Lorentz Center - Uncertainty Guidances in Science and Public Policy from 13 Nov 2017 through 17 Nov 2017
  Current Workshop  |   Overview   Back  |   Print   |   Home   |   Search   |     

    Uncertainty Guidances in Science and Public Policy
    from 13 Nov 2017 through 17 Nov 2017


Why the Weather Forecasts of the Future Forecast will Not Forecast the Future?


Speaker - Leonard Smith (London School of Economics and Political Science)

Leonard Smith is professor of Statistics at the Department of Statistics, The London School of Economics and Political Science (LSE) and Director of the Centre for the Analysis of Time Series (CATS). His original research interests focused on understanding the mathematics of nonlinear dynamical systems, the analysis of observational data, and the application of insights from those two areas to increase our understanding of actual phenomena. This core has broadened to include research questions on the communication of that understanding to the public, decision makers in industry and policy makers in government on one side, and to the philosophical foundations of uncertainty and mathematical modelling on the other.


Poincare noted that "some hypotheses are dangerous" and went on to say that mathematical physics can render us the service of identifying dangerous assumptions we make without knowing we made them. Different notions of our aims in forecasting physical systems like the weather are discussed in this light. I argue that the insights of nonlinear dynamics, in particular structural model error and the loss of topological conjugacy in nonlinear systems, prevents our making accountable probability forecasts, in much the same way that "chaos" prevents our making accurate point forecasts. An alternative future for the weather forecasting enterprise is suggested.