Started on Nov 2017

Abstract of the process

SDO (Solar Dynamic Observatory) images are stored in our database, with a 6-minute cadence, since January 1st, 2012. Each image is of size 4096 by 4096 pixels. I want to investigate the possible changes in the calibration of the cameras as time goes by analysing the changes in the colour intensity of images.

In our database, each image is also summarized in 10 different parameters. One of the parameters is the Mean intensity. That is, each 4k by 4k image is segmented into 64 by 64 pixel patches, and each of those patches is then replaced by its mean value. In this way, every image can be represented by a much smaller image, a 64 by 64 matrix of means.

First, I pulled all the (small) images since 2012-01-01 to 2015-06-30. This gave me approximately 7200 images per month, for 42 months.

Then, I summarized each of those images to some basic statistics that could represent the images as a whole. I chose Min, Q1, Median, Q3, Max, Mean, and Variance as the statistics. As a result, I created 42 CSV files, each of which approximately 1.2 MB, describing all the images in terms of their statistics.

Next, I needed to describe the daily changes in the intensity values of the captured images. I chose boxplots to show the daily distribution of the statistics. For instance, a single boxplot for the month 2012-01 shows the changes in the Mean statistics of the images captured in that period. In other words, the distribution of the Mean value of all the images captured within January 2012.

Finally, I put all the boxplots in one plot so that the changes in the bigger picture can be noticed. I repeated the same visualization for the first and third quartiles (instead of the mean value) to present a more accurate analysis.

The following visualizations are coded in R, using ggplot2 and ggthemes packages.

Below, you see the same plots, this time without the boxplots themselves.