Hansen 1988 Revisited

Hansen’s 1988 temperature projections have recently received quite a bit of attention, e.g., RealClimate, WUWT and SkS. The pro-AGW sites state than Hansen has done very well, whereas the anti-AGW say that he hasn’t. Therefore, I thought that it would be a good time to revisit Hansen’s work to determine how well he did?

Temperature Sensitivity & What Can We Learn?

Dana1981 @ SkS states that:

“The observed temperature change has been closest to Scenario C, but actual emissions have been closer to Scenario B. This tells us that Hansen’s model was “wrong” in that it was too sensitive to greenhouse gas changes. However, it was not wrong by 150%, as Solheim claims. Compared to the actual radiative forcing change, Hansen’s model over-projected the 1984-2011 surface warming by about 40%, meaning its sensitivity (4.2°C for doubled CO2) was about 40% too high.

What this tells us is that real-world climate sensitivity is right around 3°C, which is also what all the other scientific evidence tells us. Of course, this is not a conclusion that climate denialists are willing to accept, or even allow for discussion.”

Perhaps. Climate sensitivity may be ≈ 3°C but we can also learn several other things as discussed below.

How Well Did Hansen Do?

Hansen Compared With the Real World

Figure 1 shows Hansen’s scenarios compared with the GISS Land-Ocean Index (LOTI). I have also added Dana1981’s data as Scenario D. This is the Scenario B data but with the temperature sensitivity reduced from 4.2°C to 2.7 °C. Dana did this by multiplying the Scenario B data by a factor of (0.9*3/4.2), which equates to temperature sensitivity of 2.7 °C (see SkS for the data). The SkS estimate for Scenario D appears to be based on Schmidt (2009).

Figure 1: Hansen’s 1988 Scenarios compared with Real-world Temperatures

It is evident from Figure 1 that the best fit for real world temperatures is Scenario C. However, the pro-AGW in SkS state that Scenario C is irrelevant because it uses the “wrong” sensitivity of 4.2°C and incorrect emissions. Therefore, perhaps I should modify my conclusion to real-world temperatures are following Scenario D, which has the “right” temperature sensitivity of 2.7°C and emissions that are close to actual emissions. It makes no difference; Scenarios C and D are similar, although Scenario D has tended to under-predict temperatures for the last 30 years or so.

2012 Projections

Hansen’s temperature projections for 2012 are compared with the LOTI data in Table 1. It should be noted that the 2012 LOTI temperature estimate is based on the 12-month running average from Jun-2011 to May 2012.

Scenario

2012 Anomaly (°C)

Comparison

With LOTI

(%)

Source

A

1.18

226%

Hansen (1988a)

B

1.77

205%

Hansen (1988a)

C

0.60

116%

Hansen (1988a)

D

0.67

128%

Dana (2011)

LOTI

0.52

100%

GISS LOTI

Note: The comparison with LOTI is based on Scenario/LOTI.

Table 1: Comparison of Hansen’s 1988 Temperature Projections for 2012

Comparing Hansen’s temperature projections with LOTI, it is evident that Hansen’ didn’t do very well.

Scenarios A and B overestimated real-world temperatures by a whopping 126% and 105% respectively. Scenario D over-predicts by 28% and even the no-increase-in-emissions Scenario C over-predicts real-world temperatures by 16%.

What do we learn? We could argue that climate sensitivity should be reduced to ≈ 2.1°C to correspond to the 28% over-prediction in Scenario D. However, I would suggest that we wait a few more years to determine the trend more accurately.

2019 Projections

The timeline for Hansen’s temperature projections for 2019 is presented in Table2. A summary of the comments made by different commentators are included to show how the favoured scenario/projection evolved with time.

Scenario

2019 Anomaly (°C)

Comparison

With

Scenario D (%)

Source

Comments
B

1.10

160%

Hansen (1988a)

In May 1988, Hansen states in AGU paper that, “Scenario A, assumes that growth rates of trace gas emissions typical of the 1970s and 1980s will continue indefinitely…[but]…since it is exponential, must eventually be on the high side of reality in view of finite resource constraints…Scenario B is perhaps the most plausible of the three cases.”
A

1.57

227%

Hansen (1988b)

In June 1988, Hansen states to US Congressional Committee that Scenario A was “business as usual.”
B

1.10

160%

Hansen (2005)

Hansen states that, “In my testimony in 1988, and in an attached scientific paper… Scenario A was described as “on the high side of reality”…The intermediate Scenario B was described as “the most plausible”… is so far turning out to be almost dead on the money.”
B

