Connecting Machine Learning with Atmospheric Science (Invited Presentation)

In recent years, machine learning has achieved a number of high profile successes in areas such as machine vision, speech recognition, and gaming. This success has prompted a growing interest in applying machine learning to new areas; in particular, the physical sciences. A recent workshop included applications to particle physics, genomics, astrophysics, and medical imaging, for example.

A similar trend is taking place within atmospheric science. Modern machine learning algorithms are being applied to various problems in the field, such as post-processing, model emulation and nowcasting. However, focusing solely on the novelty risks distorting the picture. While machine learning does offer many new and useful techniques, they are often deeply connected to more traditional methods in statistics and numerical modelling.

This talk will discuss some possible connections between the two fields, as well as some of the practicalities of applying machine learning to atmospheric science datasets.