Digital Camera as Light-Pollution Meter: Initial Results

Some time ago, I asked on s.a.a. whether anyone had tried
using a digital camera to measure the brightness of light-
polluted skies. Not receiving an answer, I decided to
purchase on myself and try the experiment. I won't have
really good data until I get a string of clear moonless
nights, but preliminary indications are that my digital
camera (the Canon A60) can measure sky brightness in
urban and suburban conditions at least as accurately
and much more quickly than any other method I have tried.

I selected the Canon A60 because it was the cheapest
digital camera I could find ($230 US) that had all the
features I wanted, including in particular full manual
control and a reasonably fast lens. So far, I am
quite satisfied with the camera in that regard. I
also bought it as a cheap way to experiment with
general-purpose digital photography, and here I am
less satisfied; the picture quality seems even worse
than can be explained by the low (2 megapixel)
resolution. But maybe I am just spoiled by working
with film.

To measure sky brightness, I set the resolution at
640x480, set the mode to Manual, the ASA to 400,
and the Effect to Black and White. Color photography
might yield useful data, but for the moment I don't
want to be distracted by another variable. I shoot
wide open (F/2.8) at the shortest focal length, and
usually with the shutter speed set to 15 seconds,
although faster shutter speeds may be preferable
under really heavy light pollution.

So far, I have mostly been shooting the zenith, where
laying the camera on the ground works beautifully.
Once I start to get serious about sampling light
pollution at different altitudes, I will need to
use a tripod and perhaps a level. Note, however,
that with a little effort, one can tell exactly
where the camera was pointed by the recorded date
and time and by which constellations are visible
within it. Stars down to mag 4 or so show well.

After shooting, I download the pictures, open them
in Corel PhotoPaint (I am too cheap to buy Adobe
Photoshop), and look at the saturation of the pixels.
There is a lot of random fluctuation at low light
levels, so I like first to resample to 32x24 pixels,
which smooths away the pixel-to-pixel variation
quite thoroughly. Unfortunately, it also blends the
stars into the background, which may cause the sky
brightness to be overestimated under semi-dark skies.
Initial indications are that this effect is negligible
under urban skies and average suburban skies. I avoid
sampling the corners, where there is serious vignetting.

The pixels recorded in the JPEG file are on a scale
from 1 to 255, which needs to be calibrated against
the actual light intensity. To do this, I shot a
series at different speeds indoors under artifical
light, with the camera tripod-mounted. Assuming that
the shutter speeds are accurate, and assuming that
the sensed intensity is proportional to the exposure
time (i.e. no reciprocity failure), this should indicate
the correlation between intensity and pixel saturation.
I have some evidence that both assumptions are fairly
accurate, but I need to do more work in that area.

1/500 3
1/400 4
1/300 3
1/250 4
1/200 5
1/160 6
1/125 6
1/100 7
1/80 9
1/60 12
1/50 14
1/40 19
1/30 25
1/25 30
1/20 37
1/15 54
1/10 65
1/8 76
1/6 91
1/5 104
1/4 118
0.3 129
0.4 150
0.5 165
0.6 178
0.8 197
1.0 211
1.3 226
1.6 233
2.0 242

I didn't shoot any slower than 2 seconds, but beyond
that, the saturation starts to level off in a hurry,
as it must considering that it only has 13 levels
more to go before full saturation. Below 1/250,
the measurement is essentially identical to a
completely dark frame, fluctuating anwyhere from
2 to 4. The camera provides half-decent resolution
over a factor of about 500X in intensity even
within a single shot at a specific speed and f-stop.
The response is essentially linear between pixel
saturations of 7 and 65, and nearly linear down to
saturation 5 and up to saturation 100 or so.

It turns out to be important to shield the camera
from incident light; the lens obviously exhibits
considerable glare.

Here are some zenithal readings that might be of
interest. I will hold off on publishing my readings
at other altitudes for a while.

full Moon, otherwise dark sky 50
my window, 4 miles from Boston 94
7 miles from Boston 64
12 miles from Boston 50

The Moon in the first case was only 20 degrees above
the horizon, and I was in a steep-sided mountain valley
shooting through a small gap in the trees. All that
baffling may well have made the sky darker than it
would otherwise be.

The reading from my window was made after Harvard had
turned on the lights at its playing field 1/2 mile
distant. With the lights, the zenithal reading
deteriorates to 100, about 6% worse. This illustrates
just how big the effect can be from one relatively
minor source if it is fairly close. The degradation
towards the horizon in that section of the sky is
much worse, of course. Harvard keeps those lights
on every night until 10PM, come rain, fog, or snow.
I am sure that a little shielding could reduce the
impact 50% or more, but Harvard wields far too much
political clout to be coerced to mend its ways.

