Models(1) Assumptions:

Models(1) Assumptions:

There have been some strong statements in the comments of late. There has been a thread about models, including some who have noticed that my career has something to do with models and just what do I think about models. It is not something that I can write about in one blog, but I will start a thread on models, and interweave some of the other ideas that have percolated through the recent discussions.

Models: A little personal background. Many would consider me a modeler; I came to modeling with a little reluctance. I held some serious doubts about the ability to represent, quantitatively, atmospheric circulations. I imagined that there was some way that I was going to divine more information out of analytic analysis of the equations that describe atmospheric motion. What I learned, there have been many smart people who have been looking at those equations, they are difficult to solve analytically the way you might imagine solving an algebra problem, and they support an incredible variety of types of motion. With some reluctance, back in the 1970s, I cast the equations that describe planetary waves into their numerical form. Planetary waves are those long waves that you see on weather maps, anchored by the mountain ranges and the geographical distribution of the continents. I was amazed at how well the numerical solution to the equations represented not only what I saw in the atmosphere, but could also represent analytic solutions derived from more simple equations. I learned that numerical models worked. I also learned that excruciating attention and rigor are required to make a good model. (Referring back to the previous blog, these are “physical” models.)

I do not exactly consider myself a modeler. My scientific career has always been driven, first, by the observations, but I have developed and used models to interpret those observations. My development efforts have focused on building models that represent the physical mechanisms that are revealed by the observations. I had the good fortune to work with Shian-Jiann Lin (now at the Geophysical Fluid Dynamics Laboratory) and to contribute to the development of a technique used in many climate and chemical-transport models.

Many of the comments to previous blogs have stated that models are “just” a set of assumptions, and that the processes in the Earth’s climate are so complex that they defy our attempts to model them with any rigor. There are assumptions that are made when models are built, but those assumptions are not simply pulled out of a bag of magic tricks. In fact, arbitrary, unjustified assumptions are generally repelled from the modeling community because, first and foremost, the models need to describe some set of observations and the evolution of those observations, i.e. prediction. Most components of models are, just like my original modeling efforts, a representation of mathematical equations that rigorously describe the motion of the atmosphere. The idea that I and my colleagues would work from some potpourri of unjustified assumptions is, in fact, a bit offensive. (The “assumptions” in climate and weather models are often based on “statistical” models ... see previous blog.)

As for complexity, yes, each and every part of the Earth system, the atmosphere, the ocean, the land, the ice is complex. When chemistry and biology is added to the mix, the complexity increases. How to represent this complexity is a challenge; we make progress; we make mistakes. We are virtually always guided by observations. Sometimes the observations are not complete; they are only a glimpse into a process. Our attempts to model them often motivate and guide the quest to take more observations. Occasionally, we learn that our model of a particular set of observations is wrong. This is progress.

There are many directions I could go here. ---

Models are solved on computers, and this is cited by some critics as an intrinsic weakness. This criticism I do not understand. We, scientists and society, have been making such calculations for decades. Every time we fly we count on a set of model calculations for the successful flight of the airplane. The success of launching satellites and having them orbit and land where they are supposed to land relies on models solved on computers. I mention these two examples because like the atmosphere they depend on fluid dynamics (wings moving through the air), the rotation of the Earth (navigation), and basic equations of motion. The physical principles of the atmosphere, the ocean, aeronautics, planetary motion are pretty much the same. Classical physics with, now, centuries of success. The phenomena they describe are complex, complex but not magic or divine. (It amazes me - the Space Shuttle not only lands on the right runway, it lands on the little stripe in the middle of the runway.)

One more idea in this first installment on models – There are many types of models. Any observer of the weather knows that we have different scales of motion. There are tornadoes (small), hurricanes (medium), and synoptic waves, like the ones battering the California and the Sierra Nevada the last few days (large). We can write a model that describes these isolated phenomena with significant accuracy. When we put together a global model these scales are not all represented with full fidelity. A global climate model will represent, intuitively, large scales well, medium scales less well, and the small scales crudely. We use the process models and process observations to analyze the impact of the deficiencies of the climate models. We use the process models and the process observations to develop better climate models.

That’s it for the first installment. Models of the climate are complex; they are imperfect. They are based on sound physical principles that describe the motion of the atmosphere and ocean. Models are always based on observations, and their performance is evaluated with observations. While there are assumptions in the building of models, these assumptions are not arbitrary and capricious. Models are an important and validated element of weather and climate science. They should not, however, be used without critical evaluation of their capabilities.

Link to a document on modeling (and more!) below the figure. It's free to you!

r




Figure 1: The chapter linked below is in this book. (Book link.) This chapter, which I wrote, is aimed at introducing non-modeling scientists to the basics of modeling. You might find it useful. (I get no money if you were to buy this book.) If you read this, well, who knows?



Chapter 16: Fundamentals of Modeling ....