Forecasting impact rather than weather — a big-data solution for meteorologists?

- Indicates paper has been withdrawn from meeting
- Indicates an Award Winner
Tuesday, 6 January 2015: 1:30 PM
221A-C (Phoenix Convention Center - West and North Buildings)
Chris Doyle, Environment Canada, Vancouver, BC, Canada

Forecasting impact rather than weather – a big-data solution for meteorologists?

Consistent with the evolution of many meteorological service providers, the Meteorological Service of Canada is transitioning from forecasting pure weather elements. One potential direction is toward the production of forecasts of weather impacts and their probability or risk. To produce impact and risk forecasts is complex, and their successful introduction depends not only on a clear understanding of social vulnerabilities and thresholds, but also on an understanding of the historical impacts of weather conditions and events. Given a sufficiently large historical database of antecedents, events and impacts, an impact forecast can be produced by combining output from an impact database and Numerical Weather Prediction systems. Probability information can be gleaned through the use of ensembles.

Solely relying on forecaster memory of analogues to recall potential impacts given a forecast is probably not a completely reliable strategy. Automated systems designed to compare forecast conditions to antecedent impacts would increase the reliability of impact anticipation and improve the utility of forecasts, although obtaining useful historical data is likely to be progressively challenging as data is sought from earlier periods. Data produced by impact and probability-of-impact-based forecast systems would still require interpretation and skeptical assessment - a potential role for future meteorologists.