THE DIFFICULTY IN AVOIDING GLITCHES IN COMPUTER MODELS

Our Creator created a Creation full of beautiful nuance we take for granted. Only when we try to paint it on canvas, or describe it with writing, or portray it with a computer model, do we become aware how wonderfully complex quite ordinary events are.

Seeing the following glitch is made possible by two men. The first is Dr. Ryan Maue, who takes computer read-outs that would be unintelligible to me, and turns them into maps that even a bumpkin like myself can understand. The second is the meteorologist Joe Bastardi, who has been studying weather maps since he was knee high to a grasshopper, (his father was a meteorologist), and who, while apparently unable to read code and build computer models, can scan the maps produced by models and spot “model error” immediately. He is able to do so because he has scanned decades of maps, amounting to what must be a total of over a million maps, and therefore has an understanding of what ordinarily and logically follows what. When a computer model produces something illogical, it leaps out at him from the map, like a cat with the head of an elephant would to a child playing “what is wrong with this picture.”

For example, what is wrong with this picture? It shows the “anomaly” of surface temperatures, (whether they are above or below normal), as very cold air pours down over North America. (Click to enlarge.)

Great Lakes 20141115 cfs_anom_t2m_conus_2014111500_31

If you look up to Hudson Bay at the top of the map, you can see what a bumpkin like myself often sees, which is that, when very cold air moves over warmer water, the air at the surface is swiftly raised to the temperature of the water. You can see that at the west side of Hudson Bay, where sea-ice is starting to form, the air is slightly milder, as to the east side of the bay the air is above normal, even though surrounded by below normal air.  The open water has warmed the air mass.

Now look down at Lake Michigan, at the center of the map. Again very cold air is moving over open water that is milder, but in this case the air is mysteriously cooled. How can that be? Something is wrong with this picture.

It can’t happen. What Joe Bastardi assumed is that the computer model “knows” that the lake water itself is below normal, and therefore cooled the air a predetermined amount downwards, irrespective of the fact the air was colder than the water, and it is physically impossible for warmer objects to cool colder objects.

Mr. Bastardi doesn’t know for sure that this is actually the glitch in the model, however when a computer model produces something that even a bumpkin like myself can see is incorrect, one assumes the person who created the model is quite smart, but simply ran into one of the nuances the Creator wove into Creation. In this case it is the fact that, while colder-than-normal lake water will cool summer air, as it passes over, it will not do so in the winter,  when winter air passes over.

I can’t even begin to imagine how a modeler weaves all variables involved in Creation into a model. There must be countless cases of, “If A, then B, unless C, in which case D.” I also can’t imagine how frustrating it must be when reality taps your shoulder, and sends you back into the world of code, to fix a glitch.

There must be countless cases where a model produces a virtual reality that differs greatly from reality.  In these cases a trained meteorologist is invaluable, due to his ability to spot “what is wrong with the this picture.”

A glitch like the one in the above map will have consequences in the forecast a model produces. If the air over Lake Michigan actually was colder than the surrounding air, then it would be descending, and producing clear skies and a “sea breeze” (or lake breeze) at the shore. It would not be rising and producing “lake-effect snows.”

And that is only a glitch that likely would be produced in the local  forecast. Chaos Theory often talks about about the “Butterfly Effect,” where a small change in the initial conditions produces a difference that is magnified, by the process of time passing. So this glitch will be magnified downstream, and appear as errors in the non-local forecasts, over a thousand miles away on the east coast.

In conclusion, it is dangerous to rely on models alone. It is also a bit lazy. It is far harder to do all the work and studying in that went into becoming a meteorologist of the old school. However it is well worth it, because it allows one to discern between a computer’s virtual reality, and the wonderful, real reality called “Creation”.

Advertisements

Leave a Reply

Fill in your details below or click an icon to log in:

WordPress.com Logo

You are commenting using your WordPress.com account. Log Out / Change )

Twitter picture

You are commenting using your Twitter account. Log Out / Change )

Facebook photo

You are commenting using your Facebook account. Log Out / Change )

Google+ photo

You are commenting using your Google+ account. Log Out / Change )

Connecting to %s