When the opportunity came up to do an undergraduate research project studying light pollution in town I jumped at it. A recent convert to the world of GIS, I immediately saw this as a spatial problem, and that I should map the brightness around town. I got to combine my dual interests in astronomy and geospatial science. Bonus!
This began with a task of driving around parts of town at night with a handheld light meter, a specially designed light pollution meter, and a Trimble GPS to both record the position and the meter values. Some problems with this started to show up, as light pollution levels varied on a large scale (some areas were brighter than others) and the small scale (I avoided standing too close to street lights, but at what point is that altering the survey?).
During this time I was taking a field methods class and was flipping through the class guidebook (Comptons Field Methods), and I came across a section I never considered before: aerial surveying methods. Thus began my adventure into aerial light pollution surveying.
The handheld light meter would sit in the porthole with a fixture I made. This would effectively tell us the "brightness" of the belly of the plane (the double layering of the floor gave an aperture effect so that the light meter was telling us brightness directly downward and not so much from the sides). But the main light measuring instrument was something most people have: a digital SLR camera.
A foam sleeve would fit around the camera lens and keep it from falling through the porthole. Exposure and focus settings were manually set so pictures would be consistant.
The next tricky bit involved a lot of funky math (Excel is my best friend sometimes): what elevation should this nighttime aerial survey be flown at? A number of factories had to be considered. Flying too high would mean more climbing time, and the plane isn't cheap. At a higher altitude, the resolution of ground features would also be less. Flying high would also mean fewer passes, as each picture would cover a larger area, and fewer passes meant less flying time. But because it would be at night, the shutter speed was slower than normal, and flying too low would cause the lights to show up as streaks from a longer exposure time. Flying too low would mean a lot of passes, and a lot of pictures, with the possibility of gaps occuring in the continuum of photos.
By some wizardry I don't care to explain I came to an optimum value: 6000 feet above ground. This was actually the cheapest elevation to fly (a balance between climbing time, descent time, and time at elevation) to cover the entire city with the least amount of photos without sacrificing resolution etc etc. The next bit was to figure out what camera settings would produce the best image at this elevation. To do this I went up to a scenic overlook of the city and played with settings to see which would provide the best image of the lights that were about a mile away.
By knowing our elevation, the area each photo would cover, and the aircraft speed, I knew I had to take a picture every 15 seconds to ensure some overlap of the photos along each line. Three passes would cover the whole city with a little extra and, this part happened to be very coincidential but I'm going to claim otherwise, by making standard rate turns at cruising speed each turn-back would cause the next line of photos to overlap the edges the previous line by just the right amount (about 15%), agan ensuring no gaps. Just like the diagram in Compton's! It looked like I knew what I was doing.
The date was set for the night flight. We would want to do the survey during "typical astronoy hours", about an hour after sunset to 11pm. Does it matter? It might. Some businesses may shut off their lights after a certain time, and I wanted data to best reflect astronomy hours.
Equipment was ready, the flight was planned from startup to shutdown, and we had 3 Trimbles ready (just in case). One of my flight instructors would be piloting, one of my professors would be ensuring GPS-itude, and I would be sitting in the back of the plane taking the photos and measuring the light through the portal.
This was quite a strange ride sitting on the floor in the back of a small plane at night. I couldn't see outside at all. To begin the survey we would pass over the airport at 6000 above ground going eastbound. In the meantime I opened up the portal and prepared the camera setup. Through the LED viewer on the camera I could see a few streetlights, but it wasn't much to see. Every few seconds I would snap a new picture, and at the end of the line I was instructed we would be making the first turn. I took this chance to take the camera out and take a look for myself.
Looking straight down at a city at night from 6000 feet through a foot-whide hole in the bottom of a plane is something that doesn't happen very often. I will never forget it.
Two more passes were made. A slight wind was giving us a bit of a drift which my professor caught by watching the GPS track on the Trimble, which we were able to correct for. Afterward, I stuck the light meter into the ground and would call out brightness values every few seconds, which would be recorded with a time value.
What good is time, though? Fortunately, the Trimble time-stamped each GPS data point, so time and position were interchangeable. Also, the digital camera time-stamped each photo, so the position of each photo could also be quickly determined. To make sure the camera and the Trimble were synced to the same time I just took a picture of the time on the Trimble display (clever, no?).
My next task was putting this plethora of information into ArcGIS. The light meter data was easy, just a few data points with lux values. The aerial photography was a big trickier, and I spent many many hours georeferencing aerial photographs, which showed up quite well (individual streetlights, even ones next to eachother, are distinguishable).
Fortunately, street lights are easily referenced to streets, and larger buildings and parking lots could be referenced to regular satellite imagery. The end result of the photography was a georeferenced, high resolution image of Winona at night.
The important thing about the georeferencing is that I could compare both the light meter readings from the airplane and the readings from the ground. We only managed two passes using the light meter but it actually did pick up areas which were brighter than others.
The large "bright" area in the middle is the group of lights just right of center on the aerial photo, and the second smaller "bright" area is from the smaller group of lights down and to the right. This whole operation took about two hours of flying time, which was great, as ground surveying (while higher resolution) was very time intensive. By combining the information obtained with the aerial and ground surveys we were able to pinpoint specific locations in the city that were contributing most to light pollution. Specifically, businesses and commercial zoning districts were the brightest areas in the city, while only comprising a small percentage of the area. Traditional "save the night sky" advice is typically aimed at proper residential lighting but from this it seems like light pollution action should be focused at businesses.