Stumble it!Bob Carter’s Mythology: Digital Fortune Telling
posted in Global Warming by themaiden |Welcome to Part IV of ‘Bob Carter’s Mythology’, my analysis of Bob Carter’s oft quoted “The Myth of Dangerous Human-Caused Climate Change”, in which Carter purports to dismantle the so called ‘alarmist’ science of global warming. Part I is a fairly light introduction. Part II digs into Carter’s claim that we have no theory of climate and hence can’t deal with what climate information we do have. In Part III, I addressed Carter’s statements about carbon dioxide. In this one, we move into the digital arena.
Under the heading ‘Can Computer Models Predict Future Climate?’, Carter takes aim at general circulation computer models (GCMs), a mainstay of climate research.
He begins by explaining that general circulation models are deterministic, that “they specify the climate system from first principles of physics”. Immediately upon first reading this, it struck me as odd that this would be a jumping-off point for a criticism of the models, as I knew it must be. Surely a model based in basic physics must be one of the best ways to construct a model? Deterministic models though, Carter explains, are flawed because our knowledge of the physics of parts of the Earth’s complicated climate system is limited. Consequently, GCMs require the use of parameterization, which Carter characterizes as ‘educated guesses’.
While his terminology is certainly chosen for its emotional appeal, there is a sense in which he is right. We don’t understand all of the gory details of Earth’s climatic system, and we do insert ‘parameters’ to cover those gaps in knowledge. Such parameterization, which Carter called ‘guessing’, would better be described as ‘doing the best we can with what we have’. Let’s think about parameterization a little more carefully.
The sad truth is that humans have complete knowledge of the physics of pretty much nothing. I don’t think this point requires much defense. Consequently, we parameterize. Consequently, all of science– all human attempts at knowledge– use parameterization extensively. We use the information we have try to make the best of it. “Still,” one might object, “we are making decision based upon incomplete knowledge of the system, and thus Carter is right. Anything could happen. The models are worthless.”
Yes, we are making decisions based upon incomplete information. We always do. We have no choice. Consider something as common as driving a car. The driver has very limited detailed knowledge of the system. A driver cannot predict the wind or the actions of other drivers. A driver can’t predict a mechanical failure. How then does a driver navigate the freeway? That driver parameterizes. That driver may not know exactly how long it would take a truck to stop, but the driver knows about how long it would take. That driver may not know exactly when a car will swerve, but you calculate the possibility anyway. Consider, really, any decision making process whether it is deciding on a savings or investment plan, or choosing the route to take when snowboarding down a mountain, or when calculating the time it takes to get to work. Consider any decision making process and it become apparent that parameterization, though rarely called that outside of the sciences, is as common as breathing. It is unavoidable. Complaining, then that modelers parameterize is a not very impressive objection. The real question needs to be whether the models are accurate enough to be useful.
Carter uses two points that could address this question. The first is that “none of the models was able to forecast the path of the global average temperature statistic as it elapsed between 1990 and 2006.” It is difficult to figure out exactly what he means here, as his statement in the article stands alone and without supporting argument or documentation. I suspect that this is a reference to the 1998– the year in which he claims global warming stopped– spike in temperature, and that is objection is that the models did not predict that spike. If that is what he means, it is no surprise. As far as I am aware, the models are not designed to provide predictions on a year by year basis, nor does anyone pretend that they do. The models average temperatures over a number of years precisely to smooth out the yearly fluctuations. It makes no sense to object that a model which averages temperatures over, say, a ten year span does not provide year by year resolution.
The second point that Carter thinks discredits the general circulation models is this: The models predict that warming trends should increase with height, while in fact the trends are either flat or decreasing. To make his point he, peculiarly, cites a report which somewhat differently than he’d have us believe. The EPA provides this summary:
- There is no discrepancy in the rate of global average temperature increase for the surface compared with higher levels in the atmosphere. This discrepancy had previously been used to challenge the validity of climate models used to detect and attribute the causes of observed climate change.
- Errors identified in the satellite data and other temperature observations have been corrected. These and other analyses have increased confidence in the understanding of observed climate changes and their causes.
- Research to detect climate change and attribute its causes using patterns of observed temperature change shows clear evidence of human influences on the climate system due to changes in greenhouse gases, aerosols and stratospheric ozone.
- An unresolved issue is related to the rates of warming in the tropics. Here, models and theory predict greater warming higher in the atmosphere than at the surface. However, greater warming higher in the atmosphere is not evident in three of the five observational data sets used in the report. Whether this is a result of uncertainties in the observed data, flaws in climate models, or a combination of these is not yet known.
It is that last item that must concern Carter. Let’s say for a moment that the discrepancy is real and not a result of instrument failures. In context, is this really enough discrepancy to justify dismissing the models?
