PrintWeather Forecasts End, But Climate Forecasts Continue
No one has succeeded in forecasting the weather more than a week or two into the future, and we’re confident that such forecasts are impossible because of “chaos”. But, this difficulty does not interfere with the ability to project climate changes much, much further into the future.
For weather forecasting, you need to start with the current state of the atmosphere. If there is a cold front sweeping eastward across North Dakota in the US, areas just to the east in Minnesota are likely to experience the effects of that cold front soon. However, if the cold front has already passed across Minnesota, a different forecast will be more accurate.
This difficulty arises from the fact that no one can ever perfectly know the current state of the atmosphere everywhere (nor can we calculate perfectly, but let’s focus on the data here). If you give a good forecasting model the best available data, the model will produce a forecast that is demonstrably skillful for the next week or two, but the further you look into the future, the lower the skill, until the model is not able to predict the details of the weather. The model still produces “reasonable” forecasts—for summer in North Dakota, it will produce summertime conditions, not wintertime ones—but there is no skill for forecasting whether a cold front is coming in 26 days, or 27.
Suppose you now take your best data, and “tweak” them within the known uncertainties in the original measurements and the interpolations between the measurement stations. If the temperature in Fargo at noon on June 23, 2012 was 87.1 F, you don’t really know whether that was 87.100 or 87.102 or 87.009, nor do you know the exact temperature in all of the suburbs of Fargo that lack thermometers. So, take the 87.1 and try replacing it with the possible value 87.102, fully consistent with the available data. Make similar tweaks to other stations. Then, run the model again. What happens?
For the first few days, the forecast is almost unaffected. But, as you look further into the future, the forecast becomes more and more different from the original one. If you do this again, with different tweaks to the data (say, 87.009 rather than 87.100), you again will get almost the same forecast for a few days, but further out the forecast will differ from both of the prior ones. Do this a lot of times, and the odds are good that one of the runs will end up being close to what happens in the future, and that the average of the runs will be similar to the average behavior of the weather over a few decades (unless climate is changing rapidly!). But, you won’t know which individual run is the right one. This “sensitivity to initial conditions” is often called “chaos” in public discussions, and it means that weather forecasts can’t be accurate too far into the future. In the same way, you cannot predict the outcome of the roll of dice in a game until the dice have almost stopped moving.
Note that you can predict the average outcome of many rolls of dice, and you can predict the average behavior of weather over many years, which is climate. You may have met someone who argued that failure of a weather forecast casts doubt on climate-change projections, but that is like using one roll of dice to argue that if you keep gambling you’ll beat the casino. People who make that mistake at casinos are usually known as “poor” or “broke”.
Video: Chaos and Phil the Groundhog (PSUrockvideo) (3:50)
PRESENTER: (SINGING) As the Wheel-of-Fortune is spinning, slowing down, you can predict that just before it stops, where it's going to end. Whether a smile or frown, but for more than a few seconds, tops.
But you know before the spin, the million dollar pie, is skinnier than all the rest. You can predict it will be rare as a few weeks go by, confident you'll pass the test.
The game is chaotic, so you cannot know too far in the future just how it will go. But the wheels deterministic, as the averages show through the years.
The weather follows rules that we now know quite well. The physics cannot go away. But too far in the future and you cannot tell what will happen on a single day.
Because no data can be perfect, we can never know everything exactly, everywhere. Tomorrow's forecast is quite good, but the uncertainties grows 'til we can't tell what will occur then, there
But this chaos doesn't mean anything goes. Brazil's hot rainforest won't get Antarctic snows. The climatic averages show how the wind blows, in your ear.
If they widened the million wedge, the chances would rise, that any spin would hit it square. You still could not predict one, but no surprise, more millions would be spun up there.
If the sun brightens up, or less reflects back out, or there's an increase in greenhouse gas, that turns up the thermostat, there is no doubt. And climate change will come to pass.
And history, physics, data, models show our CO2 warms the surface here below. So, we're eating the climate as our emissions grow through the years.
But climate averages the weather, you still have to spin, and see just where the pointer stops. Sometimes you lose and other times you win. Some lovely days, and yes, some flops.
On February 2nd of another year, the faithful sun will surely rise. But will it bring shadows on a morning clear, or diffuse light under cloudy skies?
Phil, please tell us, what will March 1st bring? Sleet, snow, tornadoes, a warm day in spring? You're just as good for that as the computer thing, and you're cuter.
Phil, please tell us, what will my March 1st bring? Sleep, snow, tornadoes, a warm day in spring? You're just as good for that as the computer thing, and you're cuter.