When the climate doesn’t behave like we expect, whether it’s for an individual season or for several decades, we often hear scientists blaming internal variability. Scientists use this term a lot (even on Twitter) and I’ve noticed that I usually obtain a few blank faces depending on the audience. I also remember being a junior scientist in this field and wondering why everyone was going on about internal, or its counterpart, external variability. Internal/External what? And who cares? Me! And you should, too! 

In our climate and weather there are:

(1) The things that are pushed around by other (external) things
(2) The things that would change or move (internally) without any push

Lions and tigers and variability

You, yourself, have your own internal variability when it comes to your behavior! But, at times, there may also be an external forcing that causes you to deviate from what you’d otherwise do. For example, I really enjoy taking walks in the woods and try to do so whenever I can. Because there are lots of trail options where I’m walking, my path will change from day to day based on pure randomness or a need for variety. But I also really don’t like running into bears (especially fat bears). As much as I’d like to pretend they don’t exist, if I see a bear, I will strongly deviate from my intended path and choose one that gives the bear a wide berth. So bears are an external forcing on my walking path. 

Cartoon by Anna Eshelman

Bears aside, why would we care whether variability is internal or external? Well, in the climate system, we might care a lot if we want to answer questions like “Is this rainstorm caused by El Niño?” Or “Did human-caused climate change cause the polar vortex to break down?” 

In order to figure out the answers, we first have to examine the likelihood that the impact would have occurred without any push from an external force. To phrase it another way, we need to determine whether the rainstorm may have occurred without any influence from El Niño. Or whether that change in the polar vortex would have occurred without increasing greenhouse gases. Internal variability are changes that would have happened anyways, regardless of the presence of something else (footnote #1). There will always be some day-to-day variations in my walks even if every bear instantly disappeared. 

For a scientist, it can be difficult to prove whether a weather or climate event occurred due to some external influence, like increasing greenhouse gases. This is because the observed weather—our reality—only occurs once! We can’t run an alternate reality where we remove the external influence because our observations are already history. 

This is where a reliable climate model that simulates realistic weather and climate comes in handy. In model world, we can run an experiment that does not have the external forcing—for example, an atmosphere with no increases in greenhouse gases—and a second experiment that DOES have the external influence of greenhouse gases. The difference between the results is considered the part that is externally influenced by greenhouse gases. Going back to my walks, we can compare my path through the woods in a world with bears to my path in a world without bears. 

Of bears and butterflies

But there’s a catch (there’s always a catch, darn it): the butterfly effect. Just like my different paths through the woods on different days, climate model simulations will evolve differently from each other based on small differences in their starting state. This is true in a world with bears (excess greenhouse gases) and without them. That’s why scientists prefer model studies that run “large ensembles.” These generate dozens—sometimes hundreds—of my simulated walks in the woods with bears versus dozens of simulated walks without bears. 

Cartoon by Anna Eshelman

Running a bunch of simulations results in a range of possible outcomes with bears (my random variability plus external forcing) and a range of outcomes without bears (only my random variability; no external forcing). We can compare these two ranges to get an idea of how much the odds of my following a given path (climate outcome) have changed. To build even more confidence, it is ideal to compare large ensembles among several different models.

Dr. Clara Deser, a senior scientist at NCAR, has been at the forefront of large ensemble studies, and she recently wrote a commentary on internal climate variability, which you should check out. In that piece, she provided an example of how external and internal variability can influence the trends in winter precipitation across the U.S. that we may experience over the next 50 years. 

Expectations for winter precipitation trends over 2010-2060. Top map is showing the “What We Expect” pattern or the change in winter precipitation due to anthropogenic climate change, which is based on the ensemble average of the 40-member, Large Ensemble Community Earth System Model version 1 (CESM1) under the RCP8.5 radiative forcing (Kay et al., 2015). The bottom two panels show a range of “What We Get” patterns, which is the combined influence of external variability (anthropogenic climate change) and and internal (or, in this instance, natural) climate variability. These two maps represent the 5th to 95th percentile range of possible outcomes, which is based on the ensemble member with the 2nd wettest trend over the U.S. (left panel) and the ensemble member with the 2nd driest trend over the U.S. (right panel). Figure courtesy of Clara Deser and Adam Phillips (NCAR) and modified by climate.gov. 

