Transcript Hasselmann, 79 etc, Allen + Tett, 99

Detection of anthropogenic
climate change
Gabi Hegerl,
Nicholas School for the
Environment and Earth Sciences,
Duke University
01-12-2000
1
Temperature trend 1901-2000
01-12-2000
2
Fingerprint methods: lin. regression
Estimate amplitude of model-derived climate
change signals X=(xi),i=1..n
from observation y
Best Linear Unbiased Estimator

y  ai xi  u
u: noise residual
(Hasselmann, 79 etc, Allen + Tett, 99)
Vector: eg Temperature(space,time), scalar product: Inverse
noise covariance
Signal pattern from model, amplitude from observation!
01-12-2000
3
June-July-August Greenhouse gas + sulfate aerosol
01-12-2000
4
uncertainty range

Estimated from coupled model internal
variability

Safety checks:
– Use model with strong variability
– test consistency with observed noise
residual u
01-12-2000
5
Contribution of greenhouse gas and
sulfate aerosols to to trend 1949-98
o: Greenhouse gas +
sulfate aerosol simulation
+: Greenhouse gas only
o/+ inconsistent with
observation
Ellipse: 90% uncertainty
range in obs. Signal
estimate
from: Hegerl and Allen,
2002
01-12-2000
6
The longer perspective
reconstruction of
NH warm season
temperature
Forced component
Fat: best fit to paleo
Thin: 5-95% range
*: significant
01-12-2000
7
Conclusions global/NH SAT




01-12-2000
8
Significant climate change observed
Uncertainty in distinction between
forcings, but:
“Most of the recent (last 50 yrs) global
warming is likely due to greenhouse
gases”
Significant and consistent climate
signals in long temperature records
Towards detection of anthropogenic
changes in climate extremes



01-12-2000
9
How to compare course-grid model with
station data?
Can daily data be substituted by
monthly/annual and shift in distribution =>
no
Which index to use for early detection
(avoid baseball statistics!) that is
moderately robust between models?
Change in once/few times/yr events robust
and strong
Changes in precipitation extremes
stronger
01-12-2000
10
Change in rainfall wettest day/yr NAmerica
Consensus
Observations show
overall increase, too
01-12-2000
11
Annual mean precip
changes consistent
between two models
Wettest day/yr
Wettest 5 consecutive
days
01-12-2000
12
Results: Anthropogenic vs natural
signals, time-space
Bars show 5-95% uncertainty limits
01-12-2000
13
Allen et al, 2002
Annual mean rainfall change NAmerica
consensus
01-12-2000
14