Bayesian Model Averaging (BMA) and Bayesian Hierarchical Model (BHM) are statistical postprocessing techniques for calibrating precipitation forecast ensembles. BMA is a mixture model of predictive densities, while BHM is a fully Bayesian alternative to BMA. Both techniques are applied on a case-study. BMA is applied to quantitative precipitation, yielding a better calibration than the ensemble in homogeneous areas. For qualitative precipitation, both BMA and BHM forecasts are more calibrated than the ensemble. However, BHM yields a worse performance due to the “shrinkage” effect, that lets the forecasts vary across a small range of values.
Bruno F., Cocchi D. , Rigazio A. (2011). Alternative approaches for probabilistic precipitation forecasting. FOGGIA : CDP Service Edizioni.
Alternative approaches for probabilistic precipitation forecasting
BRUNO, FRANCESCA;COCCHI, DANIELA;
2011
Abstract
Bayesian Model Averaging (BMA) and Bayesian Hierarchical Model (BHM) are statistical postprocessing techniques for calibrating precipitation forecast ensembles. BMA is a mixture model of predictive densities, while BHM is a fully Bayesian alternative to BMA. Both techniques are applied on a case-study. BMA is applied to quantitative precipitation, yielding a better calibration than the ensemble in homogeneous areas. For qualitative precipitation, both BMA and BHM forecasts are more calibrated than the ensemble. However, BHM yields a worse performance due to the “shrinkage” effect, that lets the forecasts vary across a small range of values.I documenti in IRIS sono protetti da copyright e tutti i diritti sono riservati, salvo diversa indicazione.