Sunday, January 1, 2017

Modeling Climate Change

This will be another long blog.  I’m going to tell you why I’m extremely skeptical about any predicted climate results from existing models and simulations.  But first my qualifications.  I am not a climate scientist and have no expertise in climate or weather predictions.  I have not seen or reviewed existing climate models.  My background is in physics, electrical engineering, and most pertinent here, modeling and simulation.  I’ll start with some of my experiences, then get to my conclusions.

In the late 70’s, I was working at McDonnell-Douglas Astronautics.  The company was developing and trying to sell the cruise missile guidance set as well as the missiles themselves.  At the time that was the Air Launched Cruise Missile (ALCM) and the Harpoon.  The company was just about to start a government sponsored fly-off of the ALCM with their competitor.  I had been documenting some assembly language modules of the flight software, and was recently moved to the programming group.  My task was to take the code for the six-degree of freedom flight software (a complete simulation of all flight parameters) and produce a model for the ground control simulators that represented the missile’s terminal guidance maneuver.  That was the part of the flight where the missile pops up and then comes down and hits its target.

The company had tested the systems and was just about to start the fly-off.  I assumed the six-degree of freedom code was good.  I extracted the relevant equations and system behavior and produced my FORTRAN model.  I also plotted the flight in two dimensions to see how it behaved.  I launched my simulated missile and watched it fly towards the target—it missed!  Surely I had made a mistake.  I went over my code and the six-degree of freedom flight software multiple times.  I couldn’t find anything wrong.  Since I couldn’t deliver software that couldn’t hit a target, I added fudge factors and was able to hit the target.  Of course, I discussed this with my supervisor.  He suggested I talk to the flight programmers.  We were by then in the fly-off, and bothering them was not something I wanted to do without being sure of an issue.

So I went over to the programmers and brought up my concern.  That is, that their six-degree of freedom code—and our cruise missiles in the fly-off--could not hit the target.  They told me they were aware of the discrepancy and were working to find the problem; they had seen the error in the fly-off testing.  My conclusion here is that their simulation had significant problems, yet it was a relatively closed and simple (compared to an earth geosphere) physical system.  And they had previous test flights to collect data that could have pointed out the problem.

My next example is the simulation I created for my Ph.D. research.  I modeled the physics involved in shining a moving image on a small crystal called a PRIZ.  Only the moving part of the image came out the other side.  The crystal was essentially a moving target indicator.  For example, say you had a video of a cruise missile flying low against the cluttered ground, taken from a high-altitude aircraft.  You run the video through the crystal and see only the moving missile on the other side.  This is an idealized representation of the potential for the crystal.  Of course, laboratory testing didn’t do anything that complicated.  Previous graduate students had done a lot of lab testing.  The Russians did some really eloquent mathematical modeling, but only in one dimension.  We needed a two dimensional computer simulation to figure out how the device worked, and to make new predictions.

You start with a doped crystal.  Photons (light) impact the crystal, excite electrons, which later lose their energy and create new photons.  The crystal is non-linear, so there are some complex optical properties involved.  For this crystal, we also apply a voltage across two faces.  I had to model the electro-magnetics within the crystal as well as the optical behavior.  It was a very large and complicated simulation.  We had laboratory test results, and my simulation had to produce similar results.  When I got to the point where I could run the simulation and expected to get good results, I didn’t.  The behavior was not as expected.

For about three months I went over and over the code, refining and making changes.  I was starting to get worried.  The Air Force gives you a certain amount of time at school, then sends you back to the job, whether you have your degree or not.  Finally, I modified the boundary conditions; i.e., the constraints on behavior near the edges of the crystal.  And that change worked!  I was getting good results.  Now I just had to run enough predictive simulations, and write a dissertation, to make my doctoral committee happy.

The take-away here is that even with a small, closed system, there are a lot of variables that make it very difficult to create a realistic simulation.  In my case, I thought I had all of the significant physics represented, and it wasn’t working!  The physics at the boundary was my major problem.

Before getting to climate, I want to touch on weather modeling.  We all get the forecasts for the weather for today and the next few days.  Any news show will have it as well as online weather sites.  When there is a big storm barreling down on your location, they will usually get it right a day or even a few days out.  But they get the amount of precipitation wrong.  When it’s not a big storm, but light or mixed precipitation, they almost always get it significantly wrong.  There’s just too many variables.  My wife tells me there are only two major simulations being used so most forecasters agree or are close.  But their simulations are lousy, despite the fact they have weather data collection stations everywhere.

Now let’s move to a hypothetical model of the earth’s climate.  Here we aren’t just talking regional variability and a few days in time.  We are talking the entire globe, for years and decades.  Do you think this crew of programmers and scientists is any better than others?  As the number of variables grow, the size and complexity of the simulation grows tremendously.  You have the problems of debugging, correct physics, and being sure to include all significant variables.  I don’t know how any programmer or group of programmers could honestly say they have a high fidelity, realistic model of the earth’s environment (including solar irradiation) that would accurately predict climate going out years or decades--let alone minutes or microseconds.

The only way to even produce a simulation would be to greatly simplify land masses, bodies of water, currents, heat flows, reflection and conductance, atmospheric complexities, plate tectonics, core heating and solar irradiation.  You would have to throw out most variables and interactions, claiming they were insignificant, and hoping you were right.  Your boundary conditions would be the initial climate condition (worldwide) at some time in the past.  And then you would want to run simulations over time to see if they agreed with the historical record.

But just look at the disagreements on today’s (e.g., this year’s) temperatures.  They continuously adjust the collected raw data to account for variables they believe affect the sensors.  Maybe they are right, maybe not.  But if you can’t agree on today’s data.  What about that of the past centuries?  How are they confirming the fidelity and accuracy of their climate models when they are being run against suspect data?

From all the variables, interactions, and physics that affect our world, there are going to be climate changes.  Just like there are weather changes.  The effects we have on our environment are going to be a part of that change.  Locally, and in relatively short time frames, the effect we have can be significant.  For example, overpopulation in an area can overuse ground water reserves.  We can over hunt or fish areas and threaten other species.  We can change the atmospheric interactions by removing ground cover.  What I cannot predict is whether such changes will accumulate over time or be significant enough to override natural variables, interactions and physics.  Or even whether the resulting changes from human impact will be good or bad.

Obviously, I am not part of the consensus that believes we have to immediately stop using fossil fuels and shut down any manufacturing or personal emissions to save the planet.  I suspect most scientists that support such a consensus have thrown their professional integrity out the window and let themselves be governed by their emotions.  There may be a few that have sufficient data, expertise and wisdom to believe their confidence in the simulations is reasonable, but I don’t understand it from a physics or modeling and simulation point of view.

On the other hand, I love the outdoors, and believe in preserving it wherever possible.  I support reasonable conservation efforts.  And I also support research and development into new technologies and power sources that minimize our impact upon the environment.  What I don’t support is shutting down industries that put people out of work, mandating unrealistic and costly environmental regulations that have no significant effect on our environment, or shutting down our regional power plants.

You will note that I haven’t said anything directly about politicians or activists (non-scientists) who support the supposed “climate change consensus.”  Obviously, I believe they don’t have the background to technically agree or disagree.  I do not approve of governmental action that is not founded on good data, an understanding of that data, and confidence in the results of proposed actions.  None of those factors are present today in the climate change ‘consensus.’