Towards predicting climate system changes and diagnosing feedbacks from observations Gabi Hegerl, GeoSciences, U Edinburgh

1
Towards predicting climate system changes and
diagnosing feedbacks from observationsGabi
Hegerl, GeoSciences, U Edinburgh

• Thanks to Reto Knutti, Simone Morak, Susan
  Solomon, Xuebin Zhang, Francis Zwiers

Photo credits Tagesschau/NCDC
2
Estimating climate feedbacks and predicting
future changes

• Modelling approach Model as well as possible
  based on mechanisms
• Inverse / top-down approach Diagnosing responses
  and with it feedbacks from observed changes
• How to do it
• Findings
• Interpretation

3
Needed

• 1. Observations y with well-estimated
  uncertainties
• 2. Estimate of climate variability (to assess
  which observed changes can be explained without
  forcing) (observations or climate models
  checked against observed / palaeo reconstructed
  long-term variability).
• 3. Fingerprints for external forcing
  X(xi),i1..n models of any complexity that is
  appropriate for problem

4
Transient climate response relates directly to
observed attributable warming

• Estimated warming at the time of CO2 doubling in
  response to a 1 per year increase in CO2

• Separate the greenhouse gas fingerprint from
• response to natural forcings and response to
  other anthropogenic forcing (aerosol direct and
  indirect, ozone trop. And strat.)
• Estimate scaling factors ai
• u,v noise residual

5
Attributable warming.

• Scaling factors for greenhouse gas, other
  anthropogenic and natural fingerprints
• Translated into estimate of attributable warming

6
Yields an estimate of transient climate response
Fig 9.21
gt overall estimate based on rescaling diverse
individual model-based estimates Figure from
Hegerl et al., 2007 after Stott et al. 2006
7
Equilibrium climate sensitivity

• Does not relate in a simple way to observed
  warming rate, but needs estimate of ocean heat
  uptake with uncertainty
• Example last millennium

Comparison of several reconstructions with
amplitude uncertainties (dotted) with energy
balance model simulation Hegerl et al., 2006
8
Estimating ECS

• Run EBM with gt 1000 model simulations, varying
  equilibrium climate sensitivity, effective ocean
  diffusivity, and aerosol forcing
• Estimate likelyhood that residual between
  reconstruction and range of EBM simulations is
  indistinguishable from best fit residual
• Var(Res-resmin ) F(k,l)? (after Forest et al.,
  2001)?

• Account for uncertainties
• Calibration uncertainty of reconstruction
• Data noise and internal variability
• Uncertainty in magnitude of past solar and
  volcanic forcing

9
Result20th century forward modeling most
other results are top-down (asking what model
parameters yield simulations consistent with data)
remaining uncertainties ranging from large to
small
Knutti and Hegerl, 2008
10
Conclusions for large scales

• Top down/inverse approaches indicate consistent
  estimates as forward modelling, but larger
  uncertainties
• Similar approaches can be applied to constrain
  carbon cycle sensitivity (Frank et al., 2010)
• And have been applied to estimate aerosol effects
  on temperature yielding recently consistent
  estimates
• Regional changes and their effect on feedbacks,
  for example, through vegetation, are another
  matter
• Depend on seasonal changes in temperature
  distribution, and precipitation

11
Change in temperature distribution Eastern North
America 1950-2006 change in vegetation?
(Portmann et al., PNAS 2009)similarly Eastern
Asia aerosols? (from Morak)
TN90 Expected from Tmin Expected from Tmax
12
Regional circulation can be important
observation
model
SAM congruent residual

• Gillett et al., NGEO, 2008 gt attributable human
  influence

13
Feedbacks will depend on precipitation
change Mechanism Clausius Clapeyron gt wetter
when warmer Longwave forcing suppresses some of
the response Dynamics and circulation have major
influence
Equ. 2xCO2 models Transient change
Allen and Ingram, 2002
Fig. SPM-6
IPCC SPM
14
Estimate from observations

• Santer et al., 2010 detectable changes in water
  vapour
• Zhang et al., 2007 detectable changes in land
  precipitation

15
Fingerprint detection and attribution study

• Detectable signal, but larger than simulated!
• (scaling gt 0 but also gt1!)

16
Similar problems may occur in response to
shortwave forcing
Similarly, attributable change in Arctic
precipitation (Min et al., 2009) is significantly
larger than simulated

• Photo NASA after Trenberth et al., GRL, 2007

17
Global land precipitation Observed vs
models (5-yr smoothed) (from Hegerl et al.,
2007 adapted from Lambert et al., 2005)

• Shortwave Geoengineering We should be very
  worried, and not trust model simulated impacts!
• (Hegerl and Solomon, 2009)

18
Conclusions

• Changes in temperature distribution may point at
  missing processes on regional scale
• Precipitation shows a detectable human influence,
  but the observed changes are larger than
  simulated!
• Errors in models (missing / erroneous feedbacks),
  forcings, observations or all of this?
• Missing local processes and feedbacks as well as
  problems in precipitation changes will affect
  simulations of earth system feedbacks
• Top down estimates provide important evaluation
  of modelled feedbacks, and point at problems for
  regional changes and precipitation