Climate Change Weekly #425: Climate Model Problems Persist, Changes Reduce Accuracy Further

How can their simulations be “improved” when, as I discussed in Climate Change Weekly 407, the modelers themselves were forced to admit, just weeks before CAR6 was released, that the models were projecting even hotter temperatures and steeper temperature trends than the previous iteration, the simulated temperatures of which were already too hot, failing to represent measured temperatures accurately?
 
Only a government bureaucracy or a con artist could claim with a straight face a technology has improved when it does not perform its required task as well as poor-performing previous versions. It’s like confidently asserting a class of electric vehicles is improving based on laboratory modelling even as the miles they can travel between recharges is declining and the amount of time it takes to recharge them is getting longer. Worse performance is not better, unless the goal is to fail.
 
The fact that computer models are flawed and produce untrustworthy climate projections has long been recognized. Reports by the National Center for Policy Analysis (which I edited when I worked there) in 2001 by environmental scientist Kenneth Green, Ph.D. and in 2002 by David Legates, Ph.D. (who was then the director of the Center for Climatic Research at the University of Delaware-Newark) detailed the numerous flawed projections computer models had made, and they explained why the failures occurred and were likely to continue to be the norm.
 
Legates wrote,
 

Whereas computing power has improved markedly over time, modelers’ knowledge of the myriad factors and interconnections that drive climate change has not. In part, this is because the IPCC has always focused on understanding the human factors that affect climate, to the exclusion of other factors, even though it admits other factors do have some effect.
 
To their credit, CARs 1 through 5 acknowledged that natural factors—the Sun, clouds, ocean currents, etc.—play at least some role in climate change, however poorly understood at the time. The IPCC has provided lists of factors, natural and human, that affect temperatures. The list has changed over time, as have the estimates of the direction the various factors drive temperature and by what amount. What never changes, however, is the amount of confidence or degree of understanding the IPCC has about the temperature effects of non-anthropogenic factors, because study of these is largely ignored. Previous CAR reports consistently admitted the IPCC had low or very low understanding of each of the natural factors that drive temperature changes. Nonetheless, the IPCC has been confident in dismissing them as significant sources of present climate change.
 
That is like trying to understand how a car functions, admitting you know nothing about radiators, alternators, timing belts, oil pumps, and myriad other systems, but being confident that having a full tank of gasoline and a key in the ignition are the only real factors that make a car function. Then, when the car doesn’t start, you state with confidence the only reason it could have failed was because the key was broken or the car was out of gas.
 
The IPCC got away with this nonsense by claiming that when they run their models without carbon dioxide it does not produce the warming they expect, regardless of changes in assumptions about the other factors, but when they add carbon dioxide, the models produce significant warming. That’s circular reasoning at best and idiotic at worst. It should not surprise us that the assumptions modelers make about carbon dioxide and other greenhouse gas emissions are the only ones that produce the results they expect and are getting.
 
In CAR6, the IPCC abandons even the appearance of scientific curiosity about nonhuman factors’ effects on climate. If you read only AR6’s summary for policymakers, you wouldn’t know clouds existed unless humans caused them by creating aerosols. Yet water vapor is by far the dominant greenhouse gas, accounting for more than 97 percent of all the greenhouse gases in the atmosphere, and clouds have huge long-term and short-term effects on surface temperatures. The IPCC acknowledged as much in previous AR reports, admitting climate models account poorly for the role changes in cloud cover play in climate change.
 
AR6 virtually ignores any effect the Sun has on climate change. The report barely acknowledges solar irradiance as having any role at all in climate change, in a graphic on page SPM-8 (Summary for Policy Makers). There is no mention of solar cycles, which we know from history correlate with climate changes. Nor does the report even mention that increases and decreases in cosmic rays resulting from solar fluctuations affect cloud cover and thus temperatures. Except for volcanoes, all other factors—such as large-scale decadal ocean circulation patterns—are lumped into a category called “Internal Variability” to which AR6 attributes almost no effect on climate change.
 
Climate modelers’ response to the fact that their models perform poorly and their performance has worsened over time is not to admit it is a matter of “Garbage in, Garbage out” that should lead them to question their fundamental assumptions about whether human greenhouse gas emissions are the sole or even dominant factor driving temperature changes. Instead, as The Wall Street Journal reports,

 
That’s right, climate modelers’ response to the consistent failure of their models to reflect real climate conditions is to spend more money and time adding complexity to their models. Complexity is not in and of itself a virtue.
 
The climate system is certainly complex. Even so, there is no reason to believe making models more complex will make them more accurate. We don’t adequately understand all the factors that drive climate changes or how the Earth responds to different perturbations on the overall climate. What one doesn’t understand, one can’t model well. Absent that basic understanding, adding more lines of code and making increasingly complex assumptions about climate feedback mechanisms that are even more poorly understood than the basic physics only makes models more error-prone. Every line of code and every complex calculation is one more area, formula, or operator where a flawed assumption or simple mistake of math or code punctuation can cascade throughout the model. Complexity introduces more opportunities for errors or “bugs” in the code, which can throw off the projections.
 
The fact that as modelers make their models increasingly complex their simulated climate outputs increasingly diverge from real-world climate data should serve as an indicator complexity is a weakness of the models. Modelers simply don’t know what they don’t know. That’s a fact they should admit, instead of building their ignorance into their models by pretending elegant mathematical formulae reflect reality simply because they are elegant and complex. The first step in getting out of a hole you have dug is to stop digging.
 
A second indicator that complex climate models are inherently flawed is the fact that simpler climate models perform better in matching real-world temperature data. Simple models reject assumptions about how different aspects of the climate system will add to or reduce relative warming as greenhouse gas emissions rise. Absent the additional forcing from modeled feedback mechanisms or loops, simple models project a modest warming in response to rising emissions. In this respect, the simple models reflect well what Earth has actually undergone.
 
There has been no runaway warming, and there is little or no reason to expect such a thing to occur from any reasonably expected future rise in atmospheric greenhouse gas concentrations. If models don’t get right their basic projections—temperatures—there is no reason to trust their ancillary or projected secondary effects which are supposed to be driven by rising temperatures.
 
SOURCES: Intergovernmental Panel on Climate Change; The Wall Street Journal; Powerlineblog