The New Middle Ages

[endicom]

PJ Media
Richard Fernandez
Jan. 2, 2017

The scientific director of a genomic research company wrote to say my last post, Witchcraft, touched on the key themes of his book The Evolution of Scientific Knowledge: From Certainty to Uncertainty published in November 2016.  The email ended with the exhortation: "Thanks, and I wish you the best in the new year with your battle against the new Dark Ages"

Edward Dougherty's book dealt with the problem I raised in my post: whether it is possible to meaningfully model problems too big to reproduce in the lab.  Readers will recall the Witchcraft post began with an observation about the validity of global warming models from a friend:

More... https://pjmedia.com/richardfernandez/2017/01/02/the-new-middle-ages//?singlepage=true

[Suppressed]

The author gets tripped up by not outlining what "modeling" is, and what the purpose of a model is, and in doing so, he throws the baby out with the bathwater. 

A model doesn't have to be a perfect 1:1 representation to be useful, or even correct, if the right questions are asked.  For example, none of us could perfectly predict exactly how much more in debt our nation will be tomorrow than today, but we can all predict it will be a positive amount.  And we might be able to come up with an estimate, even if we can't tell the exact number.  Similarly, just because I can't tell you what the temperature will be in Central Park next July 19 at 10 am Eastern, but I can tell you that it's warmer than right now.  That's why climate models can be useful, even if weather models aren't able to predict weather 100% next week. 

In a sense, it seems that Dougherty is recognizing this, but it's also what scientists are already doing.  One approach doesn't preclude the other.

I might try giving Dougherty's book a read this weekend, if I get the chance.


"Essentially, all models are wrong, but some are useful." --- George E. Box

[Joe Wooten]

A properly built and validated computer model is a great tool for doing engineering/scientific work, though more properly they are better for engineering. I build models of power plants for purposes of finding efficiency losses. You build the model with generic components that use first principle physics to operate and then tune it to the conditions the plant operates in. Part of the tuning process involves making sure it can repeat past performance data with close fidelity using ONLY the heat input to the cycle and the cooling water temperature before you turn it loose to make predictions of current performance. I've been able to find losses as small as 500 Kw in a 1200 MW plant.

The morons that make all the apocalyptic predictions of future climate do not use proper modeling techniques because planetary climate is just too damn big and complex to properly model at this time. Their models are too simplistic and cannot reproduce past weather data  just using the heat inputs from the sun and Earth's core, and the radiation losses to space. They use the same models as the weather forecasters use, and no reputable meteorologist will guarantee any forecast for more than 12-24 hours. There are just way too may variables and way too much chaotic shifts.

[Just_Victor]

The problem with the climate models as they are being used, is that the builders of these models are pretending that the output is the science.  And in doing so they ignore the behavior of the actual system they are modeling, and ignore the science contained in the real data.

I helped build the models for a new docking system being designed by NASA.  I implemented the dynamic testing that was used to validate the models.  If under any scenario the model didn't match the actual docking system behavior, then the model was invalid.  If only the climate modelers would take the same approach.

[Joe Wooten]

Yup!

if it will not closely match actual behavior, then it is worthless.