トレンド

So what did we do? It all starts with the #CitizenScience project Globe at Night (GaN), which you've probably seen me boosting before. In GaN, participants select which of a set of star charts most closely matches the sky they see. (3/)

画像の説明を読む

ALT

1,770

Participants use the GaN webapp (available in dozens of languages!) to also say where they were, the date and time of the observation, and sky conditions. (4/)

画像の説明を読む

ALT

1,642

English webapp here: globeatnight.org/webapp/ All languages here: globeatnight.org/downloads (5/)

Different available lanuages of the webapp

画像の説明を読む

ALT

1,605

The big question - and one of the two reasons it took me ten years to publish this - is what do you do about the fact that the observations aren't in the same place from year to year? Check out ES and PT in different years: (6/)

1,513

What I realized is that we could use the World Atlas of

@FabioFalchi9

and team as a baseline from which to compare similar places: science.org/doi/10.1126/sc (7/)

Image showing the World Atlas of artificial night sky brightness.

画像の説明を読む

ALT

1,398

Sure, we might not have observations from the center of Beijing and Paris in every year. But if parts of those places were the same as each other in 2014, then we can test how areas in the same sky brightness "class" are changing. (8/)

1,334

Is that really a fair comparison? As it turns out, yes, it is. We've shown in the past that there is a very strong correlation between GaN reports of star visibility and Fabio's map: nature.com/articles/s4155 (9/)

1,311

In the graph above, NELM stands for "naked eye limiting magnitude", which is equivalent to a measure of how many stars you can see. Sky brightness here is relative to starlight, so 10 means ten times brighter than starlight. (10/)

1,225

Regression toward the mean - Wikipedia

Now with the INDIVIDUAL observations it's a different story. There's a wide range of human visual abilities, so not everyone sees the same amount of stars. The key to this working is having A LOT of observations, thanks to regression to the mean. en.wikipedia.org/wiki/Regressio (11/)

en.wikipedia.org

1,237

Thankfully,

has been attracting citizen scientists to participate around the world, year after year, since 2006 (12/)

Image showing a histogram of how many people participated in each year, and where they were.

画像の説明を読む

ALT

1,190

That image is from our new Science paper, and shows where and when people have been participating. Europe and North America have well distributed observations in time and space. The rest of the world, unfortunately, has not. (13/)

1,083

So what happens if we make a graph like in toot #9 above, but with different years? It looks like this: (14/)

1,058

(This is the same as in toot #9, even if the x-axis label looks a bit different. 15/)

1,019

What you see is that 2011 and 2021 agree with each other reasonably well when citizen scientists are looking at starlit skies, but they don't agree with each other AT ALL in places that have a lot of #LightPollution. (16/)

999

So we can see that people are reporting fewer stars every year, but how can we go from that to getting an estimate of how the brightness of the night sky is changing? This is where my former student

@yigitoa

enters the picture. (17/)

1,047

Let's say we assume that the whole world is increasing or decreasing in brightness at the same rate. Then we can easily adjust the World Atlas sky brightness (relative to 2015) for the date when a citizen scientist made their Globe at Night observation. (18/)

940

The problem is then to figure out what the rate of change is. Yigit and I did this using the "maximum likelihood method". (19/)

937

The basic idea is that we can estimate the probability a person would pick a certain star chart given where they live and how quickly the sky is brightening. (20/)

892

If you dress up what I just said in an equation, it looks like this. "P" is how likely one given citizen scientist would say "I pick chart 3" and "L" is how likely the whole dataset would look the way it does if the rate of change in skyglow was "r" (21/)

874

There are a few other parameters in there, like how often citizen scientists make mistakes (like saying they live in the wrong place), what people are likely to report if there is no #LightPollution in the sky, and how variable one person's rating is from another's. (22/)

812

We searched through all sorts of different combinations of those parameters to find the one that made L (the likelihood) the biggest. The result? The sky brightness where people who participate in Globe at Night live is increasing at about 10% per year. (23/)

メディアを再生できません。

再読み込み

863

10% per year is a lot. Like, a lot a lot. That means doubling every 7 years or so. (24/)

885

Let's put it this way: If a baby is born today in a place where her parents can see 250 stars, by the time she is 18 she will only be able to see 100. The stars are like... (25/)

メディアを再生できません。

再読み込み

1,069