Thursday, February 04, 2010

The Medea Hypothesis (Part 8): The Misuse of a Model

So, this post is a spin out of my overly long critique post on the Medea Hypothesis. I am trying to condense what I wrote into something more palatable. I realized that my commentary on the misuse of the Franck et al model deserved a post of its own.

First to the scientists and others that are misusing models. For the love of Gawd and all that is good and light (or science, at least), JUST PLAIN STOP!

I am getting exasperated by this. I, personally, dink with models, but it is not my main duty at the day job. I make sure that the computers that the scientists use are up and running. I am privileged to work with people that are developing some really intricate models for simulating a variety of subjects: the world's climate, combustion, qcd, and fusion are a few off the top of my head.

Modeling the biosphere and its interactions with the rest of the Earth system has a moderate tradition Computer models have only become viable for most scientists to use in the past thirty years and the field of earth systematics has only really come into fruition in the past twenty to thirty years. Our understanding of the carbon cycle is ever growing and while the basics have been understood for some time, the devil is in the details.

Modeling for use by scientists up and down the far, far future and into the deepest Deep Time is a relatively young science. There are few researchers working on it. They are even working on it part time and so the ground is largely contested but slowly and the models evolve and countermodels proposed very, very glacially. Getting the various cycles right in their models for the now is very difficult. We still don’t quite have it right even for modern atmospheres and climate change (we're damned close though).

Doing it for Deep Time (and the far future) is definitely a worthwhile venture, but one ought to keep in mind that if we have a hard time working out exactly the details of a world we can sample now then we will have a harder time, with greater inaccuracies, projecting back three and a half to four billion years and forward a couple billion years. The people working on these models realize this. Unfortunately, those outside of the modeling and simulation field often don’t.

Unfortunately, this means a lot of scientists have this problem, since they are not modelers themselves.

In the Medea Hypothesis, Ward relies on Franck et al's model for carbon cycling over multiple billions of years. Franck's model is very simple. It can be run on your desktop or even laptop using Mathematica. The model itself describes how carbon is cycled and sequestered in the earth system. It takes into account continental growth, some biological activity, and even weathering of kerogen. Biomass is represented as a function of carbon dioxide and temperature. It's a nice, simple, useful tool, but its not without its issues.

For example, it cannot take into account the Snowball Earth episodes: nowhere in there can they fit these important planet wide episodes in their model's data (as yet, maybe ever). Furthermore, it cannot account for the temperature fluctuations of the Pharenozoic much at all: the ice ages that have come and gone are not represented almost at all. Likewise, the temperature spikes (Permian, Eocene) cannot be included either. The increase in vulcanism attributed to supercontinental fragmentation is also not included. Mass extinctions that are oh-so important to biodiversity and biomass cannot be included or represented in this data. There are more examples, but this is sufficient for my point.

The nifty thing is that the authors of the model acknowledge this stuff. They made it as a first step, a first brick in the wall to understand how the world works. This is a simple, yet useful tool.

The best analogy for this model is a cube in wind tunnel. Understanding how air flows around the cube is useful. It teaches us a great many things. However, no one working on aerodynamic flows has any pretensions when they are doing anything other than getting simple, basic data from their cube. They certainly are not making proclamations on how airflow will be around a fighter plane based on that cube.

Unfortunately, Ward uses the cube to describe the aircraft and make great pronunciations about the future of life on the planet. This is a mistake. A terrible mistake. The model is far too simple and Ward, being familiar with the observational/paleontological data, ought to know better.

When a model cannot take into account some very fundamental and important events as they pertain to the question of what you are trying to answer, it should not be used.

When a model contradicts key observational data, it should not be used.

It does not mean that the model doesn't have a use. It doesn't mean that this model is junk. This model is an early step in developing realistic models. Everyone jokes about the spherical best, this cow here is a sphere and it is going to be for a long, long time. Even so spherical cows do have their place.

Ward really should know better. Yet he misuses the model anyways. He even builds on it as the foundation of his future projections that the Medea Hypothesis has. This is a mistake and calls into question the Medea Hypothesis as a whole, especially since it contradicts some of his other evidence that he uses for supporting his Medea Hypothesis. There are other problems with the MH, but this is a biggie.

To other researchers. Please. DO NOT DO THIS. Thank you.

For the previous posts on the Medea Hypothesis look here, here, here, here, here, here and here.

The completed Medea Hypothesis Review Table of Contents is here.


James Davis Nicoll said...

May I link to all of these articles?

Will Baird said...


One thing.

I'm not yet done.

I am going to do a condensed version of what I sent out via email. My verbosity was too much in that and I need to boil it down some.

You may want to wait for that and the TOC post that is going to cover the whole of the review.

Anonymous said...

You know, Consider a Spherical Cow was one of my first serious modeling books, omg well over half my life ago now. It had the first non-ridiculous demographic model I'd ever seen -- for China, in fact -- breaking the population down into age cohorts and using age-dependent mortality and fertility rates.

Before that, I was very frustrated with bad MIT modeling, Buckminster Fuller, and over-stylized macro. (Weirdly, I never found a good explanation of the Solow growth model then. Not sure why.)