The Use of Rank Histograms and MVL Diagrams to Characterize Ensemble Evolution in Weather Forecasting

J.A. REVELLI, M.A. RODRIGUEZ, H.S. WIO

ADVANCES IN ATMOSPHERIC SCIENCES

SCIENCE PRESS

Lugar: Marrickville NSW; Año: 2010 vol. 27 p. 1425 - 1437

0256-1530

Rank Histograms are suitable tools to assess the quality of ensembles within an ensemble predictionsystem or framework. By counting the rank of a given variable in the ensemble, we are basically makinga sample analysis, which does not allow us to distinguish if the origin of its variability is external noiseor comes from chaotic sources. The recently introduced Mean to Variance Logarithmic (MVL) Diagramaccounts for the spatial variability, being very sensitive to the spatial localization produced by infinitesimalperturbations of spatiotemporal chaotic systems. By using as a benchmark a simple model subject to noise,we show the distinct information given by Rank Histograms and MVL Diagrams. Hence, the main effects ofthe external noise can be visualized in a graphic. From the MVL diagram we clearly observe a reduction ofthe amplitude growth rate and of the spatial localization (chaos suppression), while from the Rank Histogramwe observe changes in the reliability of the ensemble. We conclude that in a complex framework includingspatiotemporal chaos and noise, both provide a more complete forecasting picture.