Global Warming: Measurement and Interpretation

Transcription

1 Global Warming: Measurement and Interpretation Ross McKitrick Professor of Economics University of Guelph Econ*4930 November Outline What do people agree on? (short list) What do people disagree on? (long list) 1

2 Outline What do people agree on? (short list) What do people disagree on? My topics: Natural variability Measurement of surface warming Testing climate models What do people agree on? CO 2 is a greenhouse gas Absorbs narrow bands of infrared energy CO 2 concentration in the atmosphere is rising This is attributable to fossil fuel use 2

3 What do people agree on? CO 2 is a greenhouse gas Absorbs narrow bands of infrared energy CO 2 concentration in the atmosphere is rising This is attributable to fossil fuel use What do people agree on? CO 2 is a greenhouse gas Absorbs narrow bands of infrared energy CO 2 concentration in the atmosphere is rising This is attributable to fossil fuel use 3

4 What do people agree on? The climate varies naturally on long and short time scales What do people agree on? (CO 2 levels vary too) 4

5 What do people agree on? There was a Little Ice Age that ended in the early 1800s Most places have likely warmed since then Glacier retreat Treeline expansion Shorter winters What do people agree on? Adding CO 2 to the air, in principle, would lead to warming in the troposphere 5

6 Why only in principle? Because the atmosphere is dynamic, not static Energy transport from the surface to the edge of space occurs through Radiation (IR flux) Convection of moist, turbulent fluids Why only in principle? Complexities that affect outcome at surface: Lapse rate Cloud formation Wind patterns Latent heat Surface albedo Ocean circulations 6

7 The feedback question If CO 2 doubles from pre-industrial levels (to 560 ppm) and everything else stays constant: Average temperature would likely rise ~ 1C To get larger effects (2 6 C) requires positive feedbacks, chiefly through water vapour accumulation Where the agreement ends In principle, adding CO 2 to the air could warm the troposphere and affect weather at the surface Questions: What are the feedback mechanisms? Is the overall effect big or small (compared to natural variability)? Is it likely to be beneficial or harmful? These cannot be answered from first principles Yet the answers have big consequences for society 7

8 Natural variability 2 contrasting views of past millennium: IPCC 1990 Natural variability 2 contrasting views of past millennium: IPCC

9 Using paleo proxies Paleoclimate analysis: trees depend on temperature, not other way around Dependent variable = f (Independent variable) Using paleo proxies Paleoclimate analysis: roles inverted Requires that we use extrapolated Dependent measures to predict Independent measures Inverse calibration Problem: uncertainty bands get inflated 9

10 The IPCC 2001 Hockey Stick 2 big claims to fame: Better library of proxies New statistical method that yielded much smaller uncertainty bands Heavily promoted by IPCC in 2001 Report The IPCC 2001 Hockey Stick Appeared 5 times in one report 10

11 The IPCC 2001 Hockey Stick This gives a fairly clear signal that this isn't just a future issue, it's happening now, Mr. Hengeveld said. Among the strongest evidence is the fact that the past century has likely been the warmest in the Northern Hemisphere in the past millennium, he said. Not only that, the 1990s ranked as the warmest decade of the millennium, and 1998 was the warmest year of the millennium in the Northern Hemisphere, which is where most of their data have been acquired. Henry Hengeveld, Canada s Chief Climate Science Advisor, Globe and Mail January 22, 2001 (emph. added) McIntyre & McKitrick work Hockey stick shape depended heavily on one small class of proxies (bristlecones) not suitable for temperature reconstructions Statistical method contained erroneous steps that underestimated the uncertainty 11

12 PC computational glitch Standard method: subtract mean, divide by standard deviation Yields series with mean=0, variance=1 PC algorithm then looks for dominant patterns PC computational glitch Mann s method: subtract mean (rather than series mean), divide by standard error 12

