Observer News

The Solar Terminator Problem

Images of sunsets

Photos taken by citizen scientists that were submitted as terminator observations. Thank you to all the citizen scientists who contributed terminator observations in order to help solve this problem.

The NASA GLOBE Clouds team at NASA Langley Research Center (LaRC) launched the (Solar) Terminator Problem, a special intense observation period (IOP) to collect citizen science solar terminator observations. The special IOP was launched to help validate a new satellite cloud detection algorithm. The goal was to identify cases where geostationary satellites (or GEO) are falsely identifying cloud coverage during the solar terminator. This is done by comparing citizen scientists' observed total cloud cover with the GEO satellites’ total cloud cover.

Ryan Moore, an undergraduate researcher and NASA intern, was a member of the team analyzing the satellite and citizen science data and is presenting the findings here. However, before we dive into the results and application of the IOP, here is some background.


The Solar Terminator

The solar terminator -- also known as the twilight zone -- is the line that separates the day-lit side of the Earth from the dark side. During the solar terminator, there are low levels of light present and not complete darkness which makes it difficult to accurately detect clouds. The NASA LaRC SatCORPS (Satellite ClOud and Radiation Property retrieval System) team has developed cloud masks or innovative ways of detecting cloudy and clear regions to improve satellite data. The SatCORPS team used the IOP to find examples when the satellites did not accurately detect clouds to test their new cloud mask.

A visual example of the terminator. Image credit: NOAA

GLOBE participants of the IOP were instructed to take observations from up to one hour after sunrise or one hour before sunset, as this is the best time to capture the solar terminator. In total, citizen scientists worldwide contributed over 600 terminator observations from March 2021 to July 2021.

This figure shows the Red, Green, Blue (RGB), and Cloud Mask SatCORPS products. The cloud mask product displays detected clouds in blue and clear skies in green.


The Process

Any large discrepancies between the citizen scientists’ data/observations and satellites’ could possibly be a case of the satellite not correctly detecting cloud coverage. For example, a citizen scientist in Oklahoma alerted the GLOBE Clouds team to the above case of false cloud detection through their observation. As seen in the above figures, the RGB product displayed minimal clouds over the Midwest; however, in the cloud mask there is a high percentage of clouds shown as the terminator is passing through.

This figure shows the GOES 16 images centered over the location of the Oklahoma citizen scientist as well as their sky photographs taken using the GLOBE Program’s GLOBE Observer app.

Once we identified more cases like the one above of satellites falsely detecting cloud cover during the terminator, we applied the new cloud mask to them in order to see if it better detected cloud cover. This new cloud mask was created by the NASA SatCORPS team at NASA LaRC. It utilizes machine learning in order to better detect clouds at the terminator.



This figure displays again the RGB and Cloud Mask SatCORPS products however this time the far right image shows the new cloud mask with the machine learning approach. As you can see in the third figure on the right, there are less instances of cloud coverage present. This was the previously shown case in Oklahoma. The new cloud mask shows a significant amount of clear skies in the Midwest that older versions of the cloud mask were classifying as cloudy areas. However, the citizen scientist in Oklahoma saw clear skies, while the new cloud mask did report some cloud cover which shows that this new cloud mask is not perfect.

This figure shows another case where the new SatCORPS machine learning algorithm improved cloud detection during the solar terminator.

And since the new algorithm needed to continue to be refined, the research team made use of another citizen scientist observation. This time from an observer in Michigan. The above case (see figures) was found because of a citizen scientist in Michigan and their terminator observation. In the old SatCORPS cloud mask, you can see some diagonal lines that resemble a ladder. These lines contain false cloud detection and are present over the location where the observation was made as well as the Midwest. The new SatCORPS cloud mask doesn’t have these lines and clears up the false cloud detection. Originally the algorithm reported 43.75% cloud coverage over the location where the observation was made, now the new cloud mask reports 0% cloud coverage. The photographs from the Michigan citizen scientist do show the presence of contrails shows that the cloud mask still needs a bit more work to detect contrails.

This figure shows the GOES 16 images centered over the location of the citizen scientist in Michigan as well as their sky photographs taken using the GLOBE Program’s GLOBE Observer app.



Without the help of citizen scientists all over the world, this problem would have been much more difficult to solve! The variables reported and photos captured for this special IOP were extremely accurate. This allowed us to easily identify cases of satellite false cloud detection during the terminator. Due to the great work of citizen scientists, the new SatCORPS algorithm can be confidently applied to the GOES-16 satellite data. In the future, it will be applied to the rest of the satellites in the geostationary constellation.

Listen to Ryan describe his research in the video below, or watch a longer version of the presentation.


About the author

Our guest blogger this month is Ryan Moore. Ryan (a junior at the University of Maryland (UMD)) worked on the Terminator Problem as an NASA GLOBE Clouds and SatCORPS teams summer intern. Ryan had two mentors this summer: Marilé Colón Robles, the project scientist for NASA GLOBE Clouds and Dr. Bill Smith, the lead for NASA LaRC SatCORPS team and who requested solar terminator observations from GLOBE participants. As an intern, Ryan was in charge of identifying citizen scientist observations where geostationary satellites incorrectly reported total cloud cover during the terminator. He used citizen science, ceilometer, and satellite data to find these cases.