European temperatures prior to the 20 th century fundamentally a consequence of unforced variability? (Bengtsson et al., 2006)

Transcription

1

2 European temperatures prior to the 20 th century fundamentally a consequence of unforced variability? (Bengtsson et al., 2006) Not really (Hegerl, Luterbacher et al NGeo) Observations All forcing natural forcing IPCC SPM and chapter

3 European temperature information Instrumental Documentary Natural archive This goes into a Mann 98 type approach to produce patterns of regional T Luterbacher et al. 2004, updated; Riedwyl et al. 2008

4 Seasonal mean temperatures Reconstruction; instrumental data Multimodel mean and ECHO-G CCSM HadCM3; spread of forcings Volcanic eruptions (liberal detection: step in reconstructed AOD; record Crowley) DJF: Much more variability JJA Less intermodel correlation Much colder Little Ice Age

5 Method Multiple regression: Fingerprint f i : 1-signal: multimodel mean response (tls estimate), 5-yr aves [3 signal (ghg+aerosol; volcanic; solar): Energy Balance Model simulation of 30-90N surface temperature in response to forcing i] Estimate of climate variability: from model runs after subtracting a scaled all forcings EBM run; compared to residual in reconstructions (which is smaller) Time periods investigated: ; (more robust recon); (robust preindustrial);

6 Winter All forcing fingerprint (blue) detectable; explains at least 75% [44-133%] of warming between the late 20th century and the late 17th ( rel. to ) Only greenhouse gas + aerosol can be separated from other forcings based on EBM results (GCMs using HadCM3 work in progress) Fingerprint of forcings only detectable if some of 20 th century used (to 1950) => influence of forcing only detectable with 20 th century, but the fingerprint has structure prior to that!

7 Summer Role of all forcing not detectable in summer for whole record summer is the only season with detectable effect of forcings effect < 1900; possibly det. solar Regression of solar EBM run (ortho. to ghg) on data: suggestive, but degenerate with volcanism Why is detectability lost over the 20 th century? (bad luck? Forcing/data problems?)

8 Recent summer trends are highly unusual yr trends from instrumental to 2009, reconstruction, and individual /averaged simulations compared to 2 sigma => significant since trend

9 Shoulder seasons Spring: similar to winter; All forcing detectable Autumn: similar to summer; Loss of variance in reconstruction early on All forcing ~5% sign.

10 Effect of volcanic eruptions on European temperature Epoch analysis: average of yrs 0-4 following a volcanic eruption, relative to yrs -5 to -1 Cut off if another eruption occurs Uncertainty analysis: pick same number of random years as volcanic years

11 Reconstruction Ave 1 yr after major volcanic eruption Multimodel mean Summer Winter Pattern significance Regression of mm fingerprint on reconstruction yrs 0, 1, 2 and 1,2 combined

12 6 tropical only volcanoes Reconstruction: sign warming Multimodel Scaling factor

13 HadCM3 simulations so far A. Schurer

14 So far consistent with NH data more so with volcanoes Questions to be addressed: What is role of volcanism vs solar (Europe and NH) for cold winters? Attributable pcp changes? Changes in T distribution in Europe?

15 Large-scale increase in the number of warm nights Assemble into Giorgi regions and compare model simulated with 2 observed datasets (hand-assembled into 5x5 boxes, Hadex); Detection analysis: regression of observations on multi-model all forcing fingerprint

16 results results (black: extension to 2003) *: 5% significant change in number of warm nights in region (non-optimized) Morak and Hegerl, submitted

17 Is this Multistep attribution? This seems a useful approach for high-noise vars.. We have detected a significant change that projects on the fingerprint of external forcings TN90 correlates strongly with SAT interannually (trend subtracted); and most of the trend in TN90 is predicted based on interannual correlation with Tmean => Observed increase in Tmean probably largely due to greenhouse gas increases leaning on T results (note we cant easily estimate the contribution)

18 Zonal seasonal precipitation changes Expressed in % change Data: Vasclimo Zonal trends from Xuebin s old dataset very similar to Vasclimo. (new one not ) Work in progress: PAGODA K. Noake

19 Detection results based on nonoptimized tls approach, using 50ish model runs Detection in all seasons but su, not robust in fall Inflation strongest in spring, only modest in DJF To be submitted to GRL?

20 Conclusions European mean temperatures have been influenced by forcing in the past, and this influence varies substantially with season Winter temperatures most strongly affected by forcing Summers show detectable influence from forcing on interdecadal variability <1900, and are warming rapidly now Volcanism (and other forcing?) trigger changes in dynamics, although with low signal-to-noise ratio; nevertheless appear largely responsible for cold conditions in LIA (to be confirmed) Zonal precipitation changes again show important differences between seasons Increase in numberclimate of warm nights detectable in regions dynamics

21

22 Summary of volcanic results

23 What influenced climate? external forcings compared to hemispheric SAT reconstructions IPCC ch6 Jansen et al

24 Winter warming: what is the mechanism? CCSM ECHO implements volcanism via solar forcing change! HadCM3: known weak circulation response to eruptions Epoch analysis of NAO reconstructions: no convincing NAO response is this due to poor reconstructions?

25 Simulation volcanoes only

26 Effect of Volcanic eruptions Lots of volcanic spikes Ammann estimates 15 large tropical, 10 since 1675 Crowley estimates 6 tropical with expected effect on Europe, and timed to produce stratospheric warming Significance of winter and summer results peaks at about 17 AOD spikes in NH since

27 Evolution of volcanic response

28 Method used to reconstruct past temperature Reconstruction is inhomogeneous in time, looses variance if datasituation poor During period when observational data available (>1675) those are closely reproduced per gridbox (e.g., CET; A Schurer p.com.) Calibration period Verification NRC 2006

29 Climate model simulations Data from three coupled climate model runs: CSM (courtesy Ammann), HadCM3 (courtesy Tett), ECHO-G (courtesy Zorita) forced with different combinations of volcanic, solar (Lean, varying size), CO 2, other anthropogenic (e.g. aerosols; not ECHO-G) and land use change (only HadCM3). All models and the reconstruction moved to common grid and landmask (ECHO-G) EBM simulations to interpret results for individual forcings (from Hegerl et al., 2003; 2007)

30 Further questions -Multiple last mill runs with individual forcings may provide input about role of individual forcings, and influence of last 500yr forcings on climate mean (eg why LIA winters so cold volcanism?) -- did cold/warm periods in the past show shifted or changed temperature distribution? If there are changes, what are their causes? Effect of volcanism and forcing on precipitation in the past? 20th c changes seem larger than expected from models