13 PC computational glitch Mann s method: subtract mean (rather than series mean), divide by standard error PC computational glitch Result: mean of series which trend up in 20 th century gets boosted PC algorithm picks weights that increase with size of this gap 13

14 PC computational glitch Example PC computational glitch Implications: Program efficiently mines for hockey sticks even where none exist Statistical significance claims in MBH99 were spurious 14

15 PC computational glitch Out of about 400 series, only about 20 have upward trend in 20 th century All but 1 are bristlecone pines from western USA Growth spurt widely acknowledged not to be temperature proxy Yet these series get all the weight in the final results Other one: Gaspé cedar tree The saga Lots of media coverage, including WSJ front page article Congressional investigations Formation of NAS panel (North et al.) + Ad hoc investigation panel (Wegman et al.) House hearings in

16 NRC Panel Chaired by Gerald North of Texas A&M March 2006: Steve and I presented to them in Washington, along with 10 other scientists (including Mann) We started by presenting answers to Boehlert s 2 nd Question Chair intervened: Panel had not seen the questions! The NRC had re-written the terms of reference to exclude any specific mention of Mann s work M&M: MBH method mines data for hockey stick shapes. NRC: Agreed. (p ) McIntyre and McKitrick (2003) demonstrated that under some conditions, the leading principal component can exhibit a spurious trendlike appearance, which could then lead to a spurious trend in the proxy-based reconstruction. (p. 106) As part of their statistical methods, Mann et al. used a type of principal component analysis that tends to bias the shape of the reconstructions. 16

17 M&M: The reconstruction failed an important verification test. This failure was not reported and indeed was misrepresented. NRC: Did not comment about the failure to report the low r 2 On insignificance: Agreed, in an elliptical way (P. 91) Reconstructions that have poor validation statistics (i.e., low CE) will have correspondingly wide uncertainty bounds, and so can be seen to be unreliable in an objective way. Moreover, a CE statistic close to zero or negative suggests that the reconstruction is no better than the mean, and so its skill for time averages shorter than the validation period will be low. Some recent results reported in Table 1S of Wahl and Ammann (in press) indicate that their reconstruction, which uses the same procedure and full set of proxies used by Mann et al. (1999), gives CE values ranging from to 0.215, depending on how far back in time the reconstruction is carried. You have to read between the lines to know that Wahl and Ammann Table 1S were reporting their emulation of unreported test scores in MBH98 and 99. Hockey stick verification stats 17

18 M&M: Hockey stick shape depends on bristlecones, which should not be used NRC: Agreed; again were very elliptical. (p. 107) For periods prior to the 16th century, the Mann et al. (1999) reconstruction that uses this particular principal component analysis technique is strongly dependent on data from the Great Basin region in the western United States. Such issues of robustness need to be taken into account in estimates of statistical uncertainties. (p. 50) Such trees are sensitive to higher atmospheric CO2 concentrations (Graybill and Idso 1993), possibly because of greater water-use efficiency (Knapp et al. 2001, Bunn et al. 2003) or different carbon partitioning among tree parts (Tang et al. 1999). While stripbark samples should be avoided for temperature reconstructions, attention should also be paid to the confounding effects of anthropogenic nitrogen deposition (Vitousek et al. 1997) Role of bristlecones Drives the conclusions Censored MM Mean MBH

19 M&M: Significance incorrectly calculated, unrealistically narrow confidence intervals NRC: Agreed. (p. 107) the choice of significance level for the reduction of error (RE) validation statistic is not appropriate. different statistics, specifically the coefficient of efficiency (CE) and the squared correlation (r 2 ), should have been used (the various validation statistics are discussed in Chapter 9). Some of these criticisms are more relevant than others, but taken together, they are an important aspect of a more general finding of this committee, which is that uncertainties of the published reconstructions have been underestimated. NAS Report Overall (p. 109) The substantial uncertainties currently present in the quantitative assessment of large-scale surface temperature changes prior to about A.D lower our confidence in this conclusion compared to the high level of confidence we place in the Little Ice Age cooling and 20th century warming. Even less confidence can be placed in the original conclusions by Mann et al. (1999) that the 1990s are likely the warmest decade, and 1998 the warmest year, in at least a millennium 19

