Nonlinear Responses Of Extreme Streamflow To Climate Variability In Europe

In recent years,as a result of climate effects,floods have frequently occurred over large space in Europe,causing a large number of casualties and huge social and economic losses in many countries.In Europe,there are many international rivers.Once a flood occurs,it will have a continuous impact involving many countries.Thus,for flood risk management purpose,it requires an overall understanding of extreme runoff across Europe.However,Europe has diverse climate types,numerous mountains and complex terrain,and the drivers affecting extreme runoff in each basin are also different.With the joint effects of atmospheric circulation,oceans and land,climate and extreme runoff in Europe exhibits complex nonlinear relationship.This brings great challenges to flood prevention in Europe.Therefore,it is necessary to understand and quantify the nonlinear impact of climate on extreme runoff for different seasons in Europe,and to form a more overall understanding on how climate affects the extreme runoff in Europe,so as to provide a scientific reference for preventing transnational flood disasters.This study collected the daily runoff data at 649 European stations from 1956 to2016,based on which maximum daily runoff of each season was calculated.In addition,11 climate factors extracted from the atmosphere and ocean data,and 3 catchment factors were used to analyze the nonlinear relationship between climate and seasonal extreme runoff in Europe.By combining nonlinear decomposition and machine learning methods,a climate-based nonlinear hybrid model of extreme runoff was constructed to quantitatively evaluate the nonlinear effects of various drivers on extreme runoff in Europe.The nonlinear relationship was studied at both inter-annual and inter-decadal scales,and also at different seasons.By setting extreme climate scenarios,the impact of extreme climate on European extreme runoff was further explored.The main findings of the thesis are summarized as follows:(1)There is a clear nonlinear relationship between climate and extreme runoff in Europe.At all stations,nonlinear relationship was found between the maximum daily runoff and different drivers,which is a non-proportional increasing or decreasing relationship.Through the Ensemble Empirical Mode Decomposition(EEMD)method,the maximum daily runoff and all factors can be decomposed into 4 periodic components and 1 trend component.The decomposed components of runoff and different factors show relative consistent periods.More specifically,the maximum daily runoff and different factors exhibit periods of 2.4～3.6 years and 4～8 years on the interannual scale;on the inter-decadal scale,they show periods of 10～16 years and 20～40years.By analyzing the variance of different components,for most factors and the maximum daily runoff,the 2.4～3.6 years component has the largest variance contribution.These series are thus dominated by inter-annual signals.However,only two factors are dominated by inter-decadal signals,which are the first principal component of North Atlantic sea surface temperature in winter and the North Atlantic sea ice concentration(SIC)in all seasons.(2)The correlation between extreme runoff in Europe and various climate factors is weak on the inter-annual scale,but strong on the inter-decadal scale.With the periods increase,in all four seasons,the proportion of stations with significant correlation increase gradually.From the relationship between different factors and the maximum daily runoff,on the inter-annual scale,the maximum daily runoff is mainly affected by the catchment factors.For example,in autumn the precipitation has the greatest impact on the maximum daily runoff,and the proportion of stations with a significant correlation reaches 50%.On the inter-decadal scale,the maximum daily runoff is mainly affected by climate factors.For example,the second principal component of the North Atlantic sea surface temperature has the greatest impact on the maximum daily runoff in winter,and the proportion of stations with a significant correlation reaches90%.(3)In about 40% of the stations,at different time scales there are opposite associations between the extreme runoff and climate drivers,most notably between the components at inter-decadal scale and the trend component.For example,in 31% of the stations,we found on the inter-decadal scale that there is a significant positive correlation between the winter maximum daily runoff and the North Atlantic Oscillation(NAO),but on the trend component,there is a significant negative correlation.These stations are in the UK,central Europe,and most parts of Finland.(4)By comparing the fitting performance of the nonlinear hybrid model and the traditional multiple linear regression model,it is found that in almost all stations the performance of the new model is better in both the training and validation periods.From the fitting results of different seasons,in general,the fitting results in spring and winter are slightly better than that in summer and autumn.In particular,the fitting result of the maximum daily runoff in winter is the best.In the validation period,the proportion of stations with a coefficient of determination greater than 0.4 reaches 38.96%.In addition,when using the climate factors with 1 to 3 months leading time to fit the maximum daily runoff,for all seasons the fitting performance decreases with the leading time increases.(5)Climate has a greater impact on extreme runoff in spring and winter than in summer and autumn,with the greatest impact in winter.In different seasons,the impact of each factor on extreme runoff varies from region to region.For example,in spring,SIC has a greater impact on the maximum daily runoff in southern England and northern Germany.While in autumn,SIC mainly affects the northeastern region of Germany.The influence of each factor on the maximum daily runoff is also different on interannual and inter-decadal scales.On inter-annual scale,the effects of catchment factors on maximum daily runoff dominates.On inter-decadal scale,ocean-related climate factors are the main factors affecting the maximum daily runoff.For example,in spring and summer,the third principal component of North Atlantic sea surface temperature has the greatest impact on the maximum daily runoff.In autumn,SIC has the greatest impact on the maximum daily runoff.In winter,the fourth principal component of North Atlantic sea surface temperature has the greatest on the maximum daily runoff.(6)In order to analyze the impact of extreme climate on maximum daily runoff in Europe,we set up three extreme climate scenarios,normal extreme,relative extreme and very extreme.In temporal scale,the impact of extreme climate on the extreme runoff in spring and summer is slightly larger than that in autumn and winter.In spatial scale,the influence of extreme climate on extreme runoff varies from region to region in Europe.In spring,compared with the historical data,the increase in the maximum daily runoff in the Alps is much larger than that in other regions.Under all three scenarios the increment is about 20%.In summer,the increment of the maximum daily runoff in Germany is the largest,which is about 15% to 20%.In autumn and winter,the impact of extreme climate is relatively low,only under very extreme climate scenarios,the maximum daily runoff in central Europe will increase by 5% to 15%.