Seasonal prediction of very hot days in Hong Kong using NCEP CFSv2

In this study, the seasonal predictability of very hot days in Hong Kong was investigated. The Wet Bulb Globe Temperature (WBGT) was adopted as the objective measure of heat stress to human, and the historical hot day statistics from the Hong Kong Observatory (HKO) and those from Climate Forecast System version 2 (CFSv2) hindcast runs were compared, for the JJAS season between 1982 and 2007. Our analysis shows that there is indeed a certain degree of predictably of hot days statistics a season ahead. The monthly mean anomalies of the CFSv2 and HKO WBGT data show a correlation with R2 being 0.4212. Furthermore, “Very Hot Days”- defined as the days with WBGT in the top 15% of its climatological distribution- were identified for each season from the HKO and CFSv2 data. The number of Very Hot Days in JJAS from the hindcast runs is also significantly correlated with observation. Also, we evaluated the performance of the binary “hot seasons” forecast based on whether the number of Very Hot Days for a particular season is larger than its climatological mean value. Our results show that for such a binary forecast, the Hit Rate is 0.769 with a False Alarm Ratio of 0.286. The corresponding Hanssen-Kuipers Discriminant is 0.462. Finally, we noted a general increasing trend in the surface temperature as found in the CFSv2 data. After removing the linear trend in the seasonal mean temperature, the resultant forecast skill is only slightly reduced.