12:00 PM - 12:15 PM
[SGD02-12] Mitigation of biases due to instrumental errors in GNSS-A observation
Keywords:GNSS-A, seafloor crustal deformation, acoustic measurement, waveform clustering
By the constant efforts for developing the methods and techniques of GNSS-A in more than two decades, we can now detect crustal deformation on the seafloor in the precision of less than a centimeter per year. The effects of sound speed variation, which potentially cause large positioning errors in GNSS-A observations, can be suppressed by directly estimating its spatial and temporal changes from the collected acoustic data (e.g., Watanabe et al., 2023, J. Geod.). In addition, some instrument biases can be reduced or directly estimated and corrected by collecting data on a geometrically balanced track. For example, the measurement error of the horizontal component (more precisely, the roll- and pitch-axis directions of the ship-born gyro) of the relative position between the GNSS antenna and the acoustic transducer (called "ATD offset") can be reduced or directly estimated by performing a set of observations with changing the ship's orientation. Errors from time synchronization in the measurement system, such as the time bias of the gyro data, can be corrected by correlating its time series with the GNSS positions.
The non-Gaussian noise errors, which may not be sufficiently corrected as described above, mainly come from modeling errors of sound speed perturbation including sound speed profile, GNSS positioning errors, measurement errors of the vertical component (the gyro's heading-axis) of the ATD offset, and acoustic signal reading errors. Because the resolution of the sound speed perturbation model depends on the data distribution, there are limitations to its improvement in the actual cases where the observation setting requires measuring from the sea-surface to the seafloor. For GNSS, if the positioning error is white noise, the effect can be suppressed as long as a sufficient number of acoustic data is obtained. However, the seafloor positioning solution is directly affected by GNSS's random walk-like errors that have time correlation on the scale of several hours to days. The measurement error of the ATD offset directly depends on the reliability of terrestrial surveys at docks. As for the acoustic signal reading error, it is necessary to correct the bias by taking into account the characteristics of the acoustic devices, similar to the GNSS phase center variation (PCV). In this study, we investigate the error characteristics of the latter two equipment-derived biases and develop methods for the error reduction.
Regarding the ATD offsets, terrestrial surveys were conducted at dock to determine the relative position from the GNSS antenna to the transducer, in each time when an acoustic transducer was installed on a survey vessel (Kawai et al., 2009; Ujihara and Narita, 2012; Akiyama et al., 2013; Akiyama and Yokota, 2014; Yoshida et al., 2021; Yoshida, 2022). However, the specific survey configurations differed from vessel to vessel. Thus, we compiled the methods and results, including the efficiency of the survey. We expect that this information is useful in indicating the necessity of re-surveying and suggesting an effective survey configuration for the future implementation.
For the characteristics of acoustic devices, we developed the Acoustic Ambiguity Reduction (AAR) method using actual data and water-tank experiment data (Yokota et al., 2024, EPS; Nagae et al., 2024, IEICE). We showed that the AAR method can improve the accuracy of crustal deformation solutions (e.g., Watanabe et al., 2024, AGU24). However, the current implementation of the AAR method still requires manual adjustments by experienced analysts, making it difficult to process the data routinely. Therefore, it is necessary to develop the automated AAR method. In this study, we developed an automatic classification method for the correlation waveforms of acoustic signals.
The non-Gaussian noise errors, which may not be sufficiently corrected as described above, mainly come from modeling errors of sound speed perturbation including sound speed profile, GNSS positioning errors, measurement errors of the vertical component (the gyro's heading-axis) of the ATD offset, and acoustic signal reading errors. Because the resolution of the sound speed perturbation model depends on the data distribution, there are limitations to its improvement in the actual cases where the observation setting requires measuring from the sea-surface to the seafloor. For GNSS, if the positioning error is white noise, the effect can be suppressed as long as a sufficient number of acoustic data is obtained. However, the seafloor positioning solution is directly affected by GNSS's random walk-like errors that have time correlation on the scale of several hours to days. The measurement error of the ATD offset directly depends on the reliability of terrestrial surveys at docks. As for the acoustic signal reading error, it is necessary to correct the bias by taking into account the characteristics of the acoustic devices, similar to the GNSS phase center variation (PCV). In this study, we investigate the error characteristics of the latter two equipment-derived biases and develop methods for the error reduction.
Regarding the ATD offsets, terrestrial surveys were conducted at dock to determine the relative position from the GNSS antenna to the transducer, in each time when an acoustic transducer was installed on a survey vessel (Kawai et al., 2009; Ujihara and Narita, 2012; Akiyama et al., 2013; Akiyama and Yokota, 2014; Yoshida et al., 2021; Yoshida, 2022). However, the specific survey configurations differed from vessel to vessel. Thus, we compiled the methods and results, including the efficiency of the survey. We expect that this information is useful in indicating the necessity of re-surveying and suggesting an effective survey configuration for the future implementation.
For the characteristics of acoustic devices, we developed the Acoustic Ambiguity Reduction (AAR) method using actual data and water-tank experiment data (Yokota et al., 2024, EPS; Nagae et al., 2024, IEICE). We showed that the AAR method can improve the accuracy of crustal deformation solutions (e.g., Watanabe et al., 2024, AGU24). However, the current implementation of the AAR method still requires manual adjustments by experienced analysts, making it difficult to process the data routinely. Therefore, it is necessary to develop the automated AAR method. In this study, we developed an automatic classification method for the correlation waveforms of acoustic signals.