What happened this past weekend along the Stilliguamish River in Washington State has been truly tragic. There have already been 24 confirmed deaths, but many more are still missing. The landslide (others will also call it mudslide or landslip) occurred when hillsides became saturated with water. In the 45 days leading up to the event, the area experienced 200% of normal rainfall.
If you’re interested in some reading material, I found the following quite helpful:
The second link was what brought my attention to the blog of Dave Petley, Professor of Hazard and Risk in the Department of Geography at Durham University, U.K.
Prof Petley has aggregated pictures from multiple sources, as well as offer further insight as to what caused the landslide, and more importantly, things to be on the lookout for in the aftermath. One image was a tweet from the National Weather Service. It shows data from the Stilliguamish River downstream from the landslide. I became a bit curious, so I took a look at the National Weather Service site. Turns out you can actually access quite a bit of data. I was able to locate the river station that collected the data in the tweet. I downloaded the data to take a look in Mathcad. I opened the data in Excel and made a copy so I could include it in an Excel component.
I then plotted the ‘stage’ column, which I interpret as depth. A comparison with the aforementioned tweet made me realize the data was being plotted backwards. Of course it was, since the most recent data was at the top.
To view the time series data in a more intuitive fashion, I had to extract the time, which surprisingly came over into Mathcad automatically. Since Excel recognizes 3:23 as a time, it converts it to a decimal corresponding to the fractional part of the day. So midnight is 0, and every six hours is 0.25 (a quarter of a day). Mathcad plots can not handle dates on the axes, so I created a new timestamp where the fractional part is the time we obtained from the Excel component and the integer part was just 321, 322, 323 (corresponding to the dates 3/21, 3/22, 3/23).
There is no new revelation here, but we can now create more custom views. We can have each measurement be represented with a symbol. And we can plot a vertical line at the time of the landslide. Keep in mind that the data here is in UTC, and the 10:47a was local time, which on 3/23 was in the Pacific Daylight Time (PDT) zone, hence the +7 hour difference.
As people have already noticed, there is actually a peak just after the landslide. This is caused by the debris entering the river and creating a surge of water. We then see a sudden drop-off as the debris has blocked off the river. Fortunately, we see that the river depth rose again just before 3/24 (or about 5pm PDT on 3/23). This is confirmed with information from the Wiki page, which stated the river started flowing through the debris on 3/23 and is now also flowing north of the debris as well.
Of course, all this is after-the-fact. What we need to now do it figure out ways to predict these events and also determine mechanisms that can help us reduce the damages. There is already much research done in these areas. Richard Iverson, an USGS hydrologist, wrote a paper modeling rain infiltration with topographic data and rainfall intensity/duration information. Retaining walls and drainage wells are just some of the ways scientists and engineers have come up with to help reduce the effects of a landslide.
Download a copy of the worksheet used to generate the above images.