Assessing the Gumbel-Logistic Model's Performance in Modeling Long-Term Rainfall Series in a High-latitude Plain Region

Abstract

Abstract The frequency of extreme events is increasing on a global scale due to various factors, accounting for natural disaster such as landslides, floods and droughts. Heilongjiang province, which is agriculturally important for China, has suffered from several extreme rainfall events in the past as a high-latitude plain region (43°26’ N-53°33’N). Analyzing historical rainfall characteristics of Heilongjiang region is useful for understanding and predicting the behavior of extreme rainfall events, providing effective strategy of reducing the economic loss of agriculture. In this paper, precipitation data from 1974 to 2017 is selected to estimate whether the monthly extreme rainfall and annual rainfall series in Heilongjiang Province obey the Gumbel-logistic model. Then we compare the Gumbel-Logistic distribution based on the data length of 5 years, 10 years, and 20 years with the original 44-year data to determine the appropriate data partition length and explore the reasons for the errors. Combining the empirical frequency and KS test, this paper concludes that the proposed model is appropriate for the representation of the joint distribution of annual monthly rainfall and corresponding amounts, which improves the rainfall assessment of these two positively correlated variables in this region. Univariate and bivariate return periods and correlation coefficients are derived as the basis for assessing the risk of extreme rainfall. The proposed model obtained from the 5-year data length has the largest error, while the distribution curve is progressively smoother as the data length increases.