BY ELEANOR TERNER
image source
Alexander Weyant is a first year phD student of climate science at Scripps Institute of Oceanography who, as an undergraduate at UCSD, participated in a research project analyzing the storm leading up to the Oroville Dam crisis. In this interview, he discusses the research models and simulations that his group used in their recently published paper, their findings, and the challenges of using ‘causation’ to tie together climate change and extreme weather events in research.
What are the main questions your research regarding the storm preceding the crisis focuses on answering?
We wanted to know whether we can say, or to what extent, climate change might have intensified the atmospheric river (AR) of early February 2017. Strictly speaking, we learned a lot about a hypothetical version of the historically-observed event. Since we will never again see the same exact conditions preceding the event it does not provide us with generalizable knowledge in the forecasting sense. There have been previous studies that have taken a broader view at the effect of climate change on precipitation. From these, we know what climate models tell us about behaviour in the aggregate, but we are not aware of very much similar research at the event scale.
Is there a difference between climate model and climate simulation?
A climate model is a set of equations (physical and empirical) which are run over time. They come in all levels of complexity and realism, though “realism” is a concept that depends upon the question being asked. The simplest models grant some intuition about bulk values(such as total outgoing infrared radiation or global mean surface temperature). The simplest “toy” models can be solved explicitly with pencil and paper, and can conceptually be used to explain phenomena such as the greenhouse effect. Other climate models are complicated behemoths which sometimes share their code with weather prediction models. These complicated models output many variables on spatio-temporal grids.
In this study, we used a Model for Prediction Across Scales (MPAS) and, in our particular application, it was similar to a weather forecasting model. This model would be considered overkill for a typical weather forecast. To initialize our weather prediction, we used a historical reconstruction of the initial conditions from reanalysis (a combination of observations and model output) as well as a background climate projection from a long-running climate model.
What does your climate model tell us about the response of precipitation to temperature?
The MPAS simulations showed us that, in the few days we let it run, the circulation pattern was only slightly dependent on temperature, and the changes in precipitation were mostly determined by a direct thermodynamic response. Simply, the “capacity” of air to hold water vapor increases with temperature. Under a warmer climate, more vapor can be transported by the same exact wind pattern. If this vapor condenses and ultimately falls to the surface (as opposed to merely becoming a cloud and reevaporating or falling and reevaporating), we observe a wetter storm. After looking at the output of our model runs, we think this rather simplistic explanation actually applies to the 2017 storm. Given that we have not yet observed a trend in the record of California precipitation, we are not sure what the effect of climate change would be on this storm.
Do you define “causation” in your paper?
We never defined causation in our paper, although it was something we talked about and struggled with a lot. If one defined it intuitively, then it would be ‘this outcome would not have happened without climate change.’ This was the counterfactual way we thought about it, which inspired the running simulation of the storm under altered initial conditions. Of course, if it were not for the exact climate change we have actually observed, the circulation pattern that “set up” the storm would have not been precisely observed, and the outcome would not have been the same. If anything at all in initial conditions were to change, then the outcomes would diverge. This is the long way of saying that the weather is chaotic. Think of a pendulum constructed from a jointed arm. The action of these pendulums is determined by simple physics equations. The initial jolt you impose the gravity, and then the inertia that is counteracted by the gravity are the main factors you would consider. The same jolt could be applied to two “identical systems,” but after a while the systems would look unrelated. Imperceptible differences in initial conditions, a tiny forcing along the way, or a rounding error (in the case of a simulation) could all bring about this sort of behavior. Even if there are simple equations iterating the system forward in time, we cannot say that the pendulum is in a particular configuration because of the jolt applied. Chaotic behavior can even arise in very simple systems, even independently of varying initial conditions, as this very fun video shows.
Defining causality in a system like this is fraught. We can't say with certainty that climate change caused any particular event, but you can look and see a trend over many events. Complicated models and simple reasoning about Earth’s radiation budget can suggest precipitation trends in the future, but not very much about events. Given that we are rather certain about many aggregate trends, it might be surprising to learn how difficult it is to unfold bulk trends and consider their implications for individual days, or hours. Even if you could define causality in a chaotic system, I am not entirely sure it would solve this problem. With this in mind, we need not be certain that climate affects particular events to know that it must be affecting events generally.
Were your results predictable?
The storm, as modeled, seemed consistent with the simplistic warmer-wetter pattern. There are many dynamical ways in which the storm could have been affected. The first suspects would be a promotion or suppression of vertical motion, an alternate steering of the moisture plume, or merely a change in the plume’s orientation relative to the topography. The hopefully realistic (and computationally expensive) model was run precisely to determine if these dynamical changes would strongly affect the storm within a few days. After looking at the model runs, it seems a bit surprising that the temperature response was so closely related to temperature, rather than displaying much more complicated behavior. I think a much more complicated result was anticipated from the beginning, and it is possible that we were wrong due to the particular model we chose, or a hole in our analysis we did not consider. If so, then there is more interesting work to do.
What concepts do you want others to focus on in regards to climate research?
Current students better stay out of my area–just kidding. Uncertainties with respect to clouds are very interesting. We generally don’t know what the overall cloud feedback will be [in response to climate change]. Different types of clouds have different radiative effects, so they can either dampen or amplify temperature changes. Clouds are not directly modeled by the complicated, gridded models I mentioned before, so right now it’s mostly pen-and-paper models, statistical notions, and constraints based on earth’s radiation budget that I base my limited understanding on. The idea that we are so uncertain about such an important part of our climate system is quite unsettling to me.
Do you have any advice for undergraduates who want to pursue climate research?
Talk to people at SIO (Scripps Institution of Oceanography), talk to people doing this type of work now. Don’t wait until your graduate school applications to apply to unknown places and work with unknown people. I don't know what I would be doing if I went in that direction. It’s a big leap to move to a different place or field. Your PI is more than a boss, they hold more over you than almost anyone can. Connect with people now. Start research yesterday! If you look up URS, undergraduate research scholarships, then you can apply to whatever application is open to you. That’s how I got started working one summer and that was instrumental in determining where I am today!
If you’d like to read more about Alexander's paper, it is published here.
Comments