What prevents us from seeing the stars?

Last night I was in Brussels, Belgium, and as the sky was clear I had the chance to look for stars at an open area just outside of the Brussels Central Station. When I first looked up, it seemed as if there weren't any stars at all, and I think this is the impression that most people would have. But when you look at the dark sky for a bit longer, and screen out some of the glaring light using your hands, it turns out you can see quite a few stars!

Here is my observation:

The limiting magnitude was around 3.5, which means that a few constellations (including Cassiopeia and the Big Dipper) were visible. But I think if I had asked passersby how many stars they thought they could see, their initial guess would be either zero or perhaps "one or two". The problem is not only that the sky is bright, but especially that glaring lights and illuminated buildings make it hard to look at the sky with dark adapted eyes. Here is a street view several blocks away from the station:
 

"Street view in Brussels" by Christopher Kyba is licensed
under a Creative Commons Attribution 4.0 International License.

If you have to look up directly into a streetlight, you're not going to be able to seem much more than the streetlight... Contrast that contemporary view with van Gogh's impression of a night scene in Arles.

Van Gogh's scene is notable in how there are no glaring lamps at all, and for the relative darkness of the building facades. The scene provides an inspiration of what I think we should aim for in urban lighting: warm feeling light, no glare, no street light directly shining on windows, good uniformity in the pedestrian areas, and above it all, bright stars shining visibly in the night sky.

---

If you look up again at my observation, you will see that for a number of stars I had to report that I wasn't sure if I could see them or not. These stars were mainly in the constellation Draco, which is quite extended, and therefore hard to orient towards when the sky is bright. I hope that in a future version of the app, we will be able to further optimize which stars the app asks you to examine. I recently requested funding from the EU for a big project, in which the Loss of the Night app plays a small part. If the proposal is successful, I would be able to produce a new version of the app sometime in 2019 or 2020. Please cross your fingers for me!

Looking for stars in Berlin

Last week we hosted the first ever international conference on artificial light at night, ALAN 2013! After two evening events of the conference, I took advantage of being out on a night with clear skies to use the Loss of the Night app in three places. I've labelled their approximate locations on this map of Berlin at Night from the Suomi-NPP VIIRS image:

The first location, "Dahlem" is in a suburb at the edge of Berlin. The second, "Spreebogen" is an unlit park across the river from the brightly lit main train station. The third was a block away from Friedrichstraße, which is the brightest place in the center of the city (at least according to VIIRS). The results of the three measurements are shown below:

The plots are similar to the ones I showed in an earlier blog post. The vertical axis shows how bright or faint the stars are. Stars I said I couldn't see are shown in red, stars I wasn't sure about are shown in blue, and stars that I said I could see are shown in black. The horizontal axis shows the time since the measurements were started (each small tick is one minute).

The horizontal lines show the estimate (maximum likelihood fit) of how bright the sky was at each location. With a few assumptions, it's possible to turn this value into a very rough estimate of how many stars you can see. In Dahlem, about 300 stars could be seen with that level of skyglow, in Spreebogenpark about 200, and near Friedrichstraße around 100. By comparing the three datasets you can see that the app is really working, and that in each location my results are quite self-consistent. Unfortunately, I didn't happen to have an SQM with me to check how closely these results compared to the actual sky radiance.

I am continuing to learn a lot about the stars and constellations by using the app!  Since September I now know learned the stars Vega and Deneb and the constellation Cygnus (the swan) very well.  I am also slowly getting used to Perseus. I think that all of the stars I wasn't sure about were in the (very sparse) constellation Draco.

Why more stars=better observation

Every observation produces useful data, but the more stars you observe, the more precise your data will be. To demonstrate this, here is an example of a minimal complete observation (7 stars), taken by my former colleague Marco Starace near the center of Dresden, Germany last night:

The black stars show stars Marco could see, and the red circle is the one star that Marco was not able to see (Fainter stars have a larger magnitude). Based on this data alone, our best guess for the limiting magnitude (faintest visible star in the sky) is probably somewhere around 4.7. But we would also have to assign a very large uncertainty to the observation, because it's possible that Marco made a mistake with star 6 or 7 (or even on other stars). If he made a mistake on the last star, the limiting magnitude could be as large as 6.5, and if he made a mistake on the 6th star, it could be as low as 3.8. It is critically important to understand exactly how large this uncertainty is, so I hope that we will be able to obtain a grant to study this in the future.

But look what happens when Marco continued past 7 stars (the blue square is a star he was unsure about):

Now we have a lot more information. Our estimate for the limiting magnitude will now be lower (near 4), and the degree of uncertainty will be far smaller. It now looks quite likely that the limiting magnitude is smaller than 4.4, and nearly certain that the limiting magnitude is above 3.

Whenever you observe more than 7 stars, you help us in two ways. First, your observation will be of a higher quality, because of the reduced uncertainty. Second, you help us to better understand the results of observers that search for only 7 stars. That's because we can make an estimate of limiting magnitude based on your first 7 stars, and then test how well that estimate holds based on your additional stars. This will help us to eventually produce unbiased estimates, and it will also help us estimate the uncertainty on those 7 star observations.

One other thing that I hope this data conveys is that incorrect decisions aren't the end of the world. If you search for a lot of stars, then we will be able to make sense of your data, even if there are occasional mistakes somewhere. With time, we will learn which stars users find it easy to make accurate decisions about quickly, and we will also learn which stars are problematic (long decision time and high percentage of incorrect decisions). This information will make future versions of the app more accurate and also easier to use.

So thanks to Marco for his observations, and thanks to all of you who go past seven stars!

(For anyone just joining us, you can download the Loss of the Night app here. Also, we need a translator for the Russian and Arabic version ASAP! If you're interested, please contact me.)