Skip to content

One of the major questions in climate science is what changes are natural, what changes are forced by increasing levels of greenhouse gasses, and how do they interplay? In a recent paper focusing on changes in northern hemisphere oceanic temperatures Steinman et al (2015) say that knowing the difference between forced and unforced is essential. They look at the North Atlantic (NA) and the North Pacific saying that they are “key drivers” of sea surface temperature (SST), although they add that there is “substantial uncertainty” in their relative contributions to the observed SST variability. To separate natural and forced variations they use a “new semi-empirical method that uses a combination of observational data and a large ensemble of coupled climate model simulations to asses the relative roles of both forced and internal variability in the Northern Hemisphere over the historical period.” In other words, they compare climate models to observations.

It’s been done before. Some researchers have simply removed the trend from the NA SST and considered the residuals as being due to natural oceanic cycles of temperature. Some have compared the NA to global SST. Others have looked at the difference between observations and the average of lots of climate models. This is the approach taken this time, and it’s not without its problems.

This is how it goes. You take a climate model that has the response to increasing levels of greenhouse gas hard-wired into it. It also includes estimates of some the natural climatic variations that you know of, though not all. There are obviously some modes of internal climate variability that you don’t know about. Because of this and other factors you know the model you are using has severe limitations. You then run the model lots of times using slightly differing parameters and starting conditions that you believe represent some aspects of natural variability.


The North Pacific: Poorly understood but blamed by some for the “pause.”

But there is a problem; no model run reproduces the observations. Indeed when it comes to the North Atlantic for example climate models display no competence in reproducing the multidecadal SST. Clearly the model is wrong either in its assumptions of greenhouse gas forcing or in the way it tackled natural internal climatic variability. You put down the differences between model and observations to extra internal climate variability you haven’t taken account of. Some recent work suggests that the disparity between models and observations might be due to errors in the forcing estimates for the accumulated effect of small volcanic eruptions.

You then take the view that internal climatic variability in the models is uncorrelated among the individual runs when considered as part of a large ensemble. Thus you maintain that internal natural climatic variability is eliminated if you take the average of them all. The mean is therefore the “true” climatic response to greenhouse gas forcing.

So you take that mean and superimpose it on the observations. There are obviously differences. These are interpreted as being due to internal climatic variability. The researchers take the difference and filter out the short-term variations and conclude that what is left is decadal internal climatic variability. Because you have defined it to be the sum of the “true” forcing climate component and the newly extracted decadal variations as being equal to the observations you feel an internal confidence that it all hangs together.

Along comes an even more difficult question: what does your analysis predict for the future? Steinman et al (2015) say that internal climatic variability in the NH oceans, in particular the Pacific, may have offset the recent expected warming and they predict a rapid rise in the future. Since in their paper they only go up to 2005 it would be interesting to see how a “prediction” of the past decade would fit into their scheme.