20 Wegman Panel: July 2006 Edward Wegman, George Mason University David W. Scott, Rice University Yasmin Said, Johns Hopkins University John T. Rigsby III, Naval Warfare Center Denise M. Reeves, MITRE Corp. Wegman Panel Findings (P. 26): While the work of Michael Mann and colleagues presents what appears to be compelling evidence of global temperature change, the criticisms of McIntyre and McKitrick, as well as those of other authors mentioned are indeed valid (p. 28) The description of the work in MBH98 is both somewhat obscure and as others have noted incomplete It is not clear that Dr. Mann and his associates even realized that their methodology was faulty at the time of writing the MBH paper. 20

21 Wegman Panel Findings (P. 4) In general, we found MBH98 and MBH99 to be somewhat obscure and incomplete and the criticisms of MM03/05a/05b to be valid and compelling. (pp. 4-5) Overall, our committee believes that Mann s assessments that the decade of the 1990s was the hottest decade of the millennium and that 1998 was the hottest year of the millennium cannot be supported by his analysis. Subsequent events Dueling reconstructions 21

22 Subsequent events Dueling reconstructions Subsequent events Dueling reconstructions 22

23 Subsequent events Dueling reconstructions Measurement of surface temperature change 3 well-known global temperature series CRU, GISS, NOAA 23

24 What is measured vs what is sought Air temperature records versus climate data T C T: Current temperature is recorded C: Climate data shows what the temperature trend would have been if nobody had ever lived in that place Data sources Land: Global Historical Climatology Network (GHCN) air temperature Oceans: ICOADS - Sea Surface Temperature 24

25 GHCN Several concerns: Collapsing sample size Incomplete sampling at surface Locations of weather stations 25

26 Incomplete sampling at surface Locations of weather stations Incomplete sampling at surface Locations of weather stations 26

27 Incomplete sampling at surface Locations of weather stations Incomplete sampling at surface Locations of weather stations 27

28 Incomplete sampling at surface Locations of weather stations GHCN Several concerns: Growing bias towards airports 28

29 GHCN Several concerns Need for extensive adjustments to correct for urbanization etc. Data come from cities/towns/villages Long recognized that temperature changes can be due to Build up of surrounding landscape Equipment changes Poor quality control Local air pollution Waste heat from buildings and traffic, etc. 29

30 Intergovernmental Panel on Climate Change (IPCC) Acknowledges local urban heat island problem, but denies it affects large-scale data patterns Classical Urban Heat Island effects Classical UHI McKendry (2003). UHI effects have been documented in, for example: South Africa (Balling and Hughes 1996), Vienna (Böhm 1998), China (Jones et. al. 1990), Alaska (Magee, Curtis and Wendler 1999), Japan (Fujibe 1995), India (Hingane 1996), Illinois (Chagnon 1999), Korea (Chung et al., 2004), Turkey (Karaca et al. 1995), Poland (Klisik 1999), Singapore and Kuala Lumpur (Tso 1995) etc 30

31 Adjustments to temperature data They are large Outside of US there is little empirical basis US method: ADJ ( t) = POP( t) 0.45 If successful, they imply Temperature trend and local socio-economic development should be uncorrelated Nonclimatic signals in temperature data: General Surface Processes De Laat and Maurellis (2004, 2006); Kalnay and Cai (2003) Inhomogeneities Linacre (1990), Mitchell and Jones (2005) 31

