Patterns vs. Models

Humans are very good at seeing patterns. We even find patterns where they don’t really exist, thus reminders such as “Correlation is not causation” and “post hoc ergo propter hoc.” We also build mental and mathematical models. We create structures to help us organize and predict what the world will do. Some of these are very durable and have stood the test of time and experience. Others . . . don’t do as well. And a few are so far out there that a lot of people enjoy listening to them but don’t believe them. [I.e. “I’m not saying that it was necessarily aliens, but . . .”]

One thing historians, archaeologists, and other people who live in the past do is look for patterns. Sometimes literally, so that we can identify the culture that created [thing], or for hints of written communication. We study aerial photos and satellite images searching for traces of missing or long-gone dwellings, forts, fields, and roads. We read accounts of events and goings on, hunting for hints about the bigger picture. A lot of environmental history, especially once you get to the centuries before modern thermometers, barometers, and the like, is combing through diaries, tax reports, inventories of foodstuffs and fiber, and government or corporate forms, trying to suss out what was going on in the background that no one bothered to write about because everyone knew what was happening. Like Dagomar Degroot pouring over ships’ logs and harbor reports to determine what the weather was in and around the Dutch Republic in the 1600s-1700s.

Our patterns are based on the past. We track recorded events and happenstances and compare them to modern, or make note of how people responded to storms, floods, freezes, and droughts. As Degroot points out, emperical data don’t tell us about how storms affected people’s lives, and how people adapted. The Dutch developed a number of adaptations that allowed them to survive the Little Ice Age in much better shape than did other places, but they still suffered. (The Wars of Independence [80-Years War and Anglo-Dutch Wars, and Louis XIV’s wars] didn’t help.) We’re looking at weather and climate events that already happened.

Weather forecasting tries to sort out, based on physics, chemistry, geology, and past events, what will happen in the near future. Anyone who puts their faith in a long-term weather forecast to, oh, plan a hiking trip, or an outdoor wedding, needs to have a back-up plan, unless he lives in one of those places with certain climates. No one in Jerusalem, for example, will plan an outdoor-only event for December that requires warm weather and sun, because winter is the rainy season. Likewise people who live in the Rocky Mountains know that thunderstorms form around two in the afternoon. Minnesotans assume that February will be cold and March-April will have mud. But to foretell on January 18 what will transpire on March 7th? Not likely. Even a week or 10 days from today is . . . fraught. If you are in Texas, it will be warm, possibly dry or not, perhaps windy or not, maybe humid or not. How warm? Above 60 F is as far as I’m willing to go, and I won’t put money on that.

Climate forecasting? Relies on models. Models are mathematical constructs of a very simplified world, with certain variables that can be adjusted. Emphasis on constructs and simplified. There is no climate prediction model yet that can deal with all the variables. Carbon Dioxide changes? Humidity changes? Heat islands? Effects of wind turbines? The occasional random equatorial volcano coughing sulfur dioxide and ash into the atmosphere? All at once? Splat! That was the model collapsing. Most of the most common models used by the IPCC and others can’t even retrocast accurately – that is, you can’t feed in the data for a date and location in the past and get the actual weather or climate for that time and place.

Models are very useful, so long as the user observes the limits in the model. The local weather guys and gals, especially the ones with several years of local experience, temper the models with “I’m just not entirely sure about this because of X, but here’s the National Weather Service/National Severe Storm Center Forecast/ European Model prediction.” Those of us with a lot of on-the-ground knowledge and regional research are not surprised when the models are off, or gee, winter can be very cold and still, or summer can be very hot and still, or that Texas gets freezes as far as Corpus Christi.

I’ve read through the diaries and reports of ranch managers and farmers from this region, going back as far as they exist, along with US Army documents and Indian Bureau reports. There were months where [due to a high pressure dome] the wind didn’t blow and windmills didn’t work. Cowboys had to wind ropes around the shaft and ride away from the thing, repeating that over and over to get the pump to bring up water for the cattle. Or there would be spells of miserable heat in an otherwise cold year. Or a hard, cold and wet winter in a decade of heat and drought. Snowvid 21 wasn’t all that unusual, really. High pressure building in and baking the Southwest, or Texas, or the Great Plains, isn’t too rare in the long-term. Even during the Little Ice Age.

Patterns and models. I work with patterns. I’m good at seeing patterns. I try not to make predictions, unless they are based on long experience and human nature. (Teenagers are going to be emotional. Toddlers will melt down. Someone’s going to tap the electric fence, because it might not really be live.) Models, especially models that claim they can determine what is going to happen and why a hundred years from now, or ten years from now? I fold my money and put it back in my pocket, as the gamblers say.


9 thoughts on “Patterns vs. Models

  1. I’d “love” it if the Climate Change folks put in the data from 30 years ago into their models to see if they can “predict” today’s “climate”.

    But they wont’. 😈

    • In at least one case my Dad read about, one of them did, and no, the model utterly failed to predict reality. To the point the scientist doing the research began wondering if the role of carbon dioxide in climate (and models) is being significantly overstated. How much attention do you think the MSM gives that?

      When there are scientists that are intellectually-honest enough to do such research, how many top-tier journals are going to accept an article on that? How many major conferences are going to allow them to present? How much attention is the media going to give it? To say nothing of funding research that involves more than existing datasets and some computer models.

