Detection of weather events in optical satellite data using deep convolutional neural networks

Early detection and forecast of extreme weather events such as tropical cyclones wildfires are becoming increasingly crucial for mitigating their catastrophic damages. Remotely-sensed satellite data enable us to detect assimilate these into models improve global-scale forecasts. Despite the significant advances in imagery image processing techniques, assimilation methods currently employ conventional approaches that can be slow inaccurate. This letter aims bridge this gap by utilizing deep convolutional neural networks (CNN) accurately rapidly patterns images. InceptionV3, ResNet50, VGG16, VGG19, four state-of-the-art CNN models, used classify 9081 optical images five different convective cell clouds, cyclones, roll wildfires, dust storms. All obtained a satisfactory average accuracy more than 80%. Among CNNs, VGG-Nets were found have difficulties distinguishing convection from patterns. Overall, InceptionV3 was best model achieving an 92% detecting systems. The demonstrates strength techniques pattern recognition instrumental atmospheric science research weather/climate models.