Abstract. The objectives of this study were to assess the predictability of exceptionally cold water in the Taiwan Strait (TS) and to develop a warning system on the base of scientific mechanism, which is a component of the information technology system currently under development in Taiwan to protect aquaculture against extreme hazards. Optimum interpolation sea surface temperature (SST) data were used to find exceptionally cold water days from January 1995 to May 2017. We found that the SST and wind speed over the TS are low and strong in La Niña winters, respectively. According to tests conducted using relative operating characteristic curves, predictions based on the Oceanic Niño Index and integrated wind speed can be employed at lead times of 60–210 and 0–30 days, respectively. This study utilized these two proxies to develop a possible warning mechanism and concluded four colors of warning light: (1) blue, meaning normal; (2) cyan, meaning warning; (3) yellow, meaning moderate risk; and (4) red, meaning high risk. Hindcasting winters over the period 1995–2017 successfully predicted the cold water hazards in the winters of 2000, 2008, 2011, and 2012 in prior to the coldest day ~ 20 days.