zbMATH — the first resource for mathematics

Bayesian model averaging and application to accumulated precipitation in Beijing. (Chinese. English summary) Zbl 1363.62026
Summary: Bayesian model averaging (BMA) is an important tool for statistical postprocessing forecast ensembles in dynamical modeling. First, BMA is introduced, including basic models, parameter estimation, and criteria for choosing days of training period and sampling from typical forecast distributions. Detailed discussions are then made about its implementation to the data of daily accumulative precipitation in Beijing from May 2 to August 31, 2011. It is suggested that BMA statistical postprocessing is superior to raw ensemble forecasts both in terms of accuracy and calibration.
62F15 Bayesian inference
62M20 Inference from stochastic processes and prediction
62P12 Applications of statistics to environmental and related topics
Full Text: DOI