Study Of Temporal Variability Of Catchment Water Storage Capacity Based On GR4J Lumped Hydrological Model

Due to the combined effects of climate change and human activities,the catchment characteristics have been changed significantly,posing serious challenges to the original catchment hydrological prediction based on the stationary assumption of the rainfall-runoff relationship.Thus,it is vital to carry out time-varying studies on the catchment characteristics under changing environments to improve the understanding of the process of catchment production and confluence,and boost the accuracy of catchment hydrological prediction.In order to improve the prediction efficiency of hydrological model,this thesis is based on the GR4J(modèle du Génie Rural à 4 paramètres Journalier)hydrological model,focusing on the Catchment Water Storage Capacity(CWSC),an important catchment property,developing studies on the time-varying characteristics and uncertainty estimation of the CWSC under two climatic conditions(i.e.,gradually changed climatic conditions and comparative climatic conditions after a long-term meteorological drought).The main research content and results of the thesis are to:(1)explore the time-varying characteristics of the CWSC under the gradually changed climatic conditions.This thesis is based on 91 catchments in the United States,developing 4time-varying functions(including linear and periodic changes)and multiple time scales(including daily scale,monthly scale,seasonal scale,annual and multi-year scale,etc.)to characterize the possible variation schemes in the CWSC.Then this thesis explores the optimal time-varying scheme and time-varying scale for each catchment according to the model prediction performance.The results of the thesis show that the CWSC of 56.0%(51/91)catchments in the study area has a periodic change characteristic;that of 23.1%(21/91)catchments possesses a linear change while that of 19.8%(18/91)catchments have both periodic and linear changes.The inclusion of the time-varying characteristics in the CWSC,compared with the stationary scheme,achieves much better model performance in the majority of catchments during the calibration and verification periods.Besides,the optimal change scale of 72.5%(66/91)catchments in the study area are between 1 day to 4 months,while that of the remaining 26.4%(24/91)catchments is between 5 months to 24 months.On average,a larger catchment has a longer time-varying scale(5-24 months),indicating that it has a stronger ability to resist climate change and has greater flexibility.(2)explore the response behavior and response time of the CWSC to the long-term meteorological drought.This thesis uses 83 catchments with a long-term meteorological drought in Australia as the research objects,constructs a Bayesian change-point analysis framework to determine the time point that the CWSC may suddenly change,and analyzes the correlation between the response behavior(response time)of CWSC and multiple catchment characteristics as well as climate variables.This thesis indicates that,during the drought,the CWSC of 48.2%(40/83)catchments were significantly increased while that of14.5%(12/83)catchments were significantly decreased.The significant decrease of the CWSC mainly occurred in catchments with silty loam(that has lower adhesion and water holding capacity)as the main soil type and the evergreen broad-leaved forest as the main land use type,while the significant increase of the CWSC occurred in catchments where loam soil(that has strong adhesion and water holding capacity),grassland and forest land are their main types.Different change directions in the CWSC had an opposite impact on the catchment runoff,i.e.,a catchment with an increased value of the CWSC will cause the catchment to produce less water under the same rainfall as before,leads to a decreased runoff coefficient.In addition,different catchments have a different response time to meteorological drought.Catchments with a relatively high proportion of evergreen broad-leaved forests usually have a relatively short response time and are more prone to changes in the CWSC.(3)reduce the estimation uncertainty of the CWSC under the gradually changed climatic conditions by considering the spatial coherence of adjacent catchments through the hierarchical Bayesian framework.The CWSC is constructed as a time-varying regression function based on temporal covariates.Two time-varying features(i.e.,periodic change within the year and inter-annual linear change)are considered in the function,and four scenarios(including scenarios that considering the spatial coherence of periodic and linear changes,that only considering the spatial coherence of periodic change,that only considering the spatial coherence of linear change,and that not considering the spatial coherence,respectively)are set to explore the most effective ways to include the spatial coherence between adjacent catchments.Finally,the proposed method was used in four adjacent catchments in the United States.The results show that the utilize of the spatial coherence of catchments through the hierarchical Bayesian framework effectively reduces the estimation uncertainty of the CWSC.Furthermore,the inclusion of the spatial coherence of the amplitude of periodic change was the most effective approach to reduce the estimation uncertainty of the CWSC,which indicates that the similarity of periodic changes between adjacent catchments is an effective input that constrains the uncertainty of the CWSC.(4)reduce the estimation uncertainty of the CWSC under the comparative climatic conditions by considering the spatial coherence of adjacent catchments through the hierarchical Bayesian framework.A drought period identification method is used to determine the drought period and the humid period in the historical sequence.Through the hierarchical Bayesian framework,the periodic changes of CWSC,two cross-verification schemes(including one scheme that parameters were calibrated in humid period and verified in the drought period,and another scheme that parameters were calibrated in drought period and verified in humid period),and four spatial-coherent scenarios(including scenarios that considering the spatial coherence of amplitude and frequency,that only considering the spatial coherence of amplitude,that only considering the spatial coherence of frequency,that not considering spatial coherence),and one scenario that both parameters are time-invariant,are developed to evaluate the uncertainty estimation of the CWSC based on the model prediction performance.Finally,the proposed method is applied to three spatially adjacent catchments in southeastern Australia that have experienced long-term drought.Our results indicate that,compared with the time-varying CWSC scheme,the inclusion of the periodic time-varying characteristics of CWSC effectively improves the model prediction accuracy,but also increases the estimation uncertainty of the results;the inclusion of the spatial coherence of both amplitude and frequency,effectively reduces the estimation uncertainty of the CWSC and improves the accuracy of the prediction results under the comparative climatic conditions.