Research Of Data Assimilation On Watershed Hydrological Processes

Watershed hydrological process research has important influence on water resources management,agricultural management and natural disaster prevention.However,application of hydrological model often encounters problems such as imperfectly defined parsameters and poor prediction accuracy.Data assimilation technology can integrate different types of real-time observation data,thus effectively improve estimation of model states and parameters.This paper reviews development progress of Ensemble Kalman filter(EnKF).Based on the semi-distributed hydrological model SWAT,a hydrological data assimilation model is established,and computational performance of the model is tested.Effect of multivariate observations(e.g.runoff,soil water and water level)is discussed by numerical experiment.Real experiment is carried out in the Little Washita River Experiment to investigate the influence of runoff and multivatiate and multi-depth soil moisture observation.The detailed contents and main conclusions are as follows:Firstly,a hydrological data assimilation algorithm based on SWAT model is established,and its capability is verified.The results show that EnKF can effectively reduce the uncertainties of parameters and states,and improve prediction of the model.Secondly,data assimilation model is applied to integrate multivariate observation.Based on numerical experiment,influence of observation dataset such as runoff,soil moisture and water level is analyzed.For runoff processes,runoff state and parameter CN2 can be accurately estimated by adding runoff observations.Adding surface soil moisture observation could predict storm runoff,but distort vertical moisture movement and base flow.Complementary groundwater information reflects vertical movement of soil moisture,and improves estimates of soil moisture.Therefore,multivariate observations lead to more accurate model states,and low-cost and available groundwater observation is moderately useful when assimilating soil water.Thirdly,value of remote sensing soil water in hydrological data assimilation is discussed based on a numerical test.Assimilation of surface soil water improves estimation of upper layer soil water.But with soil depth increasing,updating effect weakens gradually.Fourthly,valiues of runoff and multivatiate and multi-depth soil moisture observations are discussed based on Little Washita River Experiement.Adding runoff observation can effectively improve runoff simulation at the basin scale and reduce the deviation of soil water at in rainy season,but its effect is limited in dry season.Assimilating surface soil water observation,including SMOS and in-situ observation,can reduce uncertainty of soil moisture for lower layers,especially for the second soil layer.Assimilation soil moisture of multi-layers can improve runoff estimation to a certain extent.Finally,major research work and contributions are summarized,research directions which needs further improvement are proposed.