If We Can't Predict the Weather, How Can We Predict the Climate?
While weather forecasts are only skillful up to about a week, climate models can provide valuable information for hundreds of years from now.
How is this possible? Think about the differences between planning a weekend camping trip and growing a garden.
If you are planning a camping trip, you may need to know something about the weather. What will the temperature be like on that weekend? Will it rain while you are there? Will it be sunny? A weather forecast model can be used for this information, but we need to wait until just days before the trip since such models are not reliable for periods beyond this.
If you decide to start gardening, you still need think about weather, but in a different way. How long will your growing season be? On average, how much rain will you get? Will it always be sunny or will there be long stretches of overcast? A climate model can be used for this information, and if needed, we could even plan a garden several years into the future.
The only difference in the information needed for these scenarios is how it is quantified. In the first example, we want to know when and where a weather event will happen. In the second, we want to know the long-term statistics of the weather, or climatology. Climate models do not tell us what the weather will look like on a particular day, but instead that days will be warmer than nights, that Florida will be warmer than Vermont, and that it will rain more in Seattle than in Las Vegas.
Just like climate models tell us that more sunlight results in warmer temperatures in the summer, they tell us that more greenhouse gases will result in warmer temperatures, more extreme rainfall, more droughts, and more severe weather under climate change. Climate models do not tell us exactly when and where these events will occur for future camping trips, but instead what the long-term averages, trends, and extremes are for planting a garden.
So how do we know if climate models are correct?
We can verify the accuracy of models by running hindcasts. These are just like forecasts, but the model provides information about weather and climate that has already happened. If we want to examine the accuracy of a weather forecast, we can compare actual weather events to those that were forecast by our model. Likewise, if we want to examine the accuracy of a climate forecast, we can change the start date of the simulation to an earlier time, say 1915. Once it creates a 100-year forecast, we can compare the output to actual climate values at the present time. As it turns out, these models do a very good job of estimating long-term values such as temperature.
Importantly, when climate models are initialized hundreds of years ago without accounting for increasing greenhouse gas concentrations (as have been observed over this period), they forecast temperatures that are much cooler than they have actually been. It is only when we include the additional greenhouse gases that we see temperatures comparable to what has actually happened. This provides evidence of model accuracy while also supporting the science behind global warming due to increasing greenhouse gases.