How and What Do We Know About Causation: Attribution and Fingerprinting

Transcription

1 How and What Do We Know About Causation: Attribution and Fingerprinting Ben Santer Program for Climate Model Diagnosis and Intercomparison Lawrence Livermore National Laboratory, Livermore, CA Short Course on Weird(er) Weather and a Changing Climate: Piecing Together the Puzzle Sheraton Music City, Nashville, Tennessee June 25,

2 Our evolving understanding of human effects on global climate The balance of evidence suggests a discernible human influence on global climate There is new and stronger evidence that most of the warming observed over the last 50 years is attributable to human activities Most of the observed increase in globally averaged temperatures since the mid-20 th century is very likely* due to the observed increase in anthropogenic greenhouse gas concentrations 2

3 Structure Natural and human influences on climate Studying cause and effect relationships in the climate system Identifying the top ten climate models Debunking myths about climate change Conclusions 3

4 Natural mechanisms influence climate Changes in the Sun Changes in the amount of volcanic dust in the atmosphere Internal variability ( climate noise ) 4

5 Human factors also influence climate Changes in atmospheric concentrations of greenhouse gases Changes in aerosol particles from burning fossil fuels and biomass Changes in the reflectivity (albedo) of the Earth s surface Logging roads in the Brazilian rain forest Smoke from fires in Guatemala and Mexico (May 14, 1998) 5

6 Current levels of atmospheric CO 2 have not occurred for at least the last 800,000 years U.S. Climate Change Science Program State of Knowledge Report (2009) 6

7 Structure Natural and human influences on climate Studying cause and effect relationships in the climate system Identifying the top ten climate models Debunking myths about climate change Conclusions 7

8 Why do we use computer models to study cause and effect relationships in the climate system? Source: IPCC Fourth Assessment Report (2007) 8

9 Different factors that influence climate have different fingerprints Source: Global Climate Change Impacts in the United States (Karl et al., 2009; modified from Santer et al., 2006)

10 Different factors that influence climate have different fingerprints

11 Weather balloon estimates of atmospheric temperature change are consistent with human influence fingerprints Source: Allen and Sherwood, Nature Geoscience (2008) Trend ( C/decade; changes over 1970 to 2005)

12 The changing thermal structure of the atmosphere in the latest observations and model simulations Pressure (hpa)" CMIP-5 models (Human effects)" Pressure (hpa)" Observations (Santa Rosa)" Source: Santer et al., PNAS (2013b; in review) Trend ( C/decade over 1979 to 2012)" 12

13 Fingerprint detection explained pictorially. Observations 1979" 1980" 1981" 1982" 1983" 1984" " " 2010" Spatial covariance, ANT fingerprint and OBS" 2011" 0.5" 0.0" Projection onto model fingerprint -0.5" 1980" 1985" 1990" 1995" 2000" 2005" 2010" Projection time series Model ANTHRO fingerprint 13

14 Fingerprint detection explained pictorially. Control run Spatial covariance, ANT fingerprint and CTL" 1" 2" 3" 4" 5" 6" " " 3999" 4000" 0.5" 0.0" Projection onto model fingerprint -0.5" 0" 1000" 2000" 3000" 4000" Projection time series Model ANTHRO fingerprint 14

15 Estimating signal-to-noise ratios Trend length (years)" 12" 17" 22" 27" 32" Signal-to-noise ratio" 10" 8" 6" 4" 2" Santa Rosa observations Alabama observations 0" 1990" 1995" 2000" 2005" 2010" Last year of trend" 15

16 Human-caused fingerprints have now been identified in many different aspects of the climate system Surface specific humidity Water vapor over oceans Tropospheric temperatures Stratospheric temperatures Tropopause height Ocean temperatures Sea-level pressure Atmospheric temperature Zonal-mean rainfall Near-surface temperature Continental runoff 16

17 Structure Natural and human influences on climate Studying cause and effect relationships in the climate system Identifying the top ten climate models Debunking myths about climate change Conclusions 17

18 Can we identify the best climate models? Not all computer models are of equal quality Is it a model democracy ( One model, one vote? ) Or should we pay more attention to better models? Can we define the top 10 climate models?