      Remember how much climate research is federally funded. To quote President Eisenhower: “The prospect of domination of the nation’s scholars by Federal employment, project allocations, and the power of money is ever present and is gravely to be regarded. Yet, in holding scientific research and discovery in respect, as we should, we must also be alert to the equal and opposite danger that public policy could itself become the captive of a scientific-technological elite.”

      • OC2 is a lagging indicator, not leading. Think of the oceans as soda pop. When soda pop gets warm, what happens to the fizz? It goes away. When it is cold, it holds onto the fizz. Thus oceans with carbon dioxide. Plus all the activists ignore the fact that higher CO2 means more plant food means more growth means more CO2 pulled out of the air by plants. (One intriguing theory is that the “Snowball Earth” extinctions were caused by a population explosion of algae that pumped huge amounts of O2 into the air, which seriously messed up the energy balance for a while.)

  2. “you can’t feed in the data for a date and location in the past and get the actual weather or climate for that time and place.”

    Why, sure you can….. once the data has been properly “adjusted” by our experts. How? We only allow fellow cultists to see that.

  3. Most of the most common models used by the IPCC and others can’t even retrocast accurately – that is, you can’t feed in the data for a date and location in the past and get the actual weather or climate for that time and place.

    Which is a Big Problem, because that’s a very basic test. It’s one of the ways you MAKE a decent model– check for how accurately it responds to what you know happened.

  4. There are broadly, two communities with regard to forecasting fluid pressure, temperature and speed.

    One is meteorologists and climate scientists. And standards meteorologists are held to are actually quite forgiving. They have to be, because it is actually a little bit hard to experimentally replicate the case of a spherical shell of fluid, bound by gravity.

    Second is mechanical engineering, chemical engineering, aerospace engineering and maybe physics. Their approach to fluid mechanics breaks down into analytical, numerical, and experimental areas of focus. Analytical is straightforward and difficult; the continuum model starts from conservation of mass, linear momentum and energy to develop a set of equations that do not have a closed form general solution. Which means that the equations suck to work with, you have to really know what you are doing, and sometimes you are SOL anyway. Which is why numerical solutions, aka CFD, are being used more often. You still have to know what you are doing, and if you blindly trust the computer to make the right choices, it will probably spit out a wrong answer even if you feed in correct information. If you are trying to answer an unsolved question, and cannot calculate data to figure out right choices from, what do you do? Well, fluid mechanics is extremely dependent on very specific empirical testing. There is a reason why wind tunnels are a fundamental tool for aerospace engineers.

    Anyway, electrical engineering has a concept called systems identification. Feed in sensor readings, extract system behavior.

    Climate modeling appears to basically be this, but without the rigor appropriate to engineering.

    24 hours is about the shortest temperature period we can identify, and, opposite of electrical engineering, we are looking for much longer periods/smaller frequencies.

    With the Sun, the only obvious upper bound we can put on period is the life of the sun.

    Fundamentally, with earth thermal periods, we can set a much lower upper bound. Because of continental drift, and the scale of the movements we are trying to measure. Because the configuration of continents has an effect on the fluid flows we are measuring, the models are only valid for a single configuration. So, you take the length of time the continents are ‘close’ to that configuration, and that is your upper bound on estimated period. What is that length of time? Who knows? Again, situation where experimental measurements are /hard/.

    What verifies that a CFD model is correct? The model numbers are close to the measured numbers. CFD models always have an inherent error caused by the discretization/numerical solution process. You stop messing with the model when the error gets low, and call it good enough for what you are trying to do. Within ten percent of correct is maybe good enough for engineering purposes. 10% of 0 C to 25 C is around 25 or 30 C. (271 + X Kelvin, where X is 0 or 25.)

    Fundamentally, the time, space, and temperature resolution we may care about is very fine. The ‘data’ we have is crude, and the sets of equations that we can numerically solve are inherently of crude resolution. Barring magic, the predictions we can get are cruder than we would care about, and failure to carefully disclose that is dishonesty.

    This shit is far noisier than they make it seem. They are outright lying to be saying that they have done the work well and reliably enough to be making decisions of human welfare off of.

    • Ooooohhh yes. When Bjorn Lomborg popularized the term “watermelon” for politicians who use environmental concerns as a justification for enforcing socialist/communist policies, he was spot on. There is no way, even with the supercomputers we have today, to handle the complexity of an open system the size of Earth’s weather and make reliable long-term predictions about which policies will work for what outcome.

  5. Having dealt with ‘modelers’, my general response is to shoot them and get on with life… And there was a reason we called the weather people ‘weatherguessers’… Whatever they predicted, we usually got the exact opposite ONSTA… sigh

  6. Ptolemy had a model of the solar system. So did Tycho Brahe, who claimed that all the other planets went around the sun, but the whole kit and caboodle was then dragged around the earth. Brahe’s model was dominant in the popular mind at the time of Galileo but Galileo thought it was so stupid that he didn’t even address it directly in his book on Two World Systems. I asked my father many years ago what the heavens would have looked like if Ptolemy had been true. He gave me the squint eye and said everything would have looked the same; those guys were all explaining the same set of observations.

Comments are closed.