Research Of 3d Suspended Sediment Assimilation Models Based On EnKF

In the existing engineering numerical simulation of suspended sediment problems,the verification of suspended sediment concentration is based on the data of limited observation points or the surface suspended sediment distribution data analyzed by satellite images.Such models may cause the calculated results to be quite different from the actual ones.Obtaining more abundant temporal and spatial suspended sediment distribution is one of the effective ways to improve the calculation accuracy of numerical simulation of suspended sediment engineering.An ensemble Kalman filter(En KF)assimilation model of suspended sediment concentration was established to obtain a more accurate spatiotemporal distribution of suspended sediment concentration based on limited observational data.In this paper,based on the FVCOM model,the ensemble Kalman filter assimilation method is used to establish a three-dimensional suspended sediment concentration assimilation model.The calculation shows that the model will inevitably have errors due to the limitation of the number of sets: ○1 sampling error of ensemble members;○2 covariance error of the difference between the suspended sediment concentration of ensemble members and the mean value of suspended sediment concentration of the whole ensemble.By using the empirical(GC)localization method,the influence of the suspended sediment concentration at any spatial point on the whole field is limited;the constant covariance inflation method is used to increase the magnitude of the prediction error covariance.The assimilation effect of the model after this modification has been significantly improved,but its empirical parameters need to be determined through a large number of pre-experiments.Further,the localization function and radius are calculated by the adaptive localization method proposed by ANDERSON or XING et al.It can achieve the purpose of reducing the empirical parameters in the model and improving the assimilation effect.It is called "threedimensional suspended sediment concentration adaptive assimilation model".The three-dimensional suspended sediment concentration adaptive assimilation model was used to simulate the suspended sediment concentration assimilation of different types of sediment sources such as cohesive sand and non-cohesive sand under the conditions of constant and variable water depths.The influence of factors such as initial suspended sediment concentration distribution,flow field velocity,and suspended sediment concentration observation error on the simulation accuracy was analyzed.The research shows that both viscous sand and non-viscous sand can achieve a more ideal three-dimensional suspended sediment concentration assimilation effect;the localization function and radius in the model adopt ANDERSON’s sampling error correction adaptive localization method,and the expansion coefficient adopts a priori The effect of adding a posteriori adaptive expansion method is the best.For non-viscous sand,the assimilation error of SSC in each layer is between1% and 4%,which is similar to the empirical assimilation model.The assimilation error is between 16.3% and 19.2%,which is significantly better than the empirical assimilation model in the area with lower suspended sediment concentration.