1.10

160%

Hansen (2006)

Hansen assesses the predictions and states that the close agreement, “for the most realistic climate forcing (scenario B) is accidental.” He states current estimate for sensitivity is 3 ± 1°C.
B-

1.00

144%

Schmidt (2007)

RealClimate blog, Schmidt states that forcings in Scenario B are “around 10% overestimate.”
B-

1.00

144%

Schmidt (2009)

RealClimate blog, Schmidt states that Scenario B, “is running a little high compared with the actual forcings growth (by about 10%)”
B

1.00

144%

Schmidt (2011)

RealClimate blog by Schmidt, “As stated last year, the Scenario B in that paper is running a little high compared with the actual forcings growth (by about 10%)”
D

0.69

100%

Dana (2011)

Skeptical Science blog, climate sensitivity reduced from 4.2 to 2.7°C for Scenario B. Use this as the benchmark for comparison.
?

?

?

Schmidt (2012)

RealClimate blog, Schmidt states that Scenario B, “is running warm compared to the real world (exactly how much warmer is unclear)”
C

0.61

88%

Hansen (1988a))

Hansen’s original Scenario C. This is the commitment scenario with emissions held at year 2000 levels. Include this as a measure of how well the other scenarios perform.

Note: The comparison with Scenario D is based on Scenario/Scenario D

Table 2: Evolution of Hansen’s 1988 Temperature Projections for 2019

It is evident from the timeline and narrative in Table 2 that the evolution in temperature is generally downwards; apart from the brief upwards spurt for US Congressional Committee presentation in June 1988 (more on this in unethical conduct later in this blog).

The following points are also evident:

  • There is a large reduction in the estimate for the 2019 temperature anomaly from Hansen’s estimate of 1.57°C in 1988 (as presented to the US Congress) to Dana’s estimate of 0.69°C in 2011.
  • Until recently (Schmidt, 2012) the overestimate in Scenario B was portrayed as ≈ 10% but Dana at SkS (2011) showed that the overestimate was ≈ 44%.

What do we learn? All of the pro-AGW blogs states that the Hansen Scenario B was pretty good estimate. I suggest that an error of ≈ 44% is pretty bad.

Unethical Behaviour

Hansen’s paper Hansen (1988a) was published in August 1988 but it is important to note that it was accepted for publication on 6 May 1988. This date is particularly relevant because Hansen stated on 6 May 1988 that:

Yet, one month later Hansen (1988b)
in his congressional testimony here he described Scenario A as “business as usual” (see below):


Notice that Scenario A is stressed to be “business as usual”. No mention to Congress that Scenario B was “most plausible” and that Scenario A was “on the high side of reality”.

Later (2006), Hansen re-worded his 1988 congressional testimony to be Scenario A, “was described as on the high side of reality”.


From the foregoing, it is evident that Hansen did not describe to Congress in 1988 that Scenario A was on the “high side of reality”. At best, he has been economical with the truth by re-writing history and (at worst) he has been unethical and totally unprofessional.

Conclusions

I offer the following conclusions regarding Hansen 1988:

  • Temperature forecasts (sorry, should I use the politically correct term projections?) for 2019 have plummeted from 1.57°C in 1988 to 0.69°C in 2011.
  • Estimates of temperature are in error by ≈ 60 for Scenario B and 127% for Scenario A.
  • Climate sensitivity has also fallen from ≈ 4.2°C to ≈ 2.1-2.7°C, i.e., it has fallen to 50-64% of Hansen’s 1988 estimates.

These sorts of errors do not represent pretty good estimates.

Climate Models 2011: Same Data – Different Conclusions

In his blog post 2011 Updates to model-data comparisons at Real Climate, Gavin Schmidt shows the diagram in Figure 1.

Figure 1: Real World Temperatures Compared with IPCC Model Ensemble (Schmidt, 2012)

Gavin states that, “Overall, given the latest set of data points, we can conclude (once again) that global warming continues.” My perception was that there had been some cooling over the last 15 years, therefore I have decided to check Gavin’s claims.

Gavin explains that the chart shows the annual mean anomalies from the IPCC AR4 models plotted against the surface temperature records from the HadCRUT3v, NCDC and GISTEMP products (it really doesn’t matter which). Everything has been baselined to 1980-1999 (as in the 2007 IPCC report) and the envelope in grey encloses 95% of the model runs.