The reading 12 miles from Boston, just outside the
inner belt road, is probably higher than it should
be due to a Moon 4.5 days past full 8 degrees above
the horizon. Even so, I could see the Milky Way
easily through Cygnus at that site, and the Milky
Way also shows up faintly in the photograph, being
about 6 levels brighter than the sky on either side.

Curiously, the Milky Way was not at all obvious from
the site 7 miles from Boston, even though the reading
indicates a sky only 25% brighter than the 12-mile
site. The subjective difference between those sites
seems much greater than that.

- Tony Flanders

On 15 Sep 2003 08:18:56 -0700, (Tony Flanders) wrote:

Some time ago, I asked on s.a.a. whether anyone had tried
using a digital camera to measure the brightness of light-
polluted skies. Not receiving an answer, I decided to
purchase on myself and try the experiment...

Interesting experiment, Tony. I'd like to suggest a couple of things to look at
if you start refining things. First, many of these cameras become quite noisy at
higher ISO settings. In many cases, there will be less noise with an exposure of
2T at ISO 100 then an exposure of T at ISO 200. So it is worthwhile evaluating
your camera at different sensitivity settings. Second, the shutter speeds
reported are often somewhat inaccurate. On the camera display, they tend to look
like traditional speeds: 1/4, 1/60, etc. But the actual speeds may be different.
Look at the header in the JPEG file and the actual speed is usually given. Many
image viewers will display the header for you (ACDSee, for example). Also, if
you are using Windows XP, the image properties can be displayed by right
clicking. Select Advanced under the Summary tab for all the technical
information embedded by the camera.

_________________________________________________

Chris L Peterson
Cloudbait Observatory
http://www.cloudbait.com

(Tony Flanders) wrote in message m...

Some time ago, I asked on s.a.a. whether anyone had tried
using a digital camera to measure the brightness of light-
polluted skies. Not receiving an answer, I decided to
purchase on myself and try the experiment...

By the way, one things that still perplexes me is how to calibrate
the digital-camera readings against some absolute scale, like
magnitudes per square arcsecond. I have already started to
calibrate the camera against itself, so to speak, like determining
that a reading of 118 means a sky ten times as bright as a
reading of 19, and I have more ideas on how to refine and
re-affirm those calibrations. But what I would really like
is to be able to say that a reading of 118 means, say mag
17.0 per arcsecond. To do that, I need at least one
universally available surface of known brightness.

I have thought of using Bill Ferris' (or is it Brian Skiff's)
datum that the sky at full Moon is mag 18.0 per sq arcsec,
but that must vary up to almost 15% simply due to the geometry
of the Sun, Earth, and Moon, and much more depending on the
altitude of the Moon above the horizon and the transparency
of the air. A better measure might be the sky brightness at
the zenith at the end of nautical twilight, but that too
depends greatly on atmospheric conditions.

The brightness of, say, the Cygnus star cloud comes to mind, but
that seems to be right near the bottom of my camera's sensitivity
curve. The Double Cluster might serve, but it is smaller than
might be desireable.

Other thoughts are the brightness of a sheet of white paper
laid orthogonal to the direction of the full Moon. Or perhaps
I could calibrate it against some known daytime brightness,
like an 18% gray card in full Sun, and extrapolate from there.

Any ideas?

- Tony Flanders

On 16 Sep 2003 12:25:21 -0700, (Tony Flanders) wrote:

By the way, one things that still perplexes me is how to calibrate
the digital-camera readings against some absolute scale, like
magnitudes per square arcsecond...

This might work for you, if the precision of the camera is good enough and you
can work out the image scale properly:
http://home.earthlink.net/~stanleymm/CCD_topics.html

Otherwise, you can use this calculator to do the calculation using a regular
cooled astronomical CCD camera (your own, or a friend's) and then calibrate your
digicam against that.

_________________________________________________

Chris L Peterson
Cloudbait Observatory
http://www.cloudbait.com

On 16 Sep 2003 12:25:21 -0700, (Tony Flanders)
wrote:

(Tony Flanders) wrote in message m...


Other thoughts are the brightness of a sheet of white paper
laid orthogonal to the direction of the full Moon. Or perhaps
I could calibrate it against some known daytime brightness,
like an 18% gray card in full Sun, and extrapolate from there.

Any ideas?

- Tony Flanders


It is quite possible that the digicams response vs lux may not be
strictly linear.