[STT50-08] Mitigation of ionospheric noise in azimuth pixel offset based on the split-spectrum method
The pixel offset (PO) method has been utilized as a powerful tool to measure large ground movements due to earthquakes and volcanic activities, which is difficult to detect by InSAR. However, data acquired by low-frequency spaceborne SAR such as ALOS-2 are severely affected by ionosphere, which often results in periodic noise patterns called as “azimuth streaks” in the azimuth direction.
Over the past decade, remarkable advances have been made in the theoretical studies on the relation between azimuth pixel offset and ionospheric phase delay, and the split-spectrum method has been developed for ionospheric phase screen in InSAR image. Combining these two theories enables us to separate the ionospheric noise from ground displacements in the azimuth direction and to mitigate the ionospheric noise. Here we propose a new method for mitigation of the azimuth streaks, and show the effects through the application to actual SAR data.
An outline of the implementation for mitigating ionospheric azimuth shift is as follows. First, the ionospheric phase delay is obtained based on the split-spectrum method. Secondly, the approximate quadric surface for the ionospheric phase delay is estimated to be subtracted from the ionospheric phase delay. Subtracting the quadric surface corresponds to the partial elimination of ionospheric azimuth shift after coregistration. In this study, we adopt affine transformation as the coregistration method. Next, the Gaussian filter as the low-pass filter is applied to the residuals between the quadric surface and the ionospheric phase delay. The reason to apply the filter is because high frequency noise can have a negative influence on an evaluation of the azimuth gradient of the ionospheric phase delay later. Here we should be careful that the signal component of the ionospheric phase delay can be smoothed too much to be evaluated properly if the standard deviation (S.D.) of Gaussian filter is larger than the spatial scale of azimuth streaks. Subsequently, the azimuth gradient of the filtered residuals is calculated. Finally, the correction amount is evaluated using the estimated gradient to be subtracted from the azimuth pixel offset.
For the purpose of evaluating the performance of the method for mitigating azimuth streaks, we apply the method to two pairs of ALOS-2/PALSAR-2 stripmap images in Japan area which are severely affected by ionosphere. In the first case, we make Hokkaido region between 2015 and 2016 with little ground movements our target, and in the second case, we focus on the area with large ground movements induced by the 2016 Kumamoto earthquake. In the first case, while the S.D. of the azimuth offset before the correction is 87.2 cm, the value after the correction is 30.2 cm, indicating that the ionospheric azimuth shift can be considerably compensated through the correction. It is noteworthy that the S.D. after the correction is comparable to the theoretical measurement accuracy of azimuth pixel offset (~30 cm). Although we also exploit a uniform shift for coregistration, the S.D. after the correction is rather large compared to the value before correcting. In the second case, considering that the image includes significant ground movements, we examine the residuals between the azimuth offset converted from GNSS CORS displacement and the azimuth offset obtained by the PO method, and the S.D. of the residuals. The maximum residuals before and after the correction are 84.5 cm and 44.5 cm, respectively. Moreover, the S.D. of the residuals before and after the correction are 22.7 cm and 14.9 cm, respectively: the deviation gets smaller through the correction. Therefore, although the measurement accuracy after the correction is not better than that of pairs of data that are not affected by ionosphere, the corrected data are expected to be still useful as the supplements for derivation of 3D displacement from pixel offsets and finite rectangular fault models.
Over the past decade, remarkable advances have been made in the theoretical studies on the relation between azimuth pixel offset and ionospheric phase delay, and the split-spectrum method has been developed for ionospheric phase screen in InSAR image. Combining these two theories enables us to separate the ionospheric noise from ground displacements in the azimuth direction and to mitigate the ionospheric noise. Here we propose a new method for mitigation of the azimuth streaks, and show the effects through the application to actual SAR data.
An outline of the implementation for mitigating ionospheric azimuth shift is as follows. First, the ionospheric phase delay is obtained based on the split-spectrum method. Secondly, the approximate quadric surface for the ionospheric phase delay is estimated to be subtracted from the ionospheric phase delay. Subtracting the quadric surface corresponds to the partial elimination of ionospheric azimuth shift after coregistration. In this study, we adopt affine transformation as the coregistration method. Next, the Gaussian filter as the low-pass filter is applied to the residuals between the quadric surface and the ionospheric phase delay. The reason to apply the filter is because high frequency noise can have a negative influence on an evaluation of the azimuth gradient of the ionospheric phase delay later. Here we should be careful that the signal component of the ionospheric phase delay can be smoothed too much to be evaluated properly if the standard deviation (S.D.) of Gaussian filter is larger than the spatial scale of azimuth streaks. Subsequently, the azimuth gradient of the filtered residuals is calculated. Finally, the correction amount is evaluated using the estimated gradient to be subtracted from the azimuth pixel offset.
For the purpose of evaluating the performance of the method for mitigating azimuth streaks, we apply the method to two pairs of ALOS-2/PALSAR-2 stripmap images in Japan area which are severely affected by ionosphere. In the first case, we make Hokkaido region between 2015 and 2016 with little ground movements our target, and in the second case, we focus on the area with large ground movements induced by the 2016 Kumamoto earthquake. In the first case, while the S.D. of the azimuth offset before the correction is 87.2 cm, the value after the correction is 30.2 cm, indicating that the ionospheric azimuth shift can be considerably compensated through the correction. It is noteworthy that the S.D. after the correction is comparable to the theoretical measurement accuracy of azimuth pixel offset (~30 cm). Although we also exploit a uniform shift for coregistration, the S.D. after the correction is rather large compared to the value before correcting. In the second case, considering that the image includes significant ground movements, we examine the residuals between the azimuth offset converted from GNSS CORS displacement and the azimuth offset obtained by the PO method, and the S.D. of the residuals. The maximum residuals before and after the correction are 84.5 cm and 44.5 cm, respectively. Moreover, the S.D. of the residuals before and after the correction are 22.7 cm and 14.9 cm, respectively: the deviation gets smaller through the correction. Therefore, although the measurement accuracy after the correction is not better than that of pairs of data that are not affected by ionosphere, the corrected data are expected to be still useful as the supplements for derivation of 3D displacement from pixel offsets and finite rectangular fault models.