Implementing the Modified Phase-Space Tresholding despiking algorithm on the Acoustic Tomographic data to improve the measurement accuracy

The Acoustic Tomography (AT) systems are one of the remote sensing technologies to measure flow characteristics in aquatic environments. The only difference between various types of AT systems is the frequency, so that for the ocean class the frequencies ranged from 20 Hz to 1 kHz, for the sea class ranged from 1 kHz to 10 kHz and for the river, dam or lake class ranged from 10 kHz to 60 kHz. The output of this type of system also contains outlier data, just like other acoustic measurement devices. Failure to pay attention to the issue of detection and replacement of outliers causes a decrease in measurement accuracy as well as an inappropriate effect on flow modeling and data assimilation of models. The review of the previous studies shows that the removal of outlier data in this type of system is generally done by the simple method of standard deviation. In the last study, the analysis of the AT data was performed using the three-dimensional Rousseeuw Phase-Space thresholding (3D-RPS) algorithm and compared with the standard deviation method, which indicated its better performance than the simple standard deviation method.

In this study, the Modified Phase-Space Thresholding algorithm (mPST) is applied to AT data and its results are compared with the 3D-RPS method. For this purpose, the data collected in Gono River, Japan has been used, and outliers were identified and replaced with appropriate values using the two mentioned algorithms and five outlier replacement methods. Then, the results were validated with Rating-Curve (RC) data obtained from the hydrometric station of the study area as reference data. Furthermore, in another experiment in the Anzali port, using the mPST algorithm, the analysis of the AT data was performed and the volume of water entering the Caspian Sea was measured.

The results of this research showed that the use of the mPST algorithm has increased the accuracy of the flow measurement in such a way that the relative difference of the flow rate measured with the AT system after applying this method in combination with the average signal replacement methods and The last valid data with flow rate measured by RC method is limited in the range of -15% to +20%. While by applying the 3D-RPS algorithm, this value is in the range of -20% to +30%. The results of the sound recording test of Anzali port also recorded the volume of 200 m3/s flowing to the Caspian Sea.

Since AT data used in the data assimilation of ocean, sea, and river models, it is important to analyze the outlier data of the system and replace it with appropriate data. In this study, an attempt was made to improve the accuracy of the measurement by applying the mPST algorithm to the AT data. The comparison of the output results of different outlier mining methods showed that the mPST method combined with the average signal replacement methods and the last valid data before the detected outlier data has the best performance. As a result, the mPST method is recommended to be applied to AT data.