Often when late afternoon rolls around in the office here, that standard mental slump sets in, signaling it's time for the trip up to the local coffee shop with a colleague. Usually I'll realize I've been staring mindlessly at the wall and the shadows on it, and that means it's time to go.

A number of years back, I noticed that those shadows on my office wall flash for a short time at the end of the day, because they're from the pre-sunset sun shining between the cars on the I-5 bridge. It's pretty cool actually - the edges of these shadows are remarkably focused and you can clearly see the outlines of the cars and trucks as they whiz by through the rectangle of light on my wall, making the room flash between light and dark. This event only happens for a short window of time, and at a slightly different time each day, since it's essentially a proxy for the sunset time.

Well, when a scientist is mentally twiddling his thumbs and staring at moving shadows, it's not too out of the question that he's just going to start mindlessly start jotting down the times the event begins each time. And maybe plot the times as a function of date during the year and compare them year to year. And do this for, um, you know, like six years now...

So here's the plot below. Not surprisingly it looks an awful lot like the plot of sunset time over the year. We can see the earliest sunset time sometime around December 10th in the plot there pretty consistently over the years (whew, good thing!). Do note the date of earliest sunset is not the same date as the Winter Solstice which is on December 21st or 22nd. They're related but different, and the earliest sunset date is latitude dependent (it's Seattle here). I recently saw a great blurb explaining this on a Q&A page at NASA's Goddard Space Flight Center.

Finally I'll mention the name of this favorite coffee shop in Seattle's U-district that we often head up to around this time of the afternoon. Would you believe the name is Cafe Solstice!
 

plot of flashing-event-time vs day of year