Abstract:Considering the diverse operational need of Meiyu prediction, evaluation of multi-scale forecast skill of CFSv2 for Zhejiang Meiyu precipitation in 2018 was systematically conducted. Despite the forecast results on May 31, which is consistent with the observation, the initial forecast results from March 1 to May 31 could not accurately capture the trend of less precipitation in most regions of Zhejiang in June generally. On the extended-range (ER) scale, the total precipitation predicted by CFSv2 during Meiyu period exhibited approximately 30% less than observation. Based on the correlation coefficient, root mean square error and newly defined integrated forecast skill index, the forecast performance of CFSv2 on ER scale was comprehensively evaluated. The results showed that the forecast skills for total precipitation, daily regional-mean precipitation and daily stational precipitation are very limited during Meiyu period, and the forecast results for Meiyu region in Zhejiang were slightly better than for the entire Zhejiang. The evaluation results indicated that CFSv2 forecast products of Zhejiang show significant systematic bias. On the ER scale, with the decrease of forecast time scale, the improvement of forecast skill was not noticeable, and instead there is an optimal forecast time scale. Also, during the full forecasted period for each ensemble, the most skillful time-interval was commonly not the nearest time frame. To sum up, the CFSv2 forecast for Meiyu precipitation on the ER scale may not be sensitive to the initial field.