The Tyranny of Models

Paradigm 1: We now know nearly everything. Anything we don’t yet know we can work out from theory. A single observation is sufficient to validate a theory. Validated theories should not be questioned. Theories are  respected in proportion to their complexity.

Paradigm 2: We don’t know how much we don’t know. New insights are only gained by careful observation. Theories are a guide to further observation. A theory cannot be validated but a single counter-example is sufficient to invalidate a theory. The simplest theory that explains the observations is the best one to use.

Paradigm 1 is called The Science by journalists and administrators.

Paradigm 2 is called the scientific method.

To date, climate “science” has depended heavily on the first paradigm due to the predominance of numerical modelling. We use  the second approach by developing statistical methods to estimate relationships between environmental observations (i.e. between temperature data, concentration data, emissions data etc.)

Thirty years ago I presented a paper to a symposium on wave breaking at which many leading researchers in the field were present. It described a simple experiment in a wave tank which clearly demonstrated that frequency downshifting in surface gravity waves is caused by white-capping. The paper was greeted with outrage by many of those present, one of whom was given the floor for an immediate rebuttal with the argument that, if the proposal were true, many years of good work on wave models would have been wasted.

In the field of gravity wave dynamics, theory reigned supreme and meticulous experiment and observation were anathema.

Numerical models are entirely reductionist in nature and so cannot account for emergent properties of complex systems. Self-organised criticality is one such property. Fluid turbulence is another.  Even systems of non-linear differential equations exhibit the emergent property known as “chaos”.

Science has no parallel discipline similar to historiography and so there is little examination of underlying assumptions and prejudices. The seminal works of Popper and of Kuhn are thus little known among scientists themselves, although researchers in many fields pick up their ideas “by osmosis” in their post-graduate training.

Sadly this is not at all true of fluid dynamics, particularly as it applies to the environment in the form of numerical models. Generation after generation of ever more complex models succeed one another with very little actual progress, and no improvement at all in realism or relevance. Instead of the exhaustive testing describe by Popper, modellers seek only a single “validation” of their model in the manner of applied mathematicians.

Without Popperian testing, looking to disprove rather than prove, it is possible for major errors and theoretical misconceptions to be carried forward undetected.

One such error is the relationship between CO2 emissions and CO2 concentration, aka The Carbon Cycle.  I submitted a paper on this topic to Geophysical Research Letters who suggested that it would be more appropriate in a full format journal.

In other words, the editor of GRL did not disagree with it. He just did not have the courage to publish it himself.