32 Discriminating Hypothesis Spatial pattern of industrial activity does not match spatial pattern of climatic trends predicted by GCMs under GHG forcing (de Laat and Maurellis 2006) Spatial pattern of land-based temperature trends in IPCC gridded data should be independent of socioeconomic conditions in places where data are collected Previous and recent work McKitrick and Michaels (2004) Tested 218 GISS station data series and corresponding CRU gridded data Both exhibited significant imprint of socioeconomic data with v. similar coefficients Correction hypothesis rejected at high confidence level McKitrick and Michaels 2007 Complete sample of (available) surface grid cells Spatial pattern of trends in surface and troposphere compared to socioeconomic covariates Independence hypothesis rejected at high confidence level Both studies: nonclimatic signals likely add up to a net warming bias in global average 32

33 Temperature Data Linear trend in land-based CRU grid cells 1979:1 to 2002:12 Remove if < 8/12 months per year are available Remove if < 90% of years available Remove Antarctica Result: 440 grid cells Measurement Model Call measured trend θ i θ i is the sum of: [True climatic trend o C/decade] + [Effects of surface processes] + [Effects of inhomogeneities] 33

34 Measurement Model θ i = T i + f ( S i ) + g( I i ) Where T i = climatic trend o C/decade f (S i ) = function of surface processes g (I i ) = function of inhomogeneities For gridcell i Surface processes f (S i ) measured by p i = % growth in population density m i = % growth in real average income y i = % growth in real national GDP c i = % growth in national coal consumption 34

35 For gridcell i Inhomogeneities g (I i ) measured by g i = GDP density (GDP per square km) e i = availability of educated workers (sum of literacy + postsecondary education) x i = rate of missing observations (# missing months in cell) For gridcell i T i (ideal temperature trend) represented by Ti = β 0 + β1tropi + β 2PRESS i + β3dryi + β 4DSLPi + β5 WATER i + β 6 ABSLAT i TROP i = trend in troposphere over same gridcell as measured by satellites 35

36 Regression equation θ + + β 6 ABSLAT i = β 0 + β1tropi + β 2PRESS i + β3dryi + β 4DSLPi β5wateri i + β 7 i + β8mi + β 9 yi + β10ci + β11ei + β12 g i + β13 p x + u i i Surface proc. Inhom. GLS with clustering-robust std error matrix deg C Results Probability that effects are zero: Surface processes P = Inhomogeneities P = Joint P = (7x10-14 ) Effect on Trend of Doubling Level Population Density Real Average Income Real National GDP National Coal Use Variable SURF trop (8.61) slp (0.99) dry (0.09) dslp (-0.07) water (-1.34) abslat (0.49) g (3.38) e (-5.11) x (1.66) p (2.70) m (2.37) y (-2.19) c (3.25) _cons (-0.92) N 440 R ll P(I) P(S) P(all)

37 Population effect: Doubling population adds +0.4 o C/decade to local trend, after adjustments for population Regression against population (+geog) alone: Coef = +0.2, R 2 = 0.46 Effect is significant either way, but population alone lacks explanatory power for full S, I effects Generating clean trends Set GDP density and education to US levels Set all other surface and inhomogeneity effects to 0 ADJ Use model coeff s to generate adjusted predicted values θˆi Observed average surface trend: 0.30 o C/decade MSU average: 0.23 Adjusted average surface trend:

38 (Observed Adjusted) Conclusions Surface climatic data bears highly significant imprint of nonclimatic signals related to socioeconomic conditions in regions of origin Contamination of data not just local but regional and global in scope Potentially explains up to half of globallyaveraged trend 38

39 : studies critical of CRU products McKitrick and Michaels 2004 De Laat and Maurellis 2004, separate teams found warm bias in CRU data due to socioeconomic development over land CRU Chief: Phil Jones Produces CRU data for IPCC Was chosen by IPCC to write chapter assessing the quality of his own data 39

40 Conflict of interest July from Jones to Mann MM = McKitrick and Michaels Preparation of IPCC 2007 Report 1 st draft: no mention of our work Expert reviewers demanded it be discussed 2 nd draft: no mention of our work Expert reviewers repeated demand Peer review closed July

