Mayur Mudigonda, Soo Kim, Ankur Mahesh, Samira Kahou, Karthik Kashinath, Dean Williams, Vincent Michalski, Travis O’Brien and Mr Prabhat



Segmenting and Tracking Extreme Climate Events using Neural Networks


Predicting extreme weather events in a warming world is one of the most pressing and challenging problems that humanity faces today. Deep learning and advances in the field of computer vision provide a novel and powerful set of tools to tackle this demanding task. However, unlike images employed in computer vision, climate datasets present unique challenges. The channels (or physical variables) in a climate dataset are manifold, and unlike pixel information in computer vision data, these channels have physical properties. We present preliminary work using a convolutional neural network and a recurrent neural network for tracking cyclonic storms. We also show how state-of-the-art segmentation algorithms can be used to segment atmospheric rivers and tropical cyclones in global climate model simulations. We show how the latest advances in machine learning and computer vision can provide solutions to important problems in weather and climate sciences, and we highlight unique challenges and limitations