Understanding climate and climate change is an endeavor that has consumed the entire careers of a large number of scientists, and there are still unanswered questions.

Some of these questions will be answered as more and more people become climate scientists and study them. Others may never be answered.

But perhaps the most difficult concept to grasp is that some climate- and weather-related questions are simply impossible to answer with certainty. Here’s one: Will it be raining at noon at my house on this day a year from now?

Now, depending on where your house is, it may be possible to give a pretty good answer. If you live in the Sahara Desert, an answer of “no” is probably going to be right. And if you live in Indonesia, and “now” is the rainy season, “yes” is a pretty good answer. But these are just the easy cases, and they use climatology for their answers. That is, we know from the behavior of the climate that is doesn’t rain in the Sahara Desert and that it does in Indonesia in the rainy season. What this implies is that answers based on climatology have some probability of being right. In these two examples, the probablity of no rain in the Sahara Desert a year from now is probably 99% or better. And the probability of rain in Indonesia (assuming that a year from now is the rainy season) is probably 90% or better. But at noon? Well, maybe the probability goes down to 75% or maybe as low as 50%.

What this says is that there is uncertainty in the answer to this question. And, in these examples, the uncertainty is due to the fact that we’re using the climate as our guide.

There are other guides as well. For example, if we’re asking about rain tomorrow, we could use a weather forecast for guidance. Weather forecasts (at least 24-hour ones) don’t use climatology; instead, they use computer models. A computer model of the weather takes the physical principles that govern how the atmosphere works (such as Newton’s Laws), puts them in a form that a computer can solve, and uses the weather now to forecast the weather at some time in the future. And these weather models are actually pretty good—much better than climatology, for tomorrow’s weather.

So, with the guidance of a computer-generated weather forecast, we might find out that there is a 50% probability, or maybe higher or lower, of rain at noon at your house tomorrow. What this means is the following: If this particular forecast (the 50% probability forecast) is made 100 times, the forecast is correct if it rains 50 of those times. If we knew in advance which 50 times it was for sure going to rain, we would be able to make 100% probability forecasts on those times and 0% probability forecast on the other ones.

Now, this doesn’t sound like we know much about forecasting. However, forecasts like this aren’t made for particular locations (like your house) or for precise times (like noon). They’re made for larger areas (such as your town) and for time periods (such as “early afternoon”). Still, the probability of rain is often forecast as 50%, and this simply speaks to how hard this is to do.

So far, we’ve discussed climatological forecasts and computer-generated weather forecasts. There is another way to do weather forecasting, called persistence. A persistence forecast says, quite simply, that the weather is going to be doing what it’s doing right now. Needless to say, if you want to forecast the weather 10 seconds from now, or even 10 minutes from now, persistence is a good way to do it. This is because the weather generally takes longer than this to change. (Not always, of course. How about 5 minutes before sunset?)

So what forecasting method is best? It depends on the time of year and where you are, but, in general, persistence works well for very short-term forecasts, computers work well for the 1-5 day forecasts, and, after that, a climatological forecast is about the best that can be done, except for special things.

What if you use a computer weather forecasting model to try to find out if it’s going to rain at your house on this day next year? This goes back to the issue of questions that are impossible to answer. Naturally, if you turn the computer on and let it find an answer, it will give you one. But, even though it is an official computer forecast with pretty pictures and everything else, it won’t mean anything. And this gets to the heart of uncertainty.

Can you really predict this?

If you live near a little stream, there’s an experiment you can do. If you don’t, you can just think about this and probably understand it anyway.

Find a section of the stream that’s straight but that has some rocks in it that make the water swirl around. Now, find a stick and break it into little pieces, maybe 1/2" long. Carefully drop a piece of stick exactly behind one of the rocks and watch where it goes. After it’s downstream, do it again with another piece of stick, and make sure you drop it as close as you can to where you dropped the first one. Chances are that this second piece doesn’t go exactly where the first one went. And, if you keep putting sticks at exactly the same place, you are likely to see lots and lots of diferent paths that the pieces take as they float downstream.

It shouldn’t surprise you to hear that these different paths get more and more different as the stream gets faster, with more rocks and more swirls.

Now, the stream is a fluid, and the swirls are turbulence, so this system is what scientists call a turbulent fluid. And it is a fundamental property of turbulent fluids that they are not completely predictable. As time passes, they become less and less predictable.

The atmosphere, and the oceans, are also turbulent fluids, and so they are also not completely predictable. This is why the computer model can’t give a meaningful answer to the question of whether it’s going to rain at your house on this day next year. And this is why weather forecasts will always be uncertain.

But what about climate forecasts? The climatological forecasting technique mentioned above works (pretty well) for the climate we have now, if you do it with probabilities. But what if the climate changes?

First, it’s important to understand that climate models are based on weather forecasting models, but they’re more complex because they need to include the oceans and other parts of the Earth system that are relevant to climate. And, if the climate model is any good, it will predict the climate, both the present climate and changed ones. It can’t predict specific weather events (like rain at your house), but it can predict the overall behavior of the climate and its statistics.

Not, however, with complete certainty. Quantifying uncertainty—that is, putting numbers to it in a meaningful way—in climate forecasts is one of the big challenges facing climate scientists today.

Back to Environment index