19 An example of the difficulties involved in identifying the best models: The case of water vapor We found that: There is an emerging human-caused signal in the increasing moisture content of Earth s atmosphere This signal is primarily due to human-caused increases in well-mixed greenhouse gases 19

20 Although the models showed important differences in their performance, they had equal weight in the fingerprint study Santer et al., Proceedings of the U.S. National Academy of Sciences (2009) 20

21 Revisiting fingerprinting with changes in water vapor over oceans We found that: Our ability to identify an anthropogenic fingerprint in satellite-based estimates of water vapor changes is not affected by screening based on model quality Model water vapor errors are very complex in space and time 21

22 Can we identify the best climate models? Santer et al., Proceedings of the U.S. National Academy of Sciences (2009) Rankings based on 20 different performance metrics Rankings based on 50 different performance metrics 22

23 Can we identify the best climate models? Santer et al., Proceedings of the U.S. National Academy of Sciences (2009) Top 4 models Rankings based on 20 different performance metrics Rankings based on 50 different performance metrics 23

24 Is the fingerprint pattern of human-caused water vapor changes sensitive to model quality information? A Mean (top 10; non) B Mean (last 10: non) C Mean (top 10; par) D Mean (last 10: par) E Variability (top 10; non) F Variability (last 10; non) G Variability (top 10; par) H Variability (last 10; par) I ALL (top 10; non) J ALL (last 10; non) K ALL (top 10; par) L ALL (last 10; par) EOF loading

25 Is the most important natural variability pattern of water vapor sensitive to model quality information? A Mean (top 10; non) B Mean (last 10: non) C Mean (top 10; par) D Mean (last 10: par) E Variability (top 10; non) F Variability (last 10; non) G Variability (top 10; par) H Variability (last 10; par) I ALL (top 10; non) J ALL (last 10; non) K ALL (top 10; par) L ALL (last 10; par) EOF loading

26 Structure Natural and human influences on climate Studying cause and effect relationships in the climate system Identifying the top ten climate models Debunking myths about climate change Conclusions 26

27 Fact or fiction? Computer models can t simulate the small warming observed over the last 10 years Over the past ten years there has been no statistically (sic) global warming. This is not at all what was predicted by the IPCC computer models.* *Professor Will Happer, Climate Science in the Political Arena Testimony before U.S. House of Representatives Select Committee on Energy Independence and Global Warming, May 20,

28 Scientific issue: How long do we have to measure temperature changes to separate signal from noise? Signal: The climate response to human influences Noise: Purely natural changes in climate Changes in the Sun and volcanic dust El Niños, La Niñas, Pacific Decadal Oscillation, etc. Identifying a human-caused climate change signal is a signal-to-noise problem The climate science community has studied this problem for 30+ years: 28

29 In addressing Prof. Happer s claim, our focus was on the temperature of the lower troposphere El Niño La Niñas 29

30 When run with human-caused greenhouse-gas changes, can models produce 10-year periods with little warming? Santer et al. (2011), Journal of Geophysical Research 30

31 When run with human-caused greenhouse gas changes, can models produce 10-year periods with little warming? Santer et al. (2011), Journal of Geophysical Research 31

32 Identifying human effects on climate is a S/N problem! S/N > 4" S/N ~ 1" Observations (Santa Rosa 1)" Observations (Santa Rosa 2)" Observations (Alabama)" Santer et al. (2011), Journal of Geophysical Research 32

33 Cherry-picking is not permissible! Source: National Climatic Data Center 33

34 Conclusions: Human-caused climate change is not some hypothetical future event it is happening now We have identified human fingerprints in many aspects of the climate system The climate system is telling us an internally-consistent story: Natural causes alone cannot explain the observed changes in climate We routinely evaluate climate model performance in a variety of different ways Claims that models are never confronted with observations are without merit Computer model errors are complex It is not easy to identify the top ten models Current best practice is to perform model selection in different ways, and determine whether different selection approaches affect things impact analysts care about Decade-long hiatus periods in warming are not evidence of absence of human effects on climate 34