Searching for Sub-Visual Atmospheric Structures in the Daytime Sky

On sunny, warm days the sun heats the Earth’s surface and the air close to it. Periodically a parcel of air will rise from this area due to the warmed air being buoyant. This parcel is thought to rise in an elongated column of fairly large size such that several hundred tons of air are lofted skyward. In doing so many of the particles generated by Earth-bound processes (pollen, smoke, dust, pollution, water vapor, etc.)  are brought with it. These particles are commonly known as aerosols. If the column reaches an altitude where the contained water vapor condenses then a cumulus cloud will form.

It is known that aerosols have a large effect on polarization of light, up to 30% or so. My first experiment in photographing these columns of air was to take 2 sequential photos of the sky with a linear polarization filter set to 90 degrees apart. Then in accordance with the article excerpt shown below and using an image processing program (Image Magik) I calculated the degree of linear polarization (DOLP) of each pixel from the formula given in the article. The resulting pictures are interesting and strange but do not show the expected structures.

I encourage others to make their own attempt at this goal as I am really a novice at image processing. No doubt there are many other ways of looking at this problem and I welcome all comments, thoughts and ideas. Thanks!

Excerpt from the article “Digital All-Sky Polarization Imaging of Partly Cloudy Skies” from Nathan J. Pust and Joseph A. Shaw

“It is our feeling that unseen aerosols and possibly thin clouds in what has recently been called the “twilight zone” between a cloud and the clear sky are reducing the DOLP in what appears to be clear sky. We believe that this effect on the sky polarization is directly related to the recently described observations of enhanced optical depth near clouds. In partially cloudy skies, we see DOLP reductions in clear sky areas between clouds that appear to be caused by subvisual aerosols and/or clouds. (Even though clouds appear to have hard edges, they are in fact surrounded by thin clouds.) Furthermore, these DOLP reductions show up in the clear sky long before we can physically see clouds in the sky.”

To determine Degree of Linear Polarization (DOLP) in each pixel he uses this formula:
(Image1pixel value – Image2pixelvalue) / (Image1pixel value + Image2pixelvalue)

Then he normalizes and stretches the result so it fills the whole 8 bit range of 0 to 255 pixel brightness values.

Some of my resulting pictures:

Author: Deane Williams, Connecticut, USA


Posted on March 22, 2017, in experimental, polarization and tagged , , . Bookmark the permalink. 9 Comments.

  1. If I understood your post correctly, you analyzed 8-bit JPEG images. That will definitely cause problems, since JPEGs are already processed and gamma corrected, making the data highly non-linear with respect to the amount of light. They are also compressed with a lossy algorithm, causing artefacts in the image. 8-bit precision is also fairly small. I recommend doing the same with 16-bit raw images. I will also try this tomorrow if the weather permits and report back.

  2. navis, Thanks for helping me. I had wondered about the suitablity of my standard camera and use of JPEG. I am not familar with 16 bit cameras. They must be for industrial or astronomical use. I am looking forward to your results.

  3. Normal DSLR cameras are capable of outputting raw images, which are usually saved as linear 16-bit images. Technically the real bit depth is usually around 14 bits though, since it depends on the A/D converters on the camera sensor. I think my Canon EOS 60D has 14-bit A/D converters. The weather seems good today, so I will go out and try capturing some images.

  4. I managed to capture some images, but did not see any inherently interesting. Here’s one image of the degree of polarisation: I only had a CPL filter, but I think that should work. The sun was to the left of the image. The dark part of the sky on the right is 90 degrees from the sun, dark values meaning linearly polarised light. I should have used image 1 as image 2 and vice versa to make linearly polarised regions light instead of dark, but I can’t be bothered to do it again right now.

    Today the sky is very clear, but I will try the same when it’s partially cloudy, with e.g. cumulus clouds.

  5. It’s afternoon here now, and we have some clouds. I tried redoing what I did in the morning, but with similar results. Only the normal polarisation of the sky is visible. I even tried to remove the sky gradient to see any fainter variation, but didn’t see anything. I didn’t have time to fully read the paper you linked yet, I just skimmed it earlier. I’ll look into it and see if I think of something new.

  6. THanks! That is an interesting image. A sky with some cumulus will ensure that there is a possibility of capturing these columns beneath them. Please try again when you have a better sky. What image processing program are you using to do the subtraction between the two polarized images to get DOLP?

  7. We had some cumulus clouds during the afternoon when I took more images, but I couldn’t see anything interesting. I might try again later after reading the linked article more closely. I processed the images with a software called PixInsight, but since you’re just doing pixel-wise subtraction, summing and division, it’s trivial to do it with e.g. Python or Matlab too.

  8. naavis, Thanks for continuting to think about how this might be achieved. You have far more equipment and knowledge than I do at this point but I am looking into finding a suitable camera. I just found another link that shows that aerosol dust strongly depolarizes light in a LIDAR system.

  9. I was wondering if perhaps your use of a circular polarizer could be part of the reason for not seeing the expected result. The first article I mention states that they used a linear polarizer so I purchased one at Edmunds Optics.

Leave a Reply

Fill in your details below or click an icon to log in: Logo

You are commenting using your account. Log Out /  Change )

Facebook photo

You are commenting using your Facebook account. Log Out /  Change )

Connecting to %s

%d bloggers like this: