The MetObject Approach: Enabling Man and Machine Intelligence in Weather Forecasting
Speaker: Dr. Dominique Brunet, NSERC Visiting Fellow at the Cloud Physics and Severe Weather Research Section in Environment Canada.
Abstract:
The MetObject approach consists in representing weather data as a collection of geometric objects that can manipulated, stored and communicated. This conceptual simplification of data allows to link weather observations and numerical weather predictions with the skills of the human forecaster.
Three major questions emerge: 1) how to extract MetObjects from numerical weather predictions and remote sensing, 2) how to efficiently allow human interaction with MetObjects and 3) how to measure the quality of the created MetObjects. In this talk, we present our work on MetObjects in the particular scenario of Summer convective weather. We will illustrate that to answer these three questions, machine
intelligence will be necessary, but that man's intelligence should not be neglected.
Biography:
Dominique Brunet received the B.Sc. (2005) and M.Sc. (2007) degrees in Mathematics from Laval University in Québec City and the Ph.D. (2012) degree in Applied Mathematics from University of Waterloo. He is currently a NSERC Visiting Fellow at the Cloud Physics and Severe Weather Research Section in Environment Canada, working at the weather radar station in King City. Previously he was a Postdoctoral Fellow in the Department of Radiation Physics at the Princess Margaret Cancer Centre in
Toronto. He also briefly pursued a Postdoctoral Fellowship in the Department of Electrical and Computer Engineering at the University of Waterloo. His research interests include mathematical and statistical
image processing, image approximation and compression, image quality assessment, computer vision and pattern recognition, image-guided radiation therapy, and meteorological data analysis. Dr. Brunet is a
member of the SIAM Imaging Sciences group (2012-), of the SIAM Geosciences group (2015-) and of the IEEE Signal Processing Society (2013-).