I’m in thesis mode. My thesis is officially due Friday and it’s coming down to the wire, so I’ve been working on it pretty much around the clock.
My life lately has consisted of a lot of frozen dinners and 16-hour days at the office. I’m very much looking forward to getting back to real food and balanced living as of this weekend.
The other day I sat down and wrote a lengthy explanation of what my thesis is about. Since I don’t often do that for non-scientists, I thought I’d share a revised version of that explanation here. As always, questions are welcome.
The title of my thesis is:
Effect of modeled pre-industrial Greenland Ice Sheet surface mass balance bias
on uncertainty in sea level rise projections in 2100
And what that means is:
Basically, I model ice. Big ice. Namely, Greenland’s ice. (At the moment. Antarctica’s ice is coming for my PhD.)
For my master’s work, I use a coupled climate model to try to figure out things like how the Greenland Ice Sheet will respond to climate change in the future. I use these results to predict how much sea level will rise during the 21st century.
But more importantly than that, I’m worried about how much uncertainty there is in predictions of future sea level rise. And by the time you get done reading this post, you’ll be worried about it, too.
For my master’s, I start out by telling my ice sheet model what I think the net precipitation (“surface mass balance”) over the ice sheet should be. Any ice sheet model needs to know the net precipitation in order to figure out how much ice Greenland is gaining or losing (is it snowing a lot? is the ice melting?). These changes directly correspond to changes in sea level. I expect that errors in what I think the net precipitation should be will affect how much uncertainty there is in my model’s predictions of sea level rise between now and 2100. We (scientists) make a good effort at figuring out what the correct precipitation is, but we get this input wrong every time.
But it’s not because we’re bad scientists.
Often, the perceived goal of climate modeling is to perfectly reconstruct observations. The idea is that by doing so we can in some way validate that our model is “right”. I showed this approach in the two Greenland surface velocity plots I posted last week. I show there that the model matches observations really well. This suggests the model has a lot of promise, but we can’t actually say for sure that our predictions of future climate are going to be great just because we got those two plots to match.
The fact of the matter is that no model will ever be able to capture the full complexity of nature, so it will always be in some sense “wrong”. We would never expect our model results to perfectly match observations because the model is inherently imperfect. We just don’t know everything about how the Earth works and I suspect we never will. There’s a famous quote from a statistician named George Box that gets at this problem: “All models are wrong, but some are useful.”
But the other side of the problem that people often forget is that observations are also uncertain in a variety of ways. Sometimes it can even be hard to track down uncertainty in observations collected by others because it’s ignored, difficult to calculate, or assumed to be small. In real life, though, observations are also imperfect. Because of this, it’s possible for our model to be less “wrong” than it might appear at first glance, since the baseline (observations) isn’t actually a very good metric for comparison. If the observations are wrong, matching them would just ensure that our model is wrong, too.
So to really get a good grasp on climate predictions, it’s important to consider all the sources of uncertainty in the system. This is an enormous undertaking, so scientists break it down and tackle one thing at a time. For me, that’s the surface mass balance in Greenland.
The goal of my master’s work is to see if it even matters to predictions that we’re wrong. If our results are the same no matter how much I get it wrong, then I won’t expend too many brain cells trying to get it right. But if it turns out that it means the difference between a sea level rise uncertainty of 1 cm vs. 1 meter, then I care a whole lot more about improving my knowledge about this seemingly small issue of net precipitation over Greenland. It’s much easier to plan for the future when you know how much sea level is going to rise to within 1 cm than 1 m.
Current projections peg 21st century sea level rise at between roughly 0.5 meter and 2 meters. To put that in perspective, about 150 million people worldwide live within 1 meter of high-tide sea level. If the predictions are right, that means more than 150 million people could be displaced in the next 85 years. That’s nearly half the population of the United States and a full 2% of the world’s population. If we’re confident that sea level is going to rise 1 meter, it might be good idea to start moving them sooner rather than later.
But what if we’re underestimating uncertainty? What if we haven’t taken all the little uncertainties like this precipitation thing into account and really the projection should be 2 meters give or take 2 meters? In that (extreme) case, maybe sea level won’t rise at all. But maybe sea level will rise 4 meters. Those are two very different scenarios. To put that one in perspective, 250 million people live within 5 meters of sea level. Now we’re talking more like forcing the entire population of Brazil out of their homes and off their land.
No one wants to relocate 150+ million people if they don’t have to. But the ocean and the ice sheets doesn’t care what anyone wants. Beyond taking drastic steps to mitigate climate change, something we can’t figure out how to agree to do, all we can do is make the most reliable predictions possible and prepare ourselves for what may happen. If those people end up having to move and no one is prepared for it… well, that’s the sort of thing that crashes economies and starts wars.
That’s why uncertainty is so important.