J53.2 Relative Skills of Two PQPF Mechanisms for Postprocessing Medium-Term Precipitation Forecast

Thursday, 11 January 2018: 1:45 PM
Room 18A (ACC) (Austin, Texas)
Yu Zhang, Univ. of Texas at Austin, Arlington, TX; and M. Scheuerer, J. C. Schaake, L. Wu, and C. Kongoli

In this study, we compare the efficacy of two mechanisms that postprocess ensemble precipitation forecast to generate probabilistic quantitative precipitation forecast (PQPF). The first one is the mixed meta-Gaussian distribution (MMGD) model that is a component of the Meteorological Ensemble Forcing Preprocessor (MEFP); and the second is the Censored Shifted Gamma Distribution (CSGD) approach. The MMGD relies on a Gaussian copula to establish bivariate distribution of forecast and observation, and further yield the conditional distribution of forecast. CSGD is a regression-based mechanism that derives PQPF from a prescribed distribution (CSGD) by adjusting the climatological distribution according to the mean,spread, and probability of precipitation (POP) of raw ensemble forecast. We apply each

mechanism to the reforecast of Global Ensemble Forecast System (GEFS) to create postprocessed PQPFs over the lead times between 24 and 72 hours. The outcome of evaluation experiment over mid-Atlantic region of US indicates that CSGD approach broadly outperforms MMGD in terms of both ensemble mean and the reliability of distribution, though the performance gap tends to be narrower at higher precipitation thresholds (> 5 mm). Analysis of a rare storm event demonstrates the superior reliability and sharpness of CSGD PQPF. Our work suggests that the CSGD’s incorporation of ensemble spread and POP does help enhance its skill, but CSGD’s use of optimization in parameter estimation likely plays a more determining role in its outperformance.


- Indicates paper has been withdrawn from meeting
- Indicates an Award Winner