41 Final IPCC text: Page 244 Inserted after close of peer review McKitrick and Michaels (2004) and De Laat and Maurellis (2006) attempted to demonstrate that geographical patterns of warming trends over land are strongly correlated with geographical patterns of industrial and socioeconomic development, implying that urbanisation and related land surface changes have caused much of the observed warming. However, the locations of greatest socioeconomic development are also those that have been most warmed by atmospheric circulation changes (Sections and 3.6.4), which exhibit large-scale coherence. Hence, the correlation of warming with industrial and socioeconomic development ceases to be statistically significant. In addition, observed warming has been, and transient greenhouse-induced warming is expected to be, greater over land than over the oceans (Chapter 10), owing to the smaller thermal capacity of the land. First highlighted phrase is false Second highlighted phrase is presented with no supporting evidence; later work showed it is also untrue Many points of disagreement Is the GHCN data set a sufficiently consistent sample over time? Do adjustments fix the known biases? Does the surface data exaggerate warming? How good are the ICOADS sea surface data? 41

42 Alternate measurement systems will help resolve some of these Air: balloons and satellites Ocean: Argo floats Model evaluation Key prediction about spatial pattern: Concentration of GHG warming in troposphere over tropics 42

43 Problem: balloons and satellites don t show it Balloon record (CCSP p. 111) Satellite record ( Trend analyses Douglass et al. (2007): observations well below models 43

44 Trend analyses Santer et al. (2008): Model Confidence Intervals are wide enough to encompass observed trends McKitrick McIntyre & Herman 2010 Santer et al. only used data up to 1999 Using data up to 2009 and proper econometric methods: Models show 2-4x more warming than observations Observed trends in LT not all significant Observed trends in MT not significant Model-data discrepancies significant 44

45 This one matters EPA relied on Santer et al. paper in its endangerment finding Many other summary reports cite Santer conclusions CCSP report earlier called lack of tropical tropospheric warming potentially serious inconsistency Other points of disagreement Role of sun Direct heating effect Indirect effect on cloud formation Role of oceans Heat storage Chaotic coupling and reorganization of major circulation systems 45

46 Better data now coming out Argo network (oceans) since 2003 NASA Aqua satellite Marginal Damages: likely range Tol (2005) Most estimates in the $0-$10/tonne range 46

47 Marginal Damages: likely range Tol (2005) Stern review said $800 far outlier Most estimates in the $0-$10/tonne range The case of carbon dioxide MD flat and likely low MAC: Optimists and pessimists 47

48 Policy disagreements What is the best way to control CO 2 emissions? Economists: price mechanism (emissions tax) Industry lobbyists: Cap-and-trade/subsidies for wind&solar Politicians circa 1997: International Treaties Politicians circa 2006: National cap-and-trade Politicians circa 2010:??? Policy disagreements What is the best way to control CO 2 emissions? Economists: price mechanism (emissions tax) Industry lobbyists: Cap-and-trade/subsidies for wind&solar Politicians circa 1997: International Treaties Politicians circa 2006: National cap-and-trade Politicians circa 2010:??? 48

49 A last word Judith Curry, Climatologist, Chair of Earth and Atmospheric Sciences, Georgia Tech The major problem with the IPCC reports and some research that was exposed during Climategate was unwarranted confidence, she says. The climate s natural variability is unpredictable. Greenhouse gas emissions could offset a natural cooling trend or amplify a heating trend. It could even mean the plausible worst-case scenario is worse than anything we ve imagined, Curry says. It s a very complex scientific problem. There s a lot of uncertainty, she says. It s not that we re incompetent, there s just a lot of inherent variability. A lot of that is unknowable. The question then naturally arises. What is Judith Curry sure about? She pauses before giving an answer in three parts. Climate always changes, she says. Carbon dioxide, all other things being equal, will contribute to a warmer planet. And lastly, Whether in the coming century greenhouse gas will dominate natural variability remains to be seen. 49