Advanced Ensemble Prediction Systems

Line graph of forcast generation
Simulation of adaptive coevolutionary EP response to improved input information (introduced at the generation indicated by the vertical line). Shown are the minimum (blue), mean (green), and maximum (red) CSI for the set of 100 best-performing EP algorithms for each generation prior to and after the introduction of the improved ANC input. Also shown is the observed frequency of convection occurrance at each generation (thick and dotted line), and the distribution of ANC CSI values for the moving window used for all calculations (set to prior 100 forecast cycles).

Paul Roebber continues to develop a method, known as Evolutionary Programming, into a form that produces “ensemble” predictions for high-impact weather and climate scenarios, such as heavy rainfall, flash flooding or hurricanes. Ensemble prediction systems, or any collection of experts, suffer from too much agreement among individual predictions, which prevents properly capturing the actual probability distribution of an individual event. Roebber’s methods show considerable ability to provide skillful forecasts while also correcting this problem. The technique is applicable to any data (not just weather and climate), opening up a wide range of future applications, and it is fully adaptive to changing inputs.

Article about this work (PDF)

Principal Investigator

Paul Roebber

Distinguished Professor, Mathematical Sciences - Atmospheric Science Group