Copula in temporal data mining: The joint return period of extreme temperature in Beijing

Copula has become a popular tool in multivariate modeling widely applied in lots of fields, but less used in temporal data. The analysis of the extreme temperature is an important part of the study in climate change, and the data of extreme temperature is one of the temporal data. So in this study, copula is used to calculate the joint return period of extreme temperature (from station in Beijing) with the indices Frost Days (FD and Summer Days (SU₃₅). We used Anderson-Darling goodness-of-fit test (A-D test) to find the most fitted probability distribution and evaluate the 10-year return period, 50-year return period and 100-year return period based on the marginal distribution of the two univariate. After calculating the joint return period, we compared the results of univariate return period and joint return period with the reality. The results show that, the joint return period is more accurate than the univariate period, and by improving both the choice of indices and the copula method, the results should closer to the reality. This study is of significance to get a better understanding in temporal data mining by using copula method.