What happens when past climate modeling experiments do not agree with the climate data in tree rings, marine sediments or ice cores?
Early on in my career, I was chatting with a colleague at the annual Fall meeting of the American Geophysical Union in San Francisco, USA. Our focus was on the role of paleo-, or past, climate work in the larger scheme of climate research.
Specifically, for us, two paleoclimate ‘modelers’ who use complex computer programs to simulate past climate events, the focus of our conversation was on a mismatch that has become quite common in our field.
That mismatch is: That even with our best efforts at correctly setting the boundary conditions (such as the size and extent of glacial ice sheets or the distribution of vegetation) for a past climate modeling experiment, the results sometimes do not agree with the climate data derived from natural archives such as tree rings, marine sediments, or ice cores.
These kind of mismatch scenarios have played out over the last eight years for simulations describing the influence of large volcanic eruptions on the climate for the last millennium – the 1000 years preceding the pre-Industrial period. Volcanic eruptions are by far the largest climate drivers in these simulations. Indeed, if the eruption of the Philippine volcano Mt. Pinatubo in 1991 is indicative, the peak impact of volcanic events on climate can, for a few years at least, be equivalent to the climate effect of human-emissions of greenhouse gases.
In preparation for these last millennium experiments, several groups looked at ice core records for evidence of volcanism. They found this evidence in the form of layers of volcanic deposits (sulfate layers). Then, considering how particles move from high up in the atmosphere in the area above weather (~15-30 km), the stratosphere, they came up with a way to infer the size of volcanic eruptions from these ice core deposits. Essentially, the thicker and more coherent the deposits across many ice core sites, the lower latitude and higher magnitude the volcanic eruption must have been.
But, some of the volcanic events implied by these ice core records were much larger than any we have observed recently (the last 200 years) – some of them in the late 13th century appeared to be 10 times the size of the 1991 Mt. Pinatubo eruption. When we tried to simulate these enormous events – we saw an equally enormous impact on climate (in the models, that is). The past climate data -- tree rings, cave deposits -- just didn’t have the same scale of event as the models were implying. By the 13th century, many societies across the globe had developed sophisticated written records – but even these did not bear out such a huge event.
Some of the questions my colleague and I discussed:
Should we simply accept the mismatch?
Should we use the match between past climate simulation and measured indication as part of our toolbox to improve the physics and fine-tuning of our models?
Do we suggest our paleoclimate colleagues have just made a mistake?
Should we reconsider the boundary conditions (that are supposed to drive the model in the right direction) that we used in the first place?
Of course there is no single answer to these questions. But I think that this situation and the discussion between the modeling and data communities around mismatches for last millennium volcanoes has actually driven the science forward and improved our understanding overall. The community looked hard at the mismatch, from three independent directions:
Are the climate reconstructions for the past failing? The tree ring community in particular has confronted this suggestion head-on. They have evaluated whether their measure for climate fails in the face of extreme conditions, rejected this hypothesis, and in the process have defined better constraints on the uncertainty in their system. Moreover, the last millennium is a very special past climate period during which some historical records do exist, as well to supplement the reconstructed climate archives. These not only validate the climate reconstructions, but also link to the human impact of large volcanic events.
Were the boundary conditions for volcanism over the last millennium faulty? Yes, and no – they certainly could be improved, and again, the community has addressed this question head-on to create new volcanic reconstructions that seem to be much more consistent with climate response. In fact, these newer past volcanic eruption data sets work better because they account for not only the volcanic deposits in ice cores, but also the climate response from past climate archives into the reconstruction.
Are the climate models that were used to simulate the last millennium capable of accurately reproducing volcanoes over this interval? Again, yes, and no. The paleoclimate modeling community has realized that we do not understand enough about how the largest volcanoes operate. Fortunately, many of us have the privilege of working in groups also studying the dynamics of volcanic plumes, or the evolution of aerosol particles, which we can use to improve our volcanic implementation. Previously, we needed observations of the volcanic particles in the atmosphere to predict the climate response to volcanoes. But, when we ‘wait’ for these observations, we are already experiencing the climate response to the volcanic event. By instead observing and forcing the model with the eruption itself, we can predict – potentially 6 months to a few years in the future – how climate will respond to a volcanic eruption.
Quantifying and highlighting the strengths and weaknesses of the simulation in comparison to the reconstructed climate series is an integral part of the process to improve the science to understand volcanic impacts on climate and society. Incidentally this is a new working group of the Future Earth core project on Past Global Changes (PAGES), dedicated to better understanding the impact of volcanic eruptions.
These impacts are discussed at length in the latest issue of the PAGES Magazine on Volcanoes and Climate and leaves us with a much better understanding of these strong drivers of climate, and a picture of where future research should be directed.
Past-climate researchers are actively seeking new high-resolution archives that can capture these short-lived volcanic events. Climate modelers likewise are working hard to improve our models to be ready to simulate the impact of all volcanic events –not only in the past, but also in the future.