Chaos in the forecast

It was supposed to rain yesterday. It didn't.

Weather forecasting has always seemed to most people to be little more than a guessing game. Part of the problem is that forecasts are often made for very large areas, and weather can be an extremely local thing. Advances in recent decades resulted in impressive improvements in forecasting accuracy; if the morning was sunny and the forecast called for thunderstorms in the afternoon, I would carry my umbrella into work - and sometimes be the only one of us who made it back to my car dry.

But those days seem to be over. I've always watched the weather forecasts pretty closely. As a gardener and backyard astronomer, it's important to me to know when I will have to worry about watering my plants and when I can let nature water them for me, when I can count on clear skies at night and when I can plan to do my stargazing in a book. In recent months I have found that I can rely on the daily forecasts almost not at all.

Yesterday, for example, several different sources declared that it was absolutely certain that it would rain in the afternoon. There were maps and computer simulations and satellite images to back this up. Yet not a drop of rain fell here or anywhere else within about 30 miles. Fronts did not move as they were expected to move, precipitation did not develop where it had been expected to develop, and rain fell somewhere else, somewhere the rain had not been predicted. By the evening, everyone was backpedaling on their earlier predictions, suggesting that what had happened was exactly what they had expected would happen.

Long-range forecasts are worse. It is the nature of complex , recursive, non-linear systems like the weather to be inherently unpredictable to an arbitrary degree of precision over long periods of time. (This is one weapon in the arsenal of climate change deniers: "Just because all the data shows a consistent increase in temperature over time, there's no reason to expect that temperatures will continue to rise." They adopt the "argument from ignorance" position that since we don't know precisely what is going to happen in the future, there is no point in worrying about it or trying to change it.) Yet as greater and greater computing power has become available to even the local weatherman, long range forecasts have been pushed out from five days to seven or even ten days. If you can't tell me whether or not it's going to rain this afternoon, why should I believe you when you say that a week from Tuesday it will be partly cloudy with a daytime high of 82 and an overnight low of 65?

Long-range forecasts depend on two factors: viewers are easily impressed, and viewers have short memories. Ten-day forecasts exist simply because they are more impressive-looking than five-day or seven-day forecasts. And how many people remember what the weather forecast for today was a week and a half ago? Someday I think I will make a record of the long-range forecasts for a period of several weeks, noting how the prediction for a given point in the future changes from day to day and how it compares with the actual weather for that day.

But long-range forecasts aren't what I'm worried about. I want to know: do I need to water the cherry trees and blueberries that I planted last week? Will I be able to see the stars tonight? Once upon a time, not very long ago, I could count on the weather forecast to tell me. But I fear that one aspect of the climate change that is a fact of our lives is an increase in the chaotic dynamism of the weather. Fronts don't move like they're expected to move, precipitation doesn't form where it's expected to form, and the weather doesn't do what the weather forecasters expected it to do. So while long-range forecasts have always been suspect, now we can say goodbye to the accuracy of short-range forecasts, too.

Will it rain this afternoon? Your guess is as good as mine.