Climate Impacts Research Consortium

  Thinking Deeper:

Confidence Levels & Precipitation Projections:

As with weather forecasts, climate projections are all about calculating likely outcomes. When we say an outcome is more likely to happen (or more likely to not happen, as the case may be), we say we have a higher degree of confidence about the outcome. Outcomes that are calculated to be more likely are said to have a higher degree of confidence.

We bring this all up, not because we want to send you down a rabbit hole of unnecessary nuance, but because nothing (or nearly nothing) has caused us more dismay over the years than trying to explain to a general audience what our findings say about precipitation for the Pacific Northwest under climate change. So here goes: unlike temperature projections for our region, precipitation projections for the Pacific Northwest do not have a high degree of confidence to them. Let’s break this down.

There are at least three reasons for this lack of confidence about what will happen to precipitation under anthropogenic climate change. One reason is that the effect of greenhouse gas emissions and other human activities are not as direct on precipitation as on temperature. A second reason is that precipitation is simply a more complex process to model with a computer than temperature. A third reason—and the one that will always exist no matter how good our models become—is that there is a lot of natural variability when it comes to precipitation, especially when compared to the effect that increasing greenhouse gases have on how much it rains. To picture this, consider the old statistical analogy of the so-called fair die.

With a fair die—that great arbiter of numerical justice—any given roll is as likely to come up as, for example, a six as it is to come up any other number between one and six. Roll the die once, and you might get a six. Roll it again, and you might get a one, or a two, or a three, or a four, or a five, or even a six again. But roll it enough times and any number between one and six should start showing up about as often as any other number between one and six. The point is, you shouldn't see any unexpected pattern in the numbers. Okay, you get the idea.

Think of these die rolls as computer model runs, where every roll equals one model run. Now imagine an unfair die, one that, let’s say, is weighted so the die is much heavier toward the six. Think of temperature as this unfair, or weighted, die. This is a pretty good analogy for a climate model where greenhouse gas concentrations have been cranked up past natural variability; a model, in other words, where anthropogenic warming is at work. Our cranked-up die roll is so heavily weighted toward sixes, or temperature increases, you really don’t need to roll the die very often—you don’t need to run the model very often—before you start to see a pattern. All die numbers are no longer equally likely. Instead, sixes rule the day, temperature rises, and you feel confident that if you were to roll the die again—run the model again—that there would be a high probability that six would be the result. The die of precipitation is another story.

Due to its higher degree of complexity, our precipitation die is more subtly weighted. It’s not a perfectly fair die with equally likely outcomes, but it’s not a perfectly unfair die either. This means it simply takes more rolls, or model runs, to discern a pattern. Run the model once and you might get one outcome, run it again, and you might get a slightly different outcome. That’s the analogy. Here’s the reality.

The reality is the Global Climate Models (GCMs) that climate researchers use aren’t always run often enough to satisfy our exacting search for a definitive pattern. And while this is acceptable, if not ideal, for temperature projections because GCMs produce similar results for temperature under climate change, precipitation is another story. There is a lot of disagreement among models concerning precipitation And because the models aren’t always run multiple times, discerning the subtle weighting in the models is a lot harder. (For more info on why we like to have models run multiple times and in batches called ensembles, see Models: Climate, Hydrologic, Vegetation and More.)

For these reasons, our confidence in precipitation outcomes is lower than our confidence in temperature outcomes. Okay, that’s a pretty long preamble, but keep die rolls and all that in mind when you read our results below.

Under climate change, yearly precipitation in the Pacific Northwest isn’t expected to deviate significantly from its current pattern, according to our analysis. Only 60 percent of the scenarios we used in our assessment—that is models + emissions scenarios, 40 in total—showed a tendency toward an increase in annual precipitation under climate change by the end of this century. Contrast this to the temperature projections we came up with: all 40 scenarios we used showed temperature increasing for our region in this century.

What’s more, the average increase in annual precipitation we observed in our data was only about 5 percent for the scenarios we employed. We also observed a similar nudging of precipitation outside of natural variability for summers and winters. This means our already dry summers might nudge slightly drier in our climate changed-future, while our already wet winters might nudge slightly wetter. However, due to the high complexity/low confidence of these scenarios, as scientists we feel it’s important to again point out that these results should not be taken out of context. We’re taking about projections and not predictions. Probabilities are not prophecies.