Judith Curry has provided us with her take on an eight-year old paper by Carl Wunsch, a highly regarded oceanographer. I am sure I have read it before, but I certainly have not referred to it. I think her summary is excellent reading. I have made a few editorial changes.
From one point of view, scientific communities without adequate data have a distinct advantage: one can construct interesting and exciting stories and rationalizations with little or no risk of observational refutation. Colorful, sometimes charismatic, characters come to dominate the field, constructing their interpretations of a few intriguing, but indefinite observations that appeal to their followers, and which eventually emerge as “textbook truths.”
Consider the following characteristics ascribed to one particular, notoriously data-poor, field (Smolin, 2006), as having:
- Tremendous self confidence, leading to a sense of entitlement and of belonging to an elite community of experts.
- An unusually monolithic community, with a strong sense of consensus,whether driven by the evidence or not, and an unusual uniformity of views on open questions. These views seem related to the existence of a hierarchical structure in which the ideas of a few leaders dictate the viewpoint, strategy, and direction of the field.
- In some cases asense of identification with the group, akin to identification with a religious faith or political platform.
- A strong sense of the boundary between the group and other experts.
- A disregard for and a lack of interest in the ideas, opinions, and work of experts who are not part of the group, and a preference for talking only with other members of the community.
- A tendency to interpret evidence optimistically, to believe exaggerated or incorrect statements of results and to disregard the possibility that the theory might be wrong. This is coupled with a tendency to believe results are true because they are ’widely believed,’ even if one has not checked (or even seen) the proof oneself.
- A lack of appreciation for the extent to which a research program ought to involve risk.
Smolin (2006) was writing not about climate science but about string theory in physics. Nonetheless, observers of the paleoclimate scene might recognize some common characteristics.
Smolin’s point 7 is perhaps the most important in his list. Good scientists seek constantly to test the basic tenets of their field — not work hard to buttress them. Routine science usually adds a trifling piece of support to everyone’s assumptions. Exciting, novel, important, science examines the basic underpinnings of those assumptions and either reports no conflict or the contrary — that maybe it isn’t true. Imagine Darwin working hard to fit all of his observational data into the framework of Genesis (today we laugh at the so-called intelligent design community for doing just that).
The Hope for a Simple World
As both human beings and scientists, we always hope for explanations of the world that are conceptually simple yet with important predictive skills (in the wide sense of that term). Thus the strong desire that box models should explain climate change, or that simple orbital kinematics can explain the glacial cycles, or that climate change is periodic, is understandable. But some natural phenomena are intrinsically complex and attempts to represent them in over-simplified fashion are disastrous.
The pitfall, which has not always been avoided, is in claiming — because an essential element has been understood — that it necessarily explains what is seen in nature.
The extension of a simplified description or explanation outside its domain of applicability is of little or no concern to anyone outside the academic community, unless it begins to control observational strategies or be used to make predictions about future behaviour under disturbed conditions.
But strikingly little attention has been paid to examining the basic physical elements of “what everyone knows.”
The model problem
General circulation models now dominate discussions of the behaviour of the climate system. As with future climate, where no data exist at all, the models promise descriptions of climate change, past and future, without the painful necessity of obtaining supporting observations. The apparent weight given to model behaviour in discussions of paleoclimate can arise simply because they are “sophisticated” and difficult to understand, as well as appearing to substitute for missing data.
That models are incomplete representations of reality is their great power. But they should never be mistaken for the real world.
If a model fails to replicate the climate system over a few decades, the assumption that it is therefore skilful over thousands or millions of years is a non sequitur. Models have thousands of tuneable parameters and the ability to make them behave “reasonably” over long time-intervals is not in doubt. That error estimates are not easy to make does not mean they are not necessary for interpretation and use of model extrapolations.
Some of the published exaggeration of the degree of understanding, and of over-simplification is best understood as a combination of human psychology and the pressures of fund-raising. Anyone who has struggled for several years to make sense of a complicated dataset, only to conclude that “the data proved inadequate for this purpose” is in a quandary. Publishing such an inference would be very difficult, and few would notice if it were published. As the outcome of a funded grant, it is at best disappointing and at worst a calamity for a renewal or promotion. A parallel problem would emerge from a model calculation that produced no “exciting” new behaviour. Thus the temptation to over-interpret the data set is a very powerful one.
Similarly, if the inference is that the data are best rationalized as an interaction of many factors of comparable amplitude described through the temporal and spatial evolution of a complicated fluid model, the story does not lend itself to a one-sentence, intriguing, explanation (“carbon dioxide was trapped in the abyssal ocean for thousands of years;” “millennial variability is controlled by solar variations”; “climate change is a bipolar seesaw”), and the near-impossibility of publishing in the near-tabloid science media (Science, Nature) with their consequent press conferences and celebrity. Amplifying this tendency is the relentlessly increasing use by ignorant or lazy administrators and promotion committees of supposed “objective” measures of scientific quality such as publication rates, citation frequencies, and impact factors. The pressures for “exciting” results, over-simplified stories, and notoriety, are evident throughout the climate and paleoclimate literature.
The price being paid is not a small one. Often important technical details are omitted, and alternative hypotheses arbitrarily suppressed in the interests of telling a simple story. Some of these papers would not pass peer-review in the more conventional professional journals, but lend themselves to headlines and simplistic stories written by non-scientist media people. In the long-term, this tabloid-like publication cannot be good for the science, which developed peer review in specialized journals over many decades beginning in the 17th Century, for very good reasons.
I commend this eight-year-old summary to readers. I am trying to find the full article somewhere past the current paywall. And I was going to use Wunsch’s reflections here to deal with some comments over the past few weeks, but I’ll do that in the Comments section.