NASA Center for Climate Simulation

Abstract

The emergence of deep learning techniques in machine learning has seen an explosion of applications across numerous scientific disciplines. In recent years, climate science has adopted the use of convolutional neural networks to detect extreme weather events[1]. Breakthroughs in neural networks first given in [3] have numerous applications across many interdisciplinary fields. These state of the art techniques allow for extremely accurate classification of images[1][2]. In this study, we aim to detect a type of extreme weather event called a derecho. Given a set of images of wind vector fields and magnitude, we use a CNN to predict whether the image is a derecho.