Abstract: Distribution features of model variables accompanied with short-time strong rainfall events are investigated based on hourly precipitation data from 1605 automatic weather stations and ECMWF 0.125°×0.125° fine grid model products, and a method of short-time strong rainfall forecasting based on threshold determination is established. Results show that short-time strong rainfall occur more frequently in Fujian inland area and less in Fujian coastal area during pre-rainy season, and the diurnal variation exhibits double peaks with the notable one at 1700 BT and the inapparent one at 0500 BT. The box difference index is useful to check whether a variable could differentiate short-time strong rainfall events well. The box difference index of humidity variables like specific humidity at 925 hPa and total column water vapor are most prominent followed by K index and convective available potential energy (CAPE) which shows these variables have good performances in distinguishing short-time strong rainfall events. Some variables like temperature difference between 850 hPa and 500 hPa and temperature change in 24 h at 500 hPa perform poorly in differentiating short-time strong rainfall events.The minimum threshold method based on the minimum values of variables after eliminating outliers works well in judging short-time strong rainfall events, which could decrease vacancy forecast rate effectively through increasing missing forecast rate appropriately compared with adopting real minimum values as threshold. In the key area (25.9°-27.1°N, 116.4°-117.4°E), TS (threaten score) of validation set in 2016 with 12 h interval reaches 0.5 at daytime and 0.3 at nighttime just based on the minimum threshold method. Revising threshold of variables with high box difference index values could improve the accuracy with the nighttime TS of validation set in 2016 increasing from 0.3 to 0.34. TS of 2016 is relatively lower compared with that of 2014-2015, and the cause may be that short-time strong rainfalls happen much more frequently in 2016 which is a very strong El Niño year.To establish a potential forecast model of short-time strong rainfall during pre-rainy season, Fujian is divided into grids of 1°×1°, and minimum threshold method is applied in each grid followed by threshold revise of variables with high box difference index values. This model could analyze all kinds of variables comprehensively besides those considered by weather forecasters. TS with 12 h interval at daytime mainly ranges from 0.3 to 0.5 while TS at nighttime is relatively lower. TS in inland area is much better than coastal area both at daytime and nighttime mainly because short-time strong rainfall occurs more frequently in inland area than coastal area during pre-rainy season climatologically.