Philip’s Machine Learning Experiments

Climate research is a data-rich field. We have terrabytes of in-situ observations, and many petabytes each of satellite observations, reanalyses, forecasts, and simulations. Modern Machine Learning (ML) methods promise powerful new ways to analyse and improve both forecasts and reconstructions; if we can learn to use them effectively.

This page documents my attempts to learn and use ML methods. I chose to use TensorFlow (as it’s powerful, popular, and open), and to work with data from the 20th Century Reanalysis (as it’s open, accessible, and I already have software for working with it).

• Getting climate data into TensorFlow

• Getting started - a simple autoencoder
• Perturbing the simple autoencoder
• A deep autoencoder
• A convolutional autoencoder
• Perturbing the convolutional autoencoder

• Old-style assimilation
• Old-style assimilation of precipitation

• ML Utilities package
• Small Print

This document and the data associated with it are crown copyright (2019) and licensed under the terms of the Open Government Licence. All code included is licensed under the terms of the GNU Lesser General Public License.