Tuesday, June 13, 2017

Creating Useful Simulations

     As the warmistas are still banging their pots and pans together and shrieking that their climate models “prove” that anthropogenic global warming (AGW) is a real effect that will have catastrophic consequences if not drastically curtailed, it occurred to me that it might be useful to post some of what I’ve learned about simulation, so that those of us who are skeptical of the whole thing will have more ammunition for their tussles over it.

     I have some standing in the field, having spent forty years doing simulations for various purposes. While the technicata are fascinating to others doing such things, what really matters are the validity rules: what one must do to make one’s program a true simulation of a real process, rather than a fanciful look at a process that only exists within the simulation.

     The fundamental rule of simulation validity contains all the others:

The simulation must replicate all the relevant processes within the thing being simulated.

     The fun begins with the question “What constitutes a relevant process?”

     As we’re mostly interested in the problem of climate modeling, we’re concerned with energy flows into and out of the atmosphere: all the sources and sinks of energy that could affect the global temperature (assuming that a “global temperature” of some sort is truly relevant). Some energy sources and sinks can be dismissed as “trivial:” the input from the flapping of birds’ wings, for example. However, there are several others that are clearly significant:

  • Sources:
    1. Solar input.
    2. Cosmic radiation input.
    3. The production of heat by human energy generation.
    4. Other biological activity.
    5. Vulcanism.
  • Sinks:
    1. Heat lost to space.
    2. Heat absorption by the oceans.
    3. Heat absorption by the Earth’s solid mass.

     Each of these things is controlled by processes which must be simulated in their own right – and none of them are simple. For example, cosmic radiation, while a relatively small source of heat, affects cloud formation, and clouds are a critical factor in the insulation of the atmosphere. When the Sun is especially active, the solar wind is strong, which reduces the input of cosmic radiation. That causes clouds to form at a different rate, and to have a different average mass, volume, albedo, and longevity.

     Certain important processes are incompletely understood. For example, our Sun is a variable star. Its output varies by about four percent (4%) around its mean. Therefore, so does the solar input to our atmosphere’s energy system. As the Sun is overwhelmingly the largest source of atmospheric heat, this is a significant roadblock to the creation of a useful climate simulation. No climate model yet created has accounted for that variation, which is likely the key to explaining geologically recent climate changes such as the Little Ice Age and the Medieval Warm Period.

     Another incompletely understood process is convection at the top of the atmosphere. Energy, as every first-year physics student learns, is only transferred from one medium to another in three ways:

  1. Conduction;
  2. Convection;
  3. Radiation.

     Recent studies of the convection between the uppermost atmospheric layers and the solar wind suggest that there’s more energy transfer taking place there than was previously believed. However, because the top layers of the atmosphere are thin and interact with both the solar wind and the Earth’s magnetic field, researchers do not yet have an adequate understanding of that process. Therefore, climate modelers cannot plausibly claim to have sufficiently accounted for it in their models.

     The warmistas’ models concentrate on a couple of processes they believe dominant, and slough all others or treat them as negligible. Their favorite, of course, is the “greenhouse effect:” the increased heat retention of a gas allowed to trap an increasing amount of carbon dioxide. This effect has been demonstrated, but only in closed systems free of several important properties and processes in the Earth’s atmosphere. Those closed systems, for example, don’t form clouds. They don’t contain biological entities that react to changes in their environment. And of course, they can’t and don’t account for changes in radiative input such as the Sun undergoes.

     Scant wonder the climate models are all over the map, and have failed to produce predictions accurate even to within a generous error allowance. They’re not modeling the Earth’s atmosphere, with its complex energy-transfer and feedback processes; they’re modeling uniquely simpler systems that include only a portion of those processes, as if the others were of no importance, and claiming that their models reflect atmospheric reality.

     To sum up: a simulation is useful only to the extent that it incorporates all the significant mechanisms in the thing being simulated. If those mechanisms are themselves incomplete or uncertain, the simulation will be useless, possibly even misleading. I submit that that is a wholly adequate justification for dismissing claims of anthropogenic global warming made on the basis of climate models.

     However, it is reasonable to suspect that the warmistas are principally interested in something other than scientific knowledge. Their drive for political power, and the eager cooperation they’ve received from statists of various stripes, suggests that the models are mainly a means to an end. That end, to the best of my ability to discern its lineaments, will not be favorable to freedom.

     If I am correct in this surmise, the whole of the AGW drumbeating comes down to a search for a justification for totalitarian control over the world economy: a degree of power over human activity no tyrant has ever achieved, on the grounds that if it’s not granted to them, “we’re all going to fry.” That’s a card-palming any three-card monte hustler would be thrilled to execute successfully.

     I am reminded of a classic gag: A cop on a nightshift beat comes upon a drunk running his hands through the grass around a lamppost. Of course he asks the drunk what he’s doing. The reply: “I dropped my keys about fifty feet away.”

     Bemused, the cop asks the drunk, “Then why aren’t you looking for them over there?”

     The drunk replies, “The light is better here.”


Jean said...

Bingo! Right on target.

E. William Brown said...

There's another glaring problem with climate simulations that people always seem to overlook, probably due to lack of understanding about how simulations re built. When building a simulation of any complex process, the first version inevitably turns out to be wrong. In order to get a working model you have to compare the prototype's output to reality, and then go back through the code and figure out what you got wrong. Then you can run it again on a different scenario, and find another problem.

With a system of any significant complexity it always takes dozens, if not hundreds, of test runs before you reach the point where the model starts giving accurate predictions. But how would you do test runs against a system that we only have one example of, and where it takes several decades for measurable changes to happen? At that right it would take several hundred years to properly test your model.

So the mere fact that climate models have only been around for a few decades tells us that they can't possibly be accurate.

Francis W. Porretto said...

Indeed, EWB. During my early years in simulation I was often taken aback by the incredible complexity of natural systems -- and I'm a former physicist. Even a definitive enumeration of the inputs and outputs to a real-world system can be unexpectedly challenging. Anyone who thinks he can capture the intricacies of a multiply chaotic system such as the Earth's climate in a conventional digital simulation is either completely inexperienced at the job or has had way too much to drink.

Probably the best way to approach the simulation of a planetary climate is to assume that it can't be done, and then study ultimately simple, non-real-world cases: e.g., a perfect, absolutely solid sphere with an absolutely uniform atmosphere, a rotational speed of zero, no axial tilt or precession, no magnetic field, and an unvarying solar input from a sun that's always the same distance away. Even that would be tough to model accurately, because of the small variations in distance and angle of impact according to latitude. I'll bet the warmistas haven't tried it, though.