The University of British Columbia

• Atmosphere-ocean climate dynamics
• Neural network models, machine learning, computational intelligence
• Seasonal climate and extreme weather prediction

Our goal is to understand climate variability, that subtle, nonlinear interplay between atmosphere, ocean and land. To accomplish this goal, we are deeply involved in the development and application of neural networks and other methods from the field of machine learning, a branch of computational intelligence. On the pragmatic side, we aim to develop models for seasonal climate prediction and extreme weather prediction.

A prominent example of climate variability is the famous El Niño-La Niña phenomenon, an irregular fluctuation of the climate system which produces anomalous warming in the equatorial Pacific during El Niño and cooling during La Niña, with notable influence on the Canada winter climate. El Niño/La Niña episodes can now be forecasted with reasonable accuracy 3-12 months in advance. Forecast techniques range from coupled global atmosphere-ocean general circulation models run on supercomputers to elegant statistical techniques on PCs.

Our group has pioneered the use of neural network and other machine learning methods for analyzing climate data and for El Niño/La Niña prediction. We have developed neural network models for nonlinear principal component analysis, nonlinear canonical correlation analysis, and nonlinear singular spectrum analysis (our codes are freely downloadable and have users from over 60 countries). We have identified nonlinear atmospheric teleconnection patterns in the extra-tropical Northern Hemisphere associated with the El Niño/La Niña and with the Arctic Oscillation.

Neural network models can also be combined with dynamical models to improve the parametrization in dynamical models, or even to form hybrid models -- e.g. a neural network atmospheric model has been coupled to a dynamical ocean model, yielding a hybrid coupled model of the tropical Pacific. With general circulation models having spatial resolution too coarse to reveal climate variability at local scales, machine learning methods are being used to downscale the model output to finer spatial scales, especially for precipitation and streamflow.

We have also started to look at applying machine learning methods to satellite data.

My graduate-level book "Machine Learning Methods in the Environmental Sciences: Neural Networks and Kernels" was published by Cambridge Univ. Press on 30 July, 2009.

Our climate forecasts are updated monthly on our web site: http://www.ocgy.ubc.ca/projects/clim.pred/.

Selected Publications