Study Of Stochastic Daily Rainfall Models Based On Rainfall Event Characteristics Under Climate Change

In recent years,climate change mainly characterized by global warming has produced a profound influence on the earth system,in which hydrological cycle is one of the most important and directly affected links.With the increase of global average temperature,the atmosphere tends to have a lager moisture holding capacity and the hydrological cycle accelerates,which results in an increasing of rainfall frequency and intensity and the frequent occurrence of extreme hydrological events,posing a serious threat to the safety of human life and property and social and economic development.Therefore,it is of great significance to investigate the impacts of climate change on hydrological cycle and provide important information for water resource managers and decision makers in order to effectively plan and manage water resources,design hydraulic structures and prevent extreme disasters under changing environment.Rainfall is one of the important driving factors for hydrological cycle and the rainfall event characteristics,i.e.rainfall depth,rainfall duration,inter-event time and rainfall temporal pattern,greatly affect the forming of streamflow.Taking into account the rainfall event characteristics in stochastic rainfall models can not only help better interpret the mechanism of rainfall processes,generate rainfall series with sufficient lengths for gauged and ungauged stations,but also provide multiple possible realizations of rainfall as inputs for hydrological modelling,which plays an important role in vulnerability assessment and prediction for hydrology and agriculture under climate change.This thesis firstly develops a new stochastic rainfall event model based on rainfall event characteristics and evaluates its performances at gauged stations and ungauged sites in Zhejiang Province.Then,taking the Qu River basin in the east of Zhejiang Province as the study area,the impacts of climate change on rainfall event characteristics are investigated.Furthermore,in view of the limitation that this developed stochastic rainfall event model cannot be directly applied in hydrological modelling,this thesis further constructs a stochastic rainfall model by coupling Markov chain mode and the developed stochastic rainfall event model,and evaluates its performances from two aspects of meteorology and hydrology.Lastly,through combining the stochastic rainfall model with a hydrological model,this thesis assesses the future changes of streamflow including mean,high and low flows under climate change,and also quantifies the contribution of different uncertainty sources in these flows.The main conclusions of this thesis are as follows:(1)Through simulating rainfall depth and rainfall duration using best-fitted Copula function,generating a rainfall type based on the occurrence frequency of different rainfall types and simulating rainfall temporal pattern of the specific rainfall type using transformed Monte Carlo method,a daily stochastic rainfall event model is developed and applied to 39 gauged stations in Zhejiang Province for performance evaluation.Subsequently,through hydrological regionalization,determination of the hydrological region that the ungauged site belongs to,inverse distance weighted interpolation of rainfall event characteristics from gauged stations in the same hydrological region and the statistical parameter transplantation of different rainfall temporal patterns from the nearest gauged station,this rainfall event model is extended to Jiande Station that is assumed as an ungauged site and the performance of this model extended to ungauged sites is verified at Jiande Station.The results show that this stochastic rainfall event model can well produce various characteristics of rainfall events,including rainfall depth,rainfall duration and their dependence structure and rainfall temporal patterns of different rainfall types as well.In addition,the performance of extending this model to ungauged sites is also very well.The above indicates that this model can be used for generating long series of rainfall event data for gauged and ungauged stations.(2)Using four RCP emission scenarios and 17 GCMs from the fifth assessment report of IPCC,the GCM simulated rainfall data is firstly bias corrected based on observations using the double gamma distribution mapping method at the monthly scale.Secondly,the good-performing bias-corrected GCMs,appropriate for regional climate impact assessment,are selected using the MAPE(Mean Absolute Percent Error)index.Thirdly,the performances of these selected GCMs on simulating rainfall event characteristics are assessed.Finally,the changes of rainfall event characteristics in the Qu River basin in the future are investigated.The results show that compared with the historical period 1971-2000,the number of rainfall events will probably decrease,but the frequency of extreme rainfall events,short-duration events,long-and extreme-dry spell events may increase in the mid-future 2041-2070 and far-future 2071-2100,and the changes in the mid-future are larger than that in the mid-future.In addition,the rainfall temporal pattern of type A(i.e.advanced rainfall type)for light rainfall events in the future will gradually become uniform,while that of type C(i.e.center-peaked rainfall type)and type D(i.e.delayed rainfall type)for heavy and extreme rainfall events will become more non-uniform.The above results indicate that the Qu River basin may be faced with a larger risk of floods and droughts in the future.(3)By coupling with Markov chain model,the stochastic daily rainfall event model is modified to be a stochastic daily rainfall model and further be applied to hydrological modelling The performances of this rainfall model are then evaluated from the aspects of rainfall time-series statistics,extreme rainfall,rainfall event characteristics,streamflow characteristics and extreme flows including high and low flows.The results show that except small rainfall(e.g.10%and 25%quantiles)and short-duration extreme rainfall(e.g.R1d and R3d)in monsoon season being slightly overestimated,this rainfall model presents a very good performance in reproducing other rainfall time-series statistics(e.g.average rainfall and various rainfall percentiles)and rainfall event characteristics(e.g.frequency of different rainfall event classes,rainfall temporal patterns,and occurrence frequency of different rainfall types under different rainfall event classes),particularly for the long-duration extreme rainfall(e.g.R5d and R7d).In terms of runoff characteristics,this model performs very well in producing general streamflow characteristics(e.g average flow and various percentiles),high flows(e.g.annual maximum 3-day and 5-day mean flow)and low flows(e.g.annual minimum 7-day,30-day and 90 day mean flow),except slightly overestimation of annual maximum 1-day flow.The drawback of this model is that it would underestimate the low-frequency variability(also known as inter-annual variability)of rainfall and runoff to some extent.(4)Under four RCPs and nine selected GCMs,the future changes of mean flows,high flows and low flows in the Qu River basin under climate change are investigated by combination of the newly-developed daily stochastic rainfall model and hydrological model,and the contribution of different uncertainty sources in these flows is quantified using the ANOVA method.The results show that the monthly flows in the monsoon season and annual flow are expected to increase in the future,and high flows increase but low flows decrease with larger changes in the far-future than that in the mid-future.This indicates that the Qu River basin may be faced with a higher risk of drought and flood disasters in the future.In the projected mean flows,GCMs contribute the largest uncertainty,followed by internal climate variability(i.e.stochastic uncertainty)and RCP uncertainty;In projected high flows,internal climate variability and GCMs are the two major uncertainty sources,and their contributions are equivalent.In projected low flows,the uncertainty sources in the mid-future are respectively internal climate variability,GCMs and RCPs while in the far-future,GCMs are the largest uncertainty source and the contribution of RCP uncertainty is almost equal to internal climate variability.Compared with high flow projections,RCP uncertainty and GCM uncertainty play a more obvious role in low flow projections,and become gradually more pronounced in projected mean flows and low flows with the extension of projection time.