At first glance the chart seems to show a good correspondence between real world temperature and the average of the IPCC models. However, the correspondence does not look quite so good when you compare the chart with the AR4 charts. I have updated AR4 Figures 1.1 and TS.26 to include the HadCRUT data up to May 2012 and discuss these as follows.

Figure 2 is derived from Figure 1.1 of IPCC AR4.

Figure 2: Global Average Temperature Compared with FAR, SAR & TAR (after AR4 Figure 1.1)

It should be noted in Figure 2 that I could not get the HadCRUT3 temperature to match exactly with the values in Figure 1.1 in AR4. Therefore, I had to adjust the HadCRUT3 data by adding 0.026 °C. I am not sure why I had to make the adjustment in the HadCRUT3 data, perhaps it is just a printing error in the AR4 diagram but this error also repeats elsewhere. It may be coincidence but the average temperature for 1961-1990 on which HadCRUT3 is based is 0.026 °C. Therefore, it may be that the AR4 chart is normalised to a zero temperature for the 1961-1990 period. However, I can find no information that confirms that this adjustment should be made.

Notwithstanding the above, it is evident from Figure 2 that the correlation between the adjusted HadCRUT3 data and the original AR4 Figure 1.1 data is very good. This applies to both the individual data points and the smoothed data. It is also evident that the temperature trend is significantly below the FAR estimate and is at the very low ends of the SAR and TAR estimates.

In order to compare Gavin’s diagram with actual global temperatures, I use Figure TS.26 from AR4 a shown in Figure 3.

Figure 3: Model Projections of Global Mean Warming Compared with Observed Warming (after AR4 Figure TS.26)

The following points should be noted regarding Figure 3 compared with AR4 Figure TS.26:

  1. I have deleted the FAR, SAR and TAR graphic from Figure TS.26 in Figure 3 because they make the diagram more difficult to understand and because they are already presented in Figure 2, in a form that is much easier to assimilate.
  2. The temperature data shown in AR4 Figure 1.1 does not correspond to that shown in Figure TS.26. The Figure 1.1 data appear to be approximately 0.02 °C higher than the corresponding data in Figure TS.26. I have assumed that this is a typographical error. Therefore, I have used the same 0.026 °C adjustment to the HadCRUT3 data in Figure 3 that was used for Figure 2.
  3. My adjusted HadCRUT3 data points are typically higher than those presented in Figure TS.26.
  4. Despite items (1), (2) and (3) above, there is very good agreement between the smoothed data in TS.26 and the adjusted HadCRUT3 data, particularly for the 1995-2005 period. It should be noted that AR4 uses a 13-point filter to smooth the data whereas HadCRUT uses a 21-point filter. Nevertheless, AR4 states that the 13-point filter gives similar results to the 21-point filter.

Comparing Gavin’s projections in the RC chart in Figure 1 with the official AR4 projections in Figure 3, the following points are evident:

  1. The emissions scenarios and their corresponding temperature outcomes are clearly shown in the AR4 chart. Scenarios A2, A1B and B1 are included in the AR4 chart – scenario A1B is the business-as-usual scenario. None of these scenarios are shown in the RC chart.
  2. Real world temperature (smoothed HadCRUT3) is tracking below the lower estimates for the Commitment emissions scenario., i.e., emissions-held-at-year-2000 level in the AR4 chart. There is no commitment scenario in the RC chart to allow this comparison.
  3. The smoothed curve is significantly below the estimates for the A2, A1B and B1 emissions scenarios. Furthermore, this curve is below the error bars for these scenarios, yet Gavin shows this data to be well within the error bands.
  4. The RC chart shows real world temperatures compared with predictions from models that are an “ensemble of opportunity”. Consequently, Gavin states, “Thus while they do span a large range of possible situations, the average of these simulations is not ‘truth’.” [My emphasis].

In summary, TS.26 from AR4 is useful for comparing real world temperature data with the relevant emissions scenarios. To the contrary, Gavin uses a chart which compares real world temperature data with average model data for which he states does not represent “truth.” I suggest that this is not much of a comparison and I conclude that the AR4 chart is a much more informative comparison.

I also conclude that it is evident from Figure 3 (AR4 Figure TS.26) that there has been a pause in global warming and that some cooling is occurring. It is certainly not as Gavin concluded that, “Overall, given the latest set of data points, we can conclude (once again) that global warming continues.” Whether or not this cooling pause is a longer-term phenomenon or temporary pause only time will tell.