging in that variability can still occur, just around a new climatological state.

Two different experiments § Bias correcting at all levels – BC • Both stratospheric and tropospheric variability but around an improved climatological state. Improved timing of the vortex breakdown and improved tropospheric jet structure. § Bias correcting in the troposphere and nudging the zonal mean toward ERA-Interim in the stratosphere BCNUDG • Removed stratospheric variability but has an improved climatological timing of the vortex breakdown. Improved the tropospheric jet structure.

Improved tropospheric jet structure?

Annual Mean Timescales Kidston and Gerber (2010)

Annual Mean Timescales

The SAM timescale bias in CMAM does not seem to be caused by climatological circulation biases. Eddy feedback biases?

Eddy feedbacks on the SAM Eddy forcing of the SAM regressed onto the SAM Index Eddies driving the SAM driving eddies i. e. , a positive feedback See Lorenz and Hartmann (2001), Simpson et al (2013)

Quantify the feedback strengths for each simulation and the reanalysis. Focus on the DJF season.

Synoptic scale eddy feedback (k>3)

Planetary scale eddy feedback (k=1 -3)

Summary of DJF feedback strengths This is mostly coming from wavenumber 3

DJF regressions averaged over lags +7 to + 14 days FREE ERA u -u’v’, k=3

Regressions on the 300 h. Pa (+7 to +14 lag average) ERA-Interim -u’v’ (k=1 -3)

Regressions on the 300 h. Pa (+7 to +14 lag average) ERA-Interim FREE -u’v’ (k=1 -3)

Comparison with CMIP-5 historical simulations § 20 models: those with 6 hourly u and v available § Quantify DJF feedback strength

Eddy feedback, All k

Eddy feedback, k=1 -3

Virtually all GCMs exhibit this same bias in planetary wave feedbacks. Models don’t capture the negative feedback by planetary scale waves that is localised to the south west of New Zealand in the summer season.

Relation to climatological circulation biases? Our bias corrected runs tell us that climatological circulation biases are NOT the CAUSE of the eddy feedback bias. But the climatological circulation biases and eddy feedback biases could be related e. g. they could have a common cause.

Climatologically there is wave activity propagating into the mid-latitudes to the S-W of New Zealand ERA FREE

There are common climatological biases in the region around New Zealand 300 h. Pa eddy geopotential height ERA FREE-ERA

There are common climatological biases in the region around New Zealand 300 h. Pa eddy geopotential height FREE-ERA CMIP 5 - ERA

There are common climatological biases in the region around New Zealand 300 h. Pa eddy geopotential height CMIP 5 - ERA CMIP 5 CONSENSUS

Climatological jet latitude bias CMIP-5 ERA

Conclusions § Overly persistent SAM variability in the SH summer season is a common model bias. § The CMAM experiments demonstrated a bias in internal tropospheric dynamics that is not alleviated by improving the climatological circulation. § The problem is associated with a bias in the feedback by planetary scale waves in the model in the summer season. § This is true of the majority of other models in the CMIP-5 archive.

Conclusions § In order to have faith in the future predictions for the SH mid-latitude circulation in the summer season, we need to understand the planetary wave feedback localised to the SW of New Zealand why it is biased in the models.

But…. § Models do reasonably well at simulating past SAM trends. CMIP-5 DJF SAM Trends § Are we able to simulate recent SH SAM trends correctly for the correct reason? § Is our ability to simulate SAM eddy feedbacks correctly somehow less important than we imagine for our ability to simulate forced responses? Gillett and Fyfe (2013), GRL

Seasonal Variation in Timescales

Eddy feedback, k>3

Eddy feedback, k=>3

Climatologically there is wave activity propagating into the mid-latitudes to the S-W of New Zealand ERA FREE

Evidence for this relationship in simplified GCM strat-trop coupling experiments Response to Forcing Dynamical Core Experiments Timescale of natural variability Simpson et al (2010)