River flow and stage estimation with missing observation data using Multi Imputation Particle Filter (MIPF) method

Abstract

An advanced knowledge of the river condition helps for better source management. This information can be gathered via estimation using DA methods. The DA methods blend the system model with the observation data to obtain the estimated river flow and stage. However, the observation data may contain some missing data due to the hardware power limitations, unreliable channel, sensor failure and etc. This problem limits the ability of the standard method such as EKF, EnKF and PF. The Multi Imputation Particle Filter (MIPF) able to deal with this problem since it allows for new input data to replace the missing data. The result shows that the performance of the river flow and stage estimation is depending on the number of particles and imputation used. The performance is evaluated by comparing the estimated velocity obtained using the estimated flow and stage, with the measured velocity. The result shows that higher number of particles and imputation ensure better estimation result.