Models of realityGuest author | March 16, 2012
NZCSC chairman Barry Brill has suggested to Environment Waikato that its Regional Policy Statement (RPS) should not be influenced by the climate change ‘Guidance Manuals’ (here and here) issued by the Ministry for the Environment in early 2008. Like the IPCC’s Fourth Assessment Report (4AR), their recommendations have been overtaken by recent scientific papers and data. His submission notes that modelled projections of 21st century warming rely upon two components – emission volumes and climate sensitivity. Here is his comment regarding Climate Sensitivity.
CLIMATE SENSITIVITY (Model Uncertainties)
1: THE IPCC REPORT
The 17 models used for the 4AR produced a 2100 temperature range of 1.8°C – 4.4°C. Note at page 122 of the Manual, “this arises from taking the best estimate temperature change, and subtracting 40% to get the low end, and adding 60% to get the high end of the range”. The “most likely” temperature trend is 2.7C per century.
The key driver of models is net climate sensitivity. It is well accepted that doubling pre-industrial CO2 levels (to 560ppm) will cause a direct temperature increase of approximately 1°C. This initial warming then causes a plethora of positive and negative feedbacks which eventually produce a ‘net’ effect somewhere in the range of 0.4°C to 4.0°C. Clouds (and atmospheric water vapour) are the major confounding element, having both positive and negative impacts depending upon height, type, shape, etc.
None of the models have been verified or validated in any way. The IPCC requires, however, that all models are reasonably accurate in “hind-casting” actual 20th century temperatures.
Modelling is as much an art as it is a science, and the IPCC does not contend that any one of the 17 models is correct. There is no statistical rule that accuracy can be achieved by averaging any number of inaccurate values. However, as the outputs represent the opinions of highly experienced teams of climate researchers, the IPCC regards the mean of those outputs as persuasive. While accepting that view, this submission notes the very high levels of uncertainty.
In particular, 4AR considers that the feedback effects of clouds and water vapour are highly uncertain.
2: RECENT SCIENCE
The IPCC Models consistently display a “hotspot” in the upper tropical troposphere as a “fingerprint” of greenhouse gas forcing. Despite persistent efforts with weather balloons and satellites (especially during the past five years) scientists have not been able to verify the existence of the characteristic hotspot.
Reifen & Toumi (2009) examine the proposal that a model which has successfully hind-cast past climate can be relied upon to predict future climate. The researchers found “no evidence of future prediction skill delivered by past performance-based model selection” noting that “there seems to be little persistence in relative model skill”. They speculated that the cause of this behaviour was the fact that climate feedback strengths are not stationary – “models that respond accurately in one period are likely to have the correct feedback strength at that time” but that “the feedback strength enforcing is not stationary, favouring no particular model or groups of models consistently.”
During the past two years, six peer-reviewed papers have substantially affected the calculation of net climate sensitivity:
(i) Solomon et al (2010) found that the warming observed in the 1980 – 2000 period was partly attributable to an unexplained 10% increase in water vapour in the lower stratosphere. The paper estimated that 30% of the warming recorded during 1990 – 2000 was the result of this phenomenon. Stratospheric water vapour levels returned to normal in 2001 and subsequently, causing 21st century temperatures (to date) to be 25% lower than originally projected.
(ii) Schmittmer et al (2011) applied extensive historic reconstructions to establish that the range of net climate sensitivity is 1.7°C to 2.6°C, with 2.3° being most likely. The research team found “implausible” the “fat tail” of high levels reported in the 4AR, finding only “vanishing probabilities” for a value greater than 3.2°C. This conclusion supported an earlier paper by Annan & Hargreaves (2009) which proposed an upper range limit of 4.0°C.
(iii) Lindzen & Choi (2011) observed that outgoing radiation measurements (satellites 1985-2008) disclosed net negative feedback – implying that “the models are exaggerating climate sensitivity.” The net climate sensitivity was found to be 0.7°C (range 0.5°C – 1.3°C) at a 99% confidence level.
(iv) Spencer & Braswell (2011) found that the sensitivity and cloud cover assumed by models departed substantially from the earth’s measured heat-loss during 2000-10.
(v) Douglass & Knox (2012) graph the recorded changes in ocean heat noting the heat loss of -0.03 W/m2 during 2001-9. The paper records climate shifts in 2002 and 2009, criticising the calculation of a trend across a climate shift.
(vi) Davies & Malloy (2012) of the University of Auckland reported that the average height of clouds declined by 44m/decade during the period 2000 – 2010. This decline denotes a major reduction in the global temperature trend at the rate of approximately 1.0°C/century, and indicates that net cloud feedbacks must have been negative so far this century.
The updated science puts the theoretical net climate sensitivity at 2.3°C based on paleoclimatic studies (Schmittner), or 2.0°C based on physics (Solomon). The actual observed science suggests a total level (i.e. Human Warming + Natural Variability) of 0.7°C late last century (Lindzen) and about 0.0°C to date this century (Davies).
The cumulative effect of these papers suggests that the IPCC projection of Human Warming at 2.7C/century should be adjusted down to approximately 1.0°C/century. This is about the same level as would arise from an unadjusted projection of the 20th century trend.
Professor Michael Kelly of Cambridge University remarked on 29 February in a letter to The Times:
“The interpretation of the observational science has been consistently over-egged to produce alarm. All real-world data over the past 20 years has shown the climate models to be exaggerating the likely impacts — if the models cannot account for the near term, why should I trust them in the long term?”
Model certainty at 2.7C: 20%
Model certainty at 1C: 60%