Title: A brief history of time the detection and attribution of climate change
1A brief history of timethe detection and
attribution of climate change
- David Karoly
- School of Meteorology
- University of Oklahoma
2What is detection and attribution?
- Detection of significant observed climate change
and attribution of this observed change to one or
more causes is a signal-in-noise problem
identifying possible signals in the noise of
natural internal climate variations in the
chaotic climate system. - Detection is the process of demonstrating that an
observed change is significantly different (in a
statistical sense) than can be explained by
natural internal climate variability.
3What is detection and attribution?
- Attribution of climate change to specific causes
involves statistical analysis and the careful
assessment of multiple lines of evidence to
demonstrate that the observed changes are
- unlikely to be due entirely to internal climate
variability - consistent with the estimated responses to a
given combination of anthropogenic and natural
forcing and - not consistent with alternative, physically
plausible explanations of recent climate change
4Why do detection and attribution?
- To identify the causes of recent observed climate
variations - To evaluate the performance of climate models in
simulating the observed climate variations over
the last century - To constrain the projections of future climate
change
Senator James Inhofe (R, Oklahoma), Chair, Senate
Env. and Public Works Comm., in a speech to the
US Senate on Jan 4, 2005 I called the threat of
catastrophic global warming the greatest hoax
ever perpetrated on the American people
5History
Madden and Ramanathan (1980) surface warming due
to increasing carbon dioxide should be
detectable (by) the year 2000
6History
IPCC First Assessment Report in 1990 had no
attribution chapter general statement on
consistency of observed warming with model
projections
7History
IPCC SAR (1995) had detection and attribution
chapter the balance of evidence suggests a
discernable human influence on climate
8History
IPCC TAR (2001) Most of the observed warming
over the last fifty years is likely to have been
due to the increase in greenhouse
gas concentrations
9History
IPCC Fourth Assessment Report (2007) ??
10Initial studies to SAR (1995)
- Univariate analysis of global mean temperature
comparing change with internal variability - Difficult to separate different causes that
affect global radiation balance increasing
greenhouse gases, increasing solar irradiance - Use the spatial pattern of the temperature
response to differentiate between different
causes fingerprint analysis - Initially consider contrast between troposphere
and stratosphere using pattern correlation
Karoly et al (1994), Santer et al (1995)
11Changes in zonal mean atmospheric temperature
(C), 1960-1995 Modelled and observed
From Tett et al (1996), following Santer et al
(1995)
12Further studies to TAR (2001)
- Use optimal fingerprint analysis to reduce noise
by rotating detection vector away from noise
direction - Use linear regression to estimate amplitude of
forced signal from model pattern and
observational data - Applied to spatial pattern of surface temperature
change - Use more models, including models without flux
correction between the atmosphere and ocean
13Global mean surface temperature (C)
All factors
HadCM3 model (black) Observations (red)
Anthropogenic only
Natural only
Stott et al. (2000)
14Apply optimal fingerprint analysis to large-scale
variations of surface temperature at decadal
timescales
Bars show 5-95 uncertainty limits
Allen et al. (2000)
15Optimised predictions of temperature change (C),
from 1990 to 2040 under IS92a emissions (diamonds)
Allen et al, 2000
16Studies since the TAR
- Use other variables anthropogenic signal found
in ocean heat content, mean sea level pressure,
and temperature extremes - Evaluate anthropogenic signal in temperature
changes at smaller scales - Key uncertainties include relating regional
trends to anthropogenic climate change IPCC TAR
17Continental-scale studies
- Stott (2003) showed that most of the observed
warming over the last 50 years in six separate
continents, including North America, Eurasia and
Australia, was likely to have been due to the
increase in greenhouse gases
Karoly et al (2003)
18Detection of regional warming
- Calculate observed linear trend in each grid-box
and test for 95 significance (marked with )
using model control simulations to provide
estimate of natural variability of trends (Karoly
and Wu, 2005)
19Simple indices of climate variability change
- Select a small number of indices of surface
temperature variations that represent different
aspects of natural climate variability but
represent key features of patterns of
anthropogenic climate change (following Braganza
et al., 2003, 2004) - Want indices that show a common signal due to
greenhouse climate change but are nearly
independent for natural climate variations - Global mean surface temperature (GM)
- Mean land ocean temp difference (LO)
- Mean magnitude of the annual cycle over land (AC)
- Mean meridional temperature gradient in the NH
(MTG)
20Describing climate variability
21Assessing climate change
22Detection of regional warming California
- Compare observed area-mean temperature change
with model simulations for 20th century from NCAR
CCSM3 and GFDL CM2
23Attribution of regional warming California
- Probability distributions of 90-year trends in
California temperature from control model
simulations (solid line) and 20th century
simulations with increasing greenhouse gases and
aerosols (dash-dot line, 20C3M).
The observed trend agrees well with the 20C3M
simulations and cant be explained by natural
climate variations
24Attribution of regional warming Central England
temperature
- Probability distributions of 50-year trends in
CET from control model simulations (solid line)
and 20th century simulations with anthropogenic
(ANT) and natural (NAT) external forcing
The observed trend agrees well with the ANT
simulations and cant be explained by NAT
simulations
25Continental-scale temperature projections
Uncertainty plumes for changes relative to 1990s
using scalings based on continental-scale
attribution. Probabilities are represented by the
depth of shading. From Stott et al. (2006)
26Conclusions
- There have been significant advances in the
methods used for attribution of the causes of
observed climate change over the past two decades - The predictions of Madden and Ramanathan (1980)
proved to be uncannily correct - A clear anthropogenic signal can be identified in
observed climate changes over the last 50 years
in many variables and in temperature in almost
all regions