Study Of Temporal Variability Of Hydrological Model Parameters By Multivariate Data Assimilation And Functional Form

Hydrological model parameters are essential for model simulation,which may vary with time owing to climatic variations and human activities.As a result,the implementation of stationary parameters may lead to inaccurate streamflow simulation.Therefore,the identification of time-varying hydrological model parameters can enhance the model simulation,and provide technical support for hydrological response research under changing environment.Data assimilation technique has been applied to the identification of time-varying hydrological model parameters,which can automatically adjust hydrological model parameters according to changing conditions.Based on summarizing the literature of time-varying hydrological model parameters,the multivariate data assimilation technique was used first to identify the time-varying parameters of a two-parameter monthly water balance model,following the construction of the linear functions of parameters and climatic factors.The major contents and conclusions are as follows:(1)The streamflow and actual evapotranspiration(ET)combined data assimilation technique was proposed to estimate the time-varying parameters of a two-parameter monthly water balance model,and the ensemble Kalman filter(En KF)method was used to assimilate observations into the hydrological model.Four data assimilation schemes were designed and applied in a synthetic experiment and 173 catchments in USA,including one scheme that solely assimilating streamflow and three schemes that jointly assimilating streamflow and ET.The four schemes were compared in terms of deterministic and probabilistic model performances,and model parameter correlations.The results indicated that the time-varying model parameter estimation framework using the En KF method was able to identify the temporal variation of the model parameters,and the mean absolute relative error(MARE)of parameter C and SC could be within 0.08 and 0.02,respectively.The prediction of streamflow and ET can be improved by assimilating ET observations into the model,the Nash-Sutcliffe efficiency coefficient(NSE)could be increased by 0.1 and 0.7,respectively,and the correlations between parameters can be also weakened.The deterministic model simulation performances were similar among the three joint assimilation schemes,while the ensemble predictions were more reliable in the scheme that parameters were updated in one step,in which the reliability(?)was 0.94.In conclusion,the joint assimilation of streamflow and ET to update both parameters in one step,outperformed the other schemes in parameter estimation and model performance.(2)The establishment of functional forms for model parameters and the model simulation under changing environment were investigated.The En KF technique was used first to identify the time-varying parameters,following the construction of the linear functions of parameters and precipitation,potential evapotranspiration and temperature via the correlation analysis between model parameters and climatic factors.The case study was carried out in 173 catchments in USA,in which three catchments were selected for concrete analysis,and the research data were divided into drought period and wet period according to precipitation and streamflow coefficient for further analysis.The results indicated that the accuracy of streamflow simulation was improved when considering the model parameters as time-varying,especially in peak flows,and the NSE of streamflow can be increased by 0.58 in the model that two parameters were treated as time-varying.The change of streamflow was underestimated when considering the model parameters as constant,with the relative error being up to-29.4%,while its value was reduced to 7.5% when considering the two model parameters as time-varying.