Deep learning twined spatial analysis for detection of mysterious fairy circles.

The 'fairy circles' of Mali, Namibia, and Australia, intriguing geomorphological features, have gained a lot of attention since some of them have been reported as sources of natural hydrogen. Although there have been many theories on the origin of these mysterious circular depressions in specific geological terrains, they are often explained as the escaping of free gas generated through the serpentinization of rocks at deeper depths. These circles are visible in the satellite images, which can be accessed on platforms like Google Maps and Bing Maps, making them potential sources of image data. Major research in this area has been carried out primarily in Mali, Namibia, Australia, etc.; however, considering the vegetation patterns, they may also be present in different parts of the world. This study combined computational modelling and geomorphological image data to develop pre-trained CNN-based predictive models that can detect the likelihood of fairy circles from satellite images. Through rigorous training, the study reports a set of such models that can detect the presence of fairy circles with an accuracy of 98%. Using these models, satellite images can be easily screened to detect and locate such patterns across the globe.