Research On Summer Extreme Precipitation Over Northeast China Based On Nonstationary Extreme Model

In recent years,extreme weather events have occurred frequently and become more and more serious around the world.Most of the present researches adopt the stationary extreme model and assume that the parameter values in the model remain unchanged.With the development of theoretical and applied research,the assumption of constant parameter is gradually questioned.It is great theoretical and practical significance to establish the nonstationary extreme model by functionalizing the parameters in the extreme model and selecting appropriate covariates to describe these functions.Based on large-scale climate indices,medium and small-scale meteorological indices and multi-scale synergistic indices,this paper established non-stationary extreme value models,and comprehensively used a variety of modern statistical methods to study the genetic mechanism and risk evolution characteristics of summer extreme precipitation in Northeast China.The main research contents are as follows:(1)The non-stationary change characteristics of the extreme precipitation series in Northeast China are analyzed by using the trend analysis method of linear regression,quantile regression and Mann-Kendall and the significance of the non-stationary change characteristics is verified.The results show that the non-stationary of extreme precipitation sequence is widespread.In early summer,extreme precipitation has an obvious upward trend in most areas of Northeast China,while in late summer,extreme precipitation has a downward trend in most areas of Northeast China,especially in Songhuajiang and Liaohe river basins.(2)By means of simple correlation analysis,quantile regression,principal component analysis and other methods,the covariates with important influence on extreme precipitation were selected.Based on the selected multi-scale meteorological indices and synergistic indices,the non-stationary GEV model and GAMLSS model were established to analyze the genetic mechanism of summer extreme precipitation in Northeast China.Moreover,the linear/nonlinear relationship between climate indices and extreme precipitation series are analyzed according to the way that the covariates are introduced into the Northeast summer AMP extreme models.Based on 107 sites in northeast China,the results of the two models show that the non-stationary extreme models established with EASMI index,SOI index and NINO3.4 index as covariates the most common.In late summer,the GAMLSS model with NECVI index as covariates is also the main non-stationary extreme model.ENSO is the main nonstationary climatic factors for the Northeast summer AMP time series and EASMI is the second.NECVI also plays an important role for the Northeast summer AMP time series,especially in later summer.In addition,there is a non-linear relationship between the covariates and the Northeast summer AMP time series.(3)Based on the GEV model and GAMLSS model,the non-stationary time-varying moment is estimated to forecast the 2-year return period and 100-year return period extreme precipitation of summer extreme precipitation in Northeast China.The risk evolution of extreme precipitation is analyzed and the two prediction models are compared.The results show that the non-stationary time-varying moment model based on GEV model is more suitable to predict the two extreme precipitation events in Northeast China in summer.In early summer,two kinds of extreme precipitation in most areas of Northeast China show a significant upward trend year by year.While in late summer,the main areas are Nenjiang,Mudanjiang and Liaohe river basins show a significant upward trend.