Sequential aggregation of heterogeneous experts for PM10 forecasting

Özet

We experiment some methods of sequential aggregation for forecasting PM10 concentrations for thenext day, in the context of air quality monitoring in Normandy (France). Besides the field of applicationand the adaptation to the special context of the work of the forecaster, the main originality of this paperis that the set of experts contains at the same time statistical models built by means of various methodsand groups of predictors, as well as experts coming from deterministic chemical models of predictionmodeling pollution, weather and atmosphere.Numerical results on recent data from April 2013 until March 2015, on the whole network of fifteenmonitoring stations, illustrate and compare some various methods of aggregation. The obtained resultsshow that such a strategy clearly improves the performances of the best expert both in terms of predictionerrors and in terms of alerts. What is more, it reaches, for the non-convex weighting strategy, the“unbiasedness” of observed-forecasted scatterplot, extremely difficult to obtain.