By combining the detrended fluctuation analysis (DFA) method with surrogate data method, and using the heuristic segmentation algorithm as well as Chi-Square statistics, we develop a new method to determine the threshold of extreme events, e.g. stochastically re-sorting detrended fluctuation analysis(S-DFA) method. By using the S-DFA method, we obtain the thresholds of extremely low temperature events from 1961 to 2006 in China and analyze their spatiao temporal characteristics of distribution. We also validate the effectiveness of the S-DFA method through extreme event detection using the temperature series. By defining the composite index of extremely low temperature events from the angle of predictability, this composite index is integrated with the information about the frequency and the strength of the extremely low temperature events, with considering the characteristic of regional climate system. Based on the composite index, we divide the extremely low temperature events during 19612006 in China into four different zones according to their own rank. The composite index of extremely low temperature tends to be katabatic on the whole, Before the the early 80's in the 20th century, the composite index changed according to two distinct 10-year quasi-periods, and after that the composite index was in a downward trend and was well below the average. Until after the mid-90's in the 20th century, it rose to about average value once again.