Let’s provide even more context.
“This strongly suggests that there is no longer any fundamental discrepancy between modeled and observed temperature trends in the tropical atmosphere,” said Benjamin Santer, lead author of the Livermore-led Science Express paper and a scientist in LLNL’s Program for Climate Model Diagnosis and Intercomparison. “The new observational data helps to remove a major stumbling block in our understanding of the nature and causes of climate change. Our work illustrates that progress toward an improved understanding of the climate system requires a combination of observations, theory and models.”
In short, it looks like Carter is grossly overstating his case.
There may be, in fact, some discrepancy between theory and data, but that isn’t surprising. We are imperfect creatures in an imperfect world. And we always, by necessity, deal with limited information.
This leads neatly into Carter’s next point: that GCMs differ widely in their outputs. Of course they do. They are constructed by different people using different methods and different data sets. It is no surprise that they differ. The surprise is that so many GCM using those different datasets and methods, show the same trends, albeit with varying degrees of intensity.
Interestingly, Carter criticizes climate attribution studies as exercises in curve matching but seems to support, even to suggest, similar techniques when noting that GCM outputs can be further varied “by minor adjustments of some of the model parameters”. Obviously, changing the number would change the output, but isn’t that a bit unethical? Certainly minor adjustments could achieve such ends, but are the adjustments justified? That is a critical question and Carter doesn’t address it.
A climate attribution study is a study, to quote Carter, “the known 20th century meteorological record is retrodicted using models fed with known or presumed forcings…” These, he says, are “exercises in virtual reality, and not evidence of any type”. Is he correct? Let’s simplify the language. An attribution study starts with a climate model, which as already established, is based upon the best understanding we have of the physics of Earth’s climate. Researchers then manipulate certain components– the ‘forcings’ such as, to borrow Carter’s list, “carbon dioxide, volcanic gases, and other aerosols”– of the models in an attempt to match observed climate records. In a sense, this is curve fitting as Carter suggests. It is ‘curve fitting’ to the real world. I am somewhat puzzled as to why that is objectionable. It would seem that this is the empirical connection that Carter claims is lacking in much climate research.
Finally, Carter suggests an alternative to the ‘deterministic GCM approach’. His suggestion is to use another type of computer model, one of an empirical nature. “Such models use analysis of a portion of the climate record to establish the pattern of past temperature change and then project this pattern into the future.” I’m sure there is value in such studies, however, it seems that projecting the past onto the future will only work if all things remain equal. Things, quite clearly, have not remained equal since the beginning of the industrial age. Things, in fact, have changed radically from transportation to energy consumption, to agricultural patterns, to patterns of construction, to deforestation. It seems naive to put such faith in simple past to future projections.
I must note, though, that Carter chose to include a graph from one of these models of an empirical nature. That graph shows temperatures rising for around twenty years, then falling for twenty… but the fall drops by a third to a half of the rise. The result is that the graph Carter uses to demonstrate “21st century cooling” very clearly shows an average temperature rise weirdly like that predicted by the climate ‘alarmists’.
I want to close with this: Why use climate models?
Observed climate changes are the result of many different factors, and it is difficult to separate out the individual climate effects of these different factors using observations alone. Models are helpful in disentangling these effects.
In the real world, we are performing an unprecedented, and uncontrolled, geophysical experiment. Since the beginning of the Industrial Revolution, we’ve been burning fossil fuels (oil, coal, and natural gas) that emit greenhouse gases to the atmosphere. Land use changes, such as deforestation, also contribute.
We know, beyond a shadow of doubt, that these activities have changed the chemical composition of Earth’s atmosphere. We also know that greenhouse gases have important heat-trapping properties. It’s virtually certain that the burning of fossil fuels has made a major contribution to the global-scale surface warming we’ve observed over the last century. Estimating the contributions of human activities to regional-scale climate change is a more challenging problem.
Unfortunately, the geophysical experiment that we are performing with our planet doesn’t have a “control.” In other words, we do not have a parallel “undisturbed Earth,” with no human-caused changes in greenhouse gases, against which we could compare our present situation. Without such a control experiment, we have to resort to other means to estimate how the Earth’s climate might have evolved in the absence of human influences.
We do this by performing the idealized control experiments that we can’t conduct in the real world. In a computer model, we can fix atmospheric levels of greenhouse gases at pre-Industrial values. We then run the model with these fixed greenhouse gas levels, and simulate hundreds of years of undisturbed climate. This provides invaluable estimates of “climate noise”—the climate variability that arises purely from natural interactions between the atmosphere and ocean. El Niños and La Niñas are good examples of natural climate noise.
Come back for “Is there a Consensus?“
Popularity: 1%
