Evaluation Of ECMWF And CMA S2S Models For Extreme Precipitation Forecasting Capability In Southern China Under BSISO And Its Influence

The occurrence of summer extreme rainfall over southern China(SCER)is closely related to the boreal summer intraseasonal oscillation(BSISO),and whether operational models can reasonably predict the BSISO evolution and its modulation on SCER probability is crucial for disaster prevention and mitigation.Here,based on the reforecast data(1998–2012)from the China Meteorological Administration(CMA)and the European Centre for Medium-Range Weather Forecasts(ECMWF)in the Subseasonal to Seasonal(S2S)Project,both deterministic and probabilistic evaluation methods are utilized to systematically assess the prediction skill of SCER and the BSISO index.Furthermore,we also explore the modulations of two BSISO modes(including BSISO1 that propagates northeast/northward with a period of 30–90 days and BSISO2 that propagates northwest/northward with a period of 10–30 days)on SCER forecasted by these two S2 S models.Then,the sources of the associated prediction biases are diagnosed and discussed.The main conclusions are concluded as below:(1)Although both S2 S models can predict the spatial pattern of mean and standard deviations of summer rainfall over China,the prediction biases also exist in the two models,to some extent,and the ECMWF model obviously outperforms the CMA model.As for the ECMWF model predictions,the intensity of summer-mean rainfall and variability is slightly higher over the northern part of southern China(SC)and slightly lower over the southern SC.The CMA model displays an evident underestimation of amplitude for both mean and variance of rainfall over entire SC.The ECMWF(CMA)model multi-member ensemble mean can directly predict the SCER occurrences skillfully at a forecast lead time of ?14(7)days.(2)The ACCs for BSISO1(BSISO2)index tend to drop with the lead time,indicating that ensemble mean forecasts from the ECMWF and CMA models show skillful prediction of the BSISO1(BSISO2)index(with ACC larger than 0.5)at lead time of 24(15)days and 10(8)days,respectively.The elimination of phase errors would improve BSISO prediction skill much more than the elimination of amplitude errors would.For the prediction with respect to occurrence and probability changes of SCER modulated by BSISO1(BSISO2),the ECMWF model shows obviously better skill than the CMA model,which is able to effectively predict the probabilistic information of SCER under the modulations of BSISO1(BSISO2)within lead time of 2 weeks(1 week).(3)The diagnostic results of modeled moisture processes further suggest that the moisture convergence(advection)induced by the BSISO activity serves as the primary(secondary)source of subseasonal predictability of SCER.With better prediction of the moisture convergence during phases 2–4(5–7)of BSISO1(BSISO2),higher skills can be obtained in the probability prediction of SCER.The present study implies that a further improvement in predicting the BSISO and its related moisture processes is crucial to enhancing the subseasonal prediction skill of SCER probability.The S2 S models have broad application prospects for making extended-range predictions of SCER probability based on the BSISO index.