--- title: "My sky is not dark and I have the numbers" author: "Tomas Vrba (@dust)" date: 2026-05-31T08:33:38.256Z updated: 2026-05-31T08:33:38.256Z canonical: "https://jot.place/@dust/my-sky-is-not-dark-and-i-have-the-numbers" description: "Two years of SQM-L readings from one hole in my lawn, the night a floodlight went up next door, and what 0.33 mag/arcsec^2 costs in hours." tags: - "astrophotography" - "evidence" - "field-notes" - "light-pollution" - "measurement" --- # My sky is not dark and I have the numbers The meter is a Unihedron SQM-L, the narrow one with the lens, 20 degrees of sky rather than the 84 degrees the plain SQM sees. Unihedron are in Grimsby, Ontario. Same instrument since January 2024, same hole in the lawn for the tripod leg, same procedure: five readings at the zenith, throw away the first, mean the other four. Nights are included only when the moon is below the horizon, the sun is more than 18 degrees down, there is no cloud within 30 degrees of the zenith, and there is no snow on the ground. I drop snow nights because my own readings over snow come out several tenths brighter and I refuse to let that into a monthly mean without separating it, which I have not yet done properly. :::info The manufacturer quotes a systematic uncertainty of about 10 percent, which is 0.1 mag/arcsec^2. That figure gets used to dismiss small changes and it should not be. A systematic offset is common to every reading from the same unit, so it shifts my whole time series together and cancels out of any difference. My absolute numbers are good to about 0.1. My differences from one meter on one tripod are far better than that. Accuracy and precision are different quantities and surveyors get taught this in the first week. ::: ## Two years | Month | 2024 mean | n | 2025 mean | n | change | |---|---:|---:|---:|---:|---:| | Jan | 20.94 | 7 | 20.91 | 9 | -0.03 | | Feb | 20.97 | 6 | 20.95 | 5 | -0.02 | | Mar | 20.91 | 8 | 20.62 | 6 | -0.29 | | Apr | 20.88 | 9 | 20.57 | 8 | -0.31 | | May | 20.71 | 3 | 20.42 | 2 | -0.29 | | Jun | none | 0 | none | 0 | | | Jul | 20.68 | 2 | 20.39 | 2 | -0.29 | | Aug | 20.86 | 9 | 20.55 | 7 | -0.31 | | Sep | 20.95 | 10 | 20.64 | 9 | -0.31 | | Oct | 20.99 | 8 | 20.68 | 7 | -0.31 | | Nov | 21.02 | 6 | 20.71 | 6 | -0.31 | | Dec | 21.05 | 5 | 20.74 | 4 | -0.31 | 138 nights. Three things to say about the table before anyone quotes a row at me. June is empty for a reason that is not light pollution. The sun's altitude at lower culmination is $h = \varphi + \delta - 90$, and at my latitude of just under 50 north with the solar declination above 22 degrees, $h$ never reaches minus 18. So from roughly 22 May to 22 July there is no astronomical darkness here at all. Geometry, not neighbours. The only bright sky I have no complaint about. May and July are contaminated by residual twilight for the same reason and I am not deleting them, I am flagging them. Two readings is not a monthly mean, it is an anecdote with a decimal point. December looks like my darkest month and I do not believe it. December is when I throw out the most nights for snow, so what survives is a biased sample of the clearest, driest, coldest air I get all year. Treat 21.05 as a best case, not an average. Everything else in that table is Bortle class 4, brighter rural, 20.8 to 21.3 on his scale. Which was already the answer to a question I did not want to ask. ## 14 March 2025 My neighbour mounted a floodlight on his gable, aimed across the property line, roughly eleven metres from the tripod hole. Motion sensor. A cat sets it off. I had fourteen readings from 3 January to 12 March 2025: mean 20.92, standard deviation 0.04. Fourteen readings from 19 March to 30 April: mean 20.59, standard deviation 0.06. Difference 0.33 mag/arcsec^2, which is more than eight standard deviations of the before period and more than three times the scatter of either sample. I should say what I did not measure. I have not separated scattering by the air above the garden from light bouncing off my own north wall and into the meter's 20 degree cone. I know the number at the meter. The number at the meter is the number that costs me exposure time, so for my purposes it is the right number, but I am not going to dress it up as atmospheric physics. ## What 0.33 costs Sky brightness changing by $\Delta\mu$ changes sky flux by $10^{\Delta\mu/2.5}$. For 0.33 that is $10^{0.132} = 1.355$, so 36 percent more sky per pixel per second. In the sky limited case the signal to noise per pixel goes as $$\mathrm{SNR} \approx \frac{S}{\sqrt{B}}\sqrt{t}$$ so holding SNR fixed while $B$ rises by a factor $k$ means $t$ rises by the same factor $k$. Not by $\sqrt{k}$. Linearly. Thirty six percent. On the 218 hour galaxy project I finished this year, that is 77 additional hours. At my average of 3.4 hours per usable night and my long run rate of 18 percent usable nights, 77 hours is 23 nights of imaging, which takes about 128 calendar nights to collect. One lamp on one gable costs me four months per project. Here is the rest of the same arithmetic: | Sky | mag/arcsec^2 | Integration needed, relative to my garden now | |---|---|---:| | my garden before 14 March 2025 | 20.92 | 0.74 | | my garden now | 20.59 | 1.00 | | top of Bortle class 4 | 21.30 | 0.52 | | Bortle class 2 | 21.60 | 0.39 | | Bortle class 1 | 22.00 | 0.27 | 218 hours here is 113 hours at 21.3, 86 hours at 21.6, and 60 hours under a genuinely dark sky. The Jizera Dark-Sky Park has existed since 4 November 2009, about 75 square kilometres straddling the Czech and Polish border, the first transnational preserve of its kind. I have been. I cannot take a mount, a pier and three winters there. The published light pollution maps put my area inside a band, and a map cell averages over ground far larger than my village. Ivo has written about what a projection costs you before you have measured anything at all, in @nullisland/every-projection-is-wrong-here-is-how-much, and the same argument runs downward as well as outward. A raster covering a continent cannot know about a gable. My meter is one point in one cell and one lamp moved it further than the width of a Bortle boundary. I waited six weeks before knocking. Then I brought printouts, which was the wrong instrument for the job. He was perfectly friendly. He believed I was complaining about glare in a bedroom window. I was complaining about 36 percent. There is no shared unit for that conversation and I have not found one yet. My daughter is seven. She has seen the Milky Way from this garden, so I cannot claim she has not. Bortle's own description of a class 4 sky says the Milky Way well above the horizon "is still impressive, but lacks detail". That is the politest sentence in the entire scale and it is the one I keep coming back to, because she has never once seen it look like anything. Year three of readings started on 2 January. Same hole in the lawn. I keep going because a time series is the only argument that has ever moved a municipal lighting decision anywhere, and mine is 138 nights long and it is not finished.