Noctilucent clouds, also known as night-shining clouds, are appearing more often and lower in the sky than historically observed, according to scientists studying long-term shifts in Earth's weather patterns. The silvery-hued clouds scatter sunlight long after sunset and before sunrise, but they can be difficult to distinguish from lower-altitude clouds that resemble them.
To track changes in these rare clouds, NASA-supported researchers launched the Space Cloud Watch project, which asks volunteers worldwide to submit photographs. The project has struggled with the time-consuming process of manually verifying whether submitted images actually show noctilucent clouds rather than similar-looking atmospheric phenomena.
Volunteer Namai Chandra developed a machine-learning pipeline to automate the screening process. The system combines image pre-screening, cloud classification, and confidence-based review routing, training on a variety of cloud images including both noctilucent clouds and lower-altitude look-alikes commonly submitted to the project. Scientists Drs. Chihoko Cullens and Brentha Thurairajah collaborated with Chandra on the tool's development and refinement.
The identification tool is now available to Space Cloud Watch contributors who are uncertain whether they have observed noctilucent clouds, as well as project scientists using it to flag images for review.
