 |
| Camano Island on December 27th, 2022. Photo courtesy of Joan Schrammeck |
The coastal flooding event in Puget Sound on December 27th, 2022 was extreme, and notable for a whole variety of reasons that I explored in this post shortly after the event. I've also finally, after playing with the data for almost a year, convinced myself that this event had a near zero chance of occurring without the sea level rise observed in Puget Sound over the past 100 years or so. This conclusion, if it holds up to scrutiny, is also a big deal, as it could be the first documented evidence of an actual impact of observed sea level rise in Washington.
The way that I went about thinking about this was to first generate what is called a return frequency curve, using annual maximum water level observed in Seattle and recorded by the tide gauge in Seattle since its inception in 1898. I focused on Seattle first because it was impacted by flooding on 27 December 2022, and it has a nice long tide gauge record to support an analysis like this. It is worth noting that I haven't yet repeated this analysis for any of our other tide gauges in Washington, and suspect that the results wouldn't be as clear because of their shorter records. NOAA generates return frequencies too, but here is what mine looks like:
These curves are useful, as they allow us to use the history of extreme events recorded at tide gauges to analyze the expected frequency and magnitude of extreme events . On the y-axis is the maximum water level elevation, in feet relative to Seattle's Mean Higher High Water, and on the x-axis is years, from 1 to 10,000 years (the scale is logarithmic). We can use statistical models to fit a curve to these data, in this case I used a
Generalized Extreme Value (GEV) model:
The GEV model fit to the observed annual extreme water levels from Seattle suggests that that the event on December 27th, 2022, during which coastal water level in Seattle reached an elevation of almost 4 feet above Mean Higher High Water, has an annual chance of occurrence of about 0.2%, or in other words should be expected to occur about once every 500 years on average. It was an extreme event.
But sea level is also rising in Seattle, and
we know that even small amounts of sea level rise can significantly change the likelihood of these types of extreme events. How did Seattle's history of sea level rise influence the likelihood of the extreme flooding observed on December 27, 2022? another way of putting this is, what would the likelihood of the peak water level from 27 December 2022 have been if there was no sea level rise occurring in Seattle? To try to answer this question I simply removed sea level rise from the record in Seattle (assuming it is a linear rate,
which is what NOAA calculates,
but may not be the case), and then re-calculated a return frequency curve (shown below in red):
and then repeated the process of fitting this set of hypothetical extreme events from Seattle with a new GEV model:
So here the impact of sea level rise on the return frequency curves is readily visible. But if we follow the water level of the December 27, 2022 event (the highest black dot in the plot) over to the left to find where it intersects the red curve...we see that it never intersects the red curve. This suggests that, based on the best fit to the data, the extreme water level reached during the December 27th event, at least as recorded in Seattle, would have been improbable in the absence of sea level rise.
Now these statistical models are uncertain, because they are built around a very small number of events (they are, after all, extreme), and above I show the 95% confidence intervals (the red dashed lines) around the best fit to the observations (the solid red line). Here the upper confidence bound DOES intersect with the water level elevation reached on December 27, 2022. But even if we assume that the "real" return frequency curve follows that upper confidence limit we find that the event observed on December 27th 2022, with sea level rise removed, has a return frequency of, at best, once in 700 years on average. At the very least it seems safe to conclude that sea level rise that has already occurred in Seattle made the extreme water level reached on December 27th 2022 more likely (a 1-in-700 year event becomes a 1-in-500 year event) than it would have been.
**I'm not extraordinarily well-versed with extreme value analyses, so I reached out to a colleague for a review (which I don't typically do with these blogs) and wanted to thank that reviewer for their time and contributions. I'm going to keep that person anonymous only to emphasize that any errors of interpretation are mine alone. I also want to emphasize that here I've assumed two things: 1) that sea level rise is the ONLY component of coastal flooding that is changing, which I think is likely the case but I didn't test, and that 2) the rate of sea level rise in Seattle is linear, which I think is a good assumption but could be incorrect.**
