Temporal-spatial Evolution Of Extreme Precipitation In The Poyang Lake Basin,China

Extreme weather/climate events increased significantly due to global warming.The frequent occurrence of extreme events has led to serious natural disasters,which resulted in serious socio-economic losses and human injuries.Investigation of the spatial and temporal evolution of extreme precipitation events can not only deepen the theoretical understanding of regional extreme weather and climate events,but also provide technical support for disaster prevention and mitigation,which is of great significance for regional water security and water resources management.As the first largest freshwater lake in China,there is less systematic and comprehensive spatio-temporal evolution study of extreme precipitation in the Poyang Lake Basin（PLB）,especially the uncertainty of extreme precipitation risk under the influence of non-stationary characteristics increases.Thus,it is urgent to explore the temporal-spatial distribution characteristics and numrical model simulation of extreme precipitation events in the PLB.Based on the daily records from 16 meteorological stations in the PLB from 1959 to 2019,the temporal-spatial and non-stationary characteristics of extreme precipitation were investigated and analyzed comprehensively from three dimensions（intensity,frequency and duration）using Pre Whitening Mann-Kendall（PWMK）,Extreme-Point Symmetric Mode Decomposition Method（ESMD）and the Generalized Additive Models for Location,Scale,and Shape（GAMLSS）.The Weather Research and Forecasting（WRF）model was used to simulate a typical extreme precipitation event in the PLB to provide a basis for future extreme precipitation forecasting and prevention.The following are main findings:（1）The results showed that the intensity and frequency of extreme precipitation increased significant in the PLB,while the extreme precipitation duration decreased from1959 to 2019.The extreme precipitation has high intensity,high frequency,and short duration features in the PLB.There was a clear distinction between flood season and non-flood season for extreme precipitation.Extreme precipitation was concentrated in the northern and central PLB during the flood season.However,the extreme precipitation events mainly happened in the central PLB during the non-flood season.The extreme precipitation amount increased trend was 2.10 mm·a^-1 in the Xinjiang Basin,which has the most increment over the PLB.In the flood season,the extreme precipitation has longer duration but weaker intensity and smaller range.This situation was contrary during the non-flood season.（2）The intensity and frequency of extreme precipitation showed stationary characteristics in the PLB.However,the duration of extreme precipitation showed non-stationary characteristics.The extreme precipitation indices of Nanchang station showed significant non-stationary characteristics,especially the extreme precipitation intensity.The best simulation effect was found in the basin of Fuhe Basin,and the worst simulation effect was found in the whole basin.The extreme precipitation persistence indices of Fuhe Basin,Ganjiang Basin and Xiushui Basin all show non-stationary characteristics.The mean value of extreme precipitation intensity in the basin has a significant increasing trend compared with the variance,and the variance increased in some areas,which increased the risk of extreme precipitation in the future.The frequency analysis results of the GAMLSS model showed that the extreme precipitation centers were located in the north and east,and the low value centers were located in the southwest,and the frequency of heavy rainfall increases gradually from the southwest to the northeast,and the precipitation duration was longer in the northeast and southwest of the PLB.The non-stationary characteristics of the mean and variance significantly affected the trend changes and fluctuation characteristics of extreme precipitation events and increased the risk of extreme precipitation.The smooth model reduced the intensity of extreme precipitation,while the non-stationary model captured the time-varying characteristics of extreme precipitation events and was more suitable for extreme precipitation risk prevention.（3）The extreme precipitation intensity simulated by WRF model in the PLB was weak and the precipitation extreme center was northward,which failed to capture the spatial and temporal distribution characteristics of precipitation.The results of error analysis based on the observation sites showed that the average Root Mean Square Error（RMSE）,Mean Absolute Error（MAE）,and Bias of daily precipitation simulated by WRF are 47.24 mm·d^-1,30.37 mm·d^-1,and-8.33 mm·d^-1,respectively.The Pearson Correlation Coefficient（CC）ranges from-0.24 to 0.71.The total precipitation simulation results from July 2 to 10showed that the WRF model underestimated the precipitation in the western part of the Xinjiang Basin and the northern part of the Fuhe Basin by more than 300 mm,and overestimated the total precipitation in the northern part of the Xiushui Basin and the northern part of the Raohe Basin by more than 100 mm.The non-stationary characteristics can lead to increased biases of extreme precipitation events intensity,frequency,and duration of WRF simulations,especially underestimated the intensity of extreme precipitation.Therefore,in future extreme precipitation simulation and forecasting studies,it is necessary to enhance the selection of the physical parameterization scheme of the WRF model and to evaluate the coupling of multiple models.