Abstract: Low temperature and cold damage in autumn has great adverse effect on late rice production, accurate early warning is of great significance for the high and stable yield of late rice. Using the principle of fuzzy mathematics, we studied the starting time early warning output model of low temperature and cold damage in Tianyang, selected a membership function suitable for the initial date emergence period of the low temperature and cold damage in autumn, calculated the membership degree of each prediction factor, built the multiple linear regression with the knowledge of mathematical statistics, and made the first early-warning of initial date emergence period of the low temperature and cold damage in autumn. Then, we used the principle of fuzzy mathematics and the logical relation between parallel and intersection operations to make the second early-warning. The results showed that the historical fitting rate of the first early-warning reached 87.3%, the historical fitting rate of the second early-warning reached 88.5%. During four years’test early-warning, the first early-warning was correct in three years, the correct rate reached 75.0%. The second early-warning was tested for one year, and the result was correct. The fuzzy mathematics warning output model for low temperature and cold damage in autumn is scientific and easy to understand, the correct rate of historical fitting and the accuracy rate of test early-warning are ideal.