The top panel here shows what we’d expect if we averaged together the results of this particular climate model in order to identify the influence of human-caused climate change (an external forcing). It projects a much wetter future over the U.S. in response to climate change, especially over the eastern and western U.S. But the bottom panel shows two equally plausible outcomes drawn from the model ensemble with the exact same external forcing (footnote # 2). 

Clearly, the two maps are very different—the bottom right panel showing a considerably drier winter over the U.S. and the one on the bottom left indicating a wetter winter.  How can that be? Even though the human influence on the climate is exactly the same in all simulations within the model ensemble, internal variability is large enough to create a range of outcomes that can be rather distinct (footnote #3). The internal part is largely unpredictable—there is a certain amount of variability that is baked into the cake and will occur regardless of global warming. 

When El Niño is the bear

Another reason internal versus external variability matters is because it helps us understand what we can or cannot predict (footnote #4). From day to day, the exact path I take for my walk in the woods is mostly unpredictable—there’s randomness to it. As much as a bear is scary to see, it imparts some predictability on the walk because I will go well out of my way, around the bear, to avoid it. The predictable part is looping around the bear, not walking right up to the bear and asking to be eaten. So, it helps to have external variability in the weather and climate system—without it, it would be difficult to predict at all. Maybe we should be thankful for some fat bears after all. 

---
Side note: Clara and I were chatting about internal variability and how it may be time to come up with a less obscure term. She came up with inherent variability, which seems to better convey variability that is inherent to the climate system. So, blog readers, what do you think? Is it time to phase out “internal variability?” What do you think of “inherent variability” instead? Leave your thoughts in the comments!

Personal Note: Geert Jan van Oldenborgh died earlier this month and with that climate science, and even more than that, climate services, has lost a great scientist, a pioneer, and a truly good person. He openly struggled with cancer even while pushing us forward. While I did not know him well, I was lucky enough to have been touched by his insights and passion in our collaboration on a relative SST index for ENSO monitoring. I think the best way we can uphold his legacy is to apply his level of enthusiasm to our work, be curious, be humble, and remember that science is ultimately about serving others so that we can be our best selves on this fragile planet. RIP Geert Jan. 

Footnotes: 

(1) Instead of “internal variability” you may sometimes hear “natural variability,” but this term might be a little confusing b/c, depending on context, both internal and external fluctuations could have some “natural origens.” For example, a volcanic eruption is a natural occurrence, but the emissions from the eruption are an external forcing on the climate system, with possible effects on ENSO. 

(2) Primarily increasing greenhouse gases plus some other factors included in the RCP8.5 scenario of projected radiative changes.

(3) Model ensembles aren’t the only way to estimate internal variability. The two references below show that you can "scramble" the observed precipitation data to mimic a large ensemble, and obtain a similar spread of 50-year trends as those in the CESM1 model.

McKinnon, K. A and C. Deser, 2018: Internal variability and regional climate trends in an Observational Large Ensemble. J. Climate, 31, 6783–6802, doi:10.1175/JCLI-D-17-0901.1.

McKinnon, K. A. and C. Deser, 2021: The inherent uncertainty of precipitation variability, trends, and extremes due to internal variability, with implications for Western US water resources. J. Climate.

(4) As you can imagine, what is considered internal or external will change depending on the context and the particular question asked. For example, El Niño can be the “external thing” that is pushing rainfall around--- how much of this storm due to El Niño?  But different questions can be asked, like whether El Niño events are getting stronger due to increasing greenhouse gases. In this second example, greenhouse gases are now in the external forcing role and El Niño is the internal variability. As Tom pointed out, in the latest IPCC report scientists noted that El Niño has tremendous swings and variations even without changes in the amount of greenhouse gases. El Niño’s large internal variability and consequent lack of consistent changes in various model projections are one of the main challenges in determining whether greenhouse gas increases are changing its amplitude or frequency.