Great observation from Portugal

I wanted to share a great observation made a few months ago by a Loss of the Night project participant from Portugal:

As you can see, there is a very clear separation (white line) between the stars that were visible (yellow, orange, and brown) and those that were not (black). You can also see how thanks to the observer examining so many stars, you can have a lot of confidence in the result, and can even measure how consistent the observer's result is. When you first start using the app, you might not have results as consistent as this, but with a bit of practice it becomes easier and easier to cover a lot of stars quickly.

This also demonstrates why the app is a better method for estimating limiting magnitude in bright places than star chart based methods like Globe at Night. In this case, the naked eye limiting magnitude was around 3.9 ± 0.1. The star charts of Globe at Night only allow you to choose between integer limiting magnitudes, which in this case would be "about 4". In addition, with Globe at Night we can't be sure how careful a participant is, and we found that compared to skyglow models, the standard deviation of Globe at Night observations is about 1.2 magnitudes.

While the app can provide more accurate data, I want to stress that it's not a replacement for Globe at Night! The app doesn't include stars with limiting magnitudes above about 5.2, so in areas with little light pollution, Globe at Night is a better method. In addition, the Globe at Night time series goes back over 10 years, and there is therefore a lot of value in continuing to contribute to it. In my opinion, it's the best system we have for tracking global changes in skyglow. So please consider contributing to both projects!

If you make an observation with the app, you can easily see a similar plot for your own results. Just head to My Sky at Night, zoom in to the area where you made your observation, and click on it to bring up this chart.

Good Morning Twitter!

Together with Tatjana Scheffler, I recently published a paper where we looked at what time German Twitter users wake up, and how it varies throughout the year. Here's a really short explanation.

What we did

Dr. Scheffler collected and saved (nearly) all the German language tweets for an entire year. She then selected all the tweets that included the German phrase for "good morning" (Guten Morgen). We then analyzed the data to find what time each morning the phrase really starts to take off, and called that the "onset of twitter activity" (i.e. what time people woke up).

What we found

The data can be best explained visually. Here is a plot of what time the sun rises (in Frankfurt) throughout the year. Winter is near the middle, and the "fall back" and "spring ahead" of Daylight Savings Time are shown as dashed lines. We don't change the clocks on this plot, to make it easier to see how wake times relate to the sun.

Now here's what wake times look like on weekdays:

Throughout the year, the typical* wake time on a weekday is around 5:50 am (in local time). You can see that there are a few weekdays that look very different from typical. Those are holidays, and the bunch in the middle are the days between Christmas and New Years.

Now here's what it looks like on Saturdays:

And here's Sundays:

You can see that during the late fall winter, and early spring, the wake up times on weekends are closely related to the time that the sun comes up. But then Daylight Savings Time comes and the relationships break down. Here's what it looks like with all the data together:

The gap between the blue compared to the black and red lines shows you how badly people are punishing themselves by using an alarm clock. About 80% of people in Europe use alarm clocks, which means that they aren't getting a full night's sleep. But what can we do about it?

Take a look at the difference between the blue and red lines. You can see that at the end of March, they're almost reaching each other. If they were to touch, it would mean that millions of people would manage to wake up fully rested without needing an alarm clock.

The plot shows that Daylight Savings Time lasts too long. If it started later in the spring and ended earlier in the summer (or if it was eliminated altogether), millions more people would get a good nights sleep. We would be healthier, feel better, be less likely to be involved in car crashes, and our entire society would be more productive. High school and university students would benefit in particular, because they need to sleep in the latest. Schools would have fewer disciplinary problems, students would fall asleep less often in class, and because they are alert, they would learn more.

If you'd like to read our full paper, it's freely available online. If you'd like to have a laugh and hear more about how terrible Daylight Savings Time is, here's John Oliver asking how it's still a thing:

* This might seem a bit early, and it is. It's related to the method we chose to select a single time. Long story short, some people wake up earlier than this, most people wake up later, but the time shown here is the most stable measure for start of twitter activity.

Model of exponential growth in skyglow at one location

Andrej Mohar produced a video modeling skyglow changes from 1950-2040 at Matajur mountain on the border of Slovenia and Italy:

The simulation is based on a calibrated all-sky image taken at the site, with an estimate of 5% annual growth in skyglow (which is a pretty reasonable assumption for the later half of the 20th century). But there's no reason that skyglow must continue growing at such a rate! If we get smart about how, where, and when we light, we could recover the starry skies that many of us remember from our childhood.

Which stars are people observing?

The Loss of the Night app works by having people report whether they can see stars or not - but which stars are they looking for? Here are all of the stars that had been searched for up to August 22:

The big dipper and cassiopia are big and colored red, to help your orientation (it's an equal area projection, so it looks kind of weird compared to what you see in the sky). The plot is not filled in evenly, because the app has only been around since April (more about this in a moment).

Our app only asks users to look for the 1601 bright stars that were already part of Google's Sky Map app. But it doesn't ask people to look for each star with the same probability. It always starts with a bright star, and the subsequent selected stars depend on what you can see. Because there are so many more faint stars, some stars are searched for much more frequently than others. Here's a plot showing how often each of the stars has been searched for:

Black dots are 1-4 times, red 5-19 times, blue 20-100 times, and purple over 100 times. The most commonly searched for stars so far are Arcturus (602 times), Altair (327), Vega (322), Alioth (260), and Dubhe (184). The most commonly searched for stars are all in the upper half of the image above, because most of the app users live in the Northern Hemisphere.

You might also have noticed the big gap of missing stars on the right. That's there because the stars in the sky change throughout the year, as the night side of Earth points in a different direction as the Earth orbits the Sun. Each month when the moon goes down, users will be confronted with a new set of stars, in new positions, and with new magnitudes (brightnesses).

Here are the stars that were searched for at least once during April/May:

And during June: 

During July:

And up until August 22:

You can see that the searched for stars are marching slowly but surely "left", as the year advances. Very soon, Orion will be a risin', and hopefully many users will tell us whether or not they can still see his stars a blazin' from the middle of Central Park!

Our partners from Cosalux who programmed the app have produced a short video showing where the app had been used (up to early August), check it out:

We're tremendously happy that it's been used on all of the (inhabited) continents, and we hope that we can continue to make inroads in areas with rapid development. Sometime this fall we hope to release a new version with extended language support, including Chinese, French, Japanese, Spanish, Catalan, Polish, Romanian, Italian, and Czech. As I've mentioned before, we're looking for translators for Russian and Arabic, but we would need the translation done in the next days. If you could do the translation, or know someone who might be able to, please send me an email!

[Note: updated with higher resolution version of the video on Aug 31, because blogger reduced the resolution.]