| Located in the northeast of China, the Songhua River basin belongs to the EastAsian monsoon area with drastic climate changes. Its unique monsoon circulationdetermines the complexity and variability of the hydrological cycle, which is shownthrough frequent droughts and floods caused by the uneven temporal and spatialdistribution of precipitation. The Songhua River Basin is developed in population,industry and agriculture. The Songnen Plain is one of the major commodity grainproduction bases of China, so drought and flood disasters can caused large direct andindirect economic losses per year in this area. Some research data also have shownthat the extreme precipitation events had increasing trends, which would be continuein the future. Estimation of extreme precipitation of the Songhua River basin hasimportant significance in water resources allocation, irrigation project design andoperation and commodity grain base construction.Firstly, the statistical methods of linear tendency estimation, cumulative filterand commensurability method are used to study on the general characteristics,variation tendency, and periodic characteristics of flood and drought years based onthe precipitation data of different hydrological stations from1970to2003. The resultsshow that precipitation is not distributed evenly in a year; dry years and flood yearsalternately appeared during different years, what’s more, precipitation has large spacedifferences, which is caused by unique geographic and climate of the Songhua RiverBasin. The commensurable figures shows that flood/drought years have twodimensions periods:10~30-year period and2~5-year period, and the former one ismore obvious; astronomical factors of solar activity may be the primary cause of thisrule.Next, Annual Maximum sampling method and Partial Duration Series sampling method is used to establish AM model and POT model. With Maximum likelihoodmethod, parameters of6kinds of commonly used distribution functions are estimated.Then, Kolmogorov-Smirnov testing method and P-P/Q-Q Plots are combined toexamine the fitting degree, thus to determine the optimal fitting distribution, then tocalculate the extreme precipitation at different return periods. Results show that:different kinds of distribution functions have different fitting degree, and the optimaldistribution of AM model is the Generalized Extreme Value distribution, while, theoptimal distribution of POT model is the Generalized Pareto Distribution; exponentialdistribution has relatively low fitting results, which is closely related with its simpleparametric form and distribution model; Return level(that is extreme values atdifferent return periods) of the POT model is smaller compared with the AM model,while, considering the rationality of the sampling method, the POT model is viewedas optimal one which can better restore the original distribution of precipitation inSonghua River Basin.At last, considering the charoctoristics of precipitation distribution, extrememodels about flood-season and non-flood season are established respectively. Then,quantitatively calculate extreme precipitation at different return level. At the samereturn period, extreme precipitation in Baishan and Fengman station is more thanJiangqiao and Nierji station; precipitation change rate of Nierji station is lower withthe longer of return period.Simulation quality of extreme value models is closely related to many factors,such as the original data length and variation points, sampling methods, the thresholdvalue selection and parameter estimation, where the variation point of original datahas been paid more and more attention in recent years. Due to limited precipitationdata, relatively less in typical stations quantity and short time sequence has certaininfluence on the establishment and analysis of extreme value model. Strengthen theanalysis about effect of original data variation on extreme value simulation, flexibleselecting parameter estimation methods will become the research emphasis anddifficulties in the future work. |
PDF Full Text Request