5:15 PM - 7:15 PM
[STT40-P03] Evaluation of InSAR atmospheric delay correction in Japan with Japan Meteorological Agency's local analysis (LA) data
Keywords:InSAR, atmospheric delay, correction
Interferometric synthetic aperture radar (InSAR) is a space geodetic technique to measure the displacement of ground with high accuracy and is used for monitoring earthquakes and volcanoes. It is known that InSAR phase signal contains measurement errors which is caused by water vapor in the atmosphere. This error, called as the atmospheric delay, prevents accurate measurements and thus should be adequately corrected. One of the proposed correction methods is to estimate atmospheric delay using a numerical weather model. This method performs subtracting amount of delay that is estimated based on geophysics from InSAR image. Global Analysis Data (GANAL)(Shimada, 1999) and Meso-Scale Model (MSM) (Ozawa et al, 2010; Sailellah et al, 2023), both produced by the Japan Meteorological Agency (JMA), were used in previous studies of InSAR atmospheric delay correction for Japan. In addition to these models, JMA now provides Local Analysis data (LA). LA is a regional model of Japan and has a temporal and spatial resolutions of 1h and 5 km, respectively. The LA model data is obtained with the three-dimensional variational method based on MSM forecast values. It is important for the development of InSAR atmospheric delay correction to show the effect of corrections to models that have not yet been validated. Therefore, this study evaluates the correction effect of atmospheric delay correction by LA.
In this study, atmospheric delay correction by LA was performed on InSAR images generated from ALOS-2 observations. The delay correction by MSM was also performed for comparison. The study areas are set at Yonezawa (Path-Frame:18-2850), Chichibu (Path-Frame:19-2890), Mt.Ontake (Path-Frame:20-2890), Osaka (Path-Frame:20-2920), Kagawa (Path-Frame:21-2930), Izu Peninsula (Path-Frame:126-690). First, InSAR images of these regions were created using RINC (Ozawa et al, 2016). Ionospheric delay correction using the Split-Spectrum method (Gomba et al, 2015) was applied. The SAR images were selected at a time when there was no effect of ground motion, and InSAR images were created in short term pairs so that only the effect of atmospheric delay appeared. Since the period of the prepared LA was from January 2018 to December 2020, SAR images from this period were used. Next, The SAR atmospheric delay is obtained by calculating the line-of-sight component of the atmospheric delay in the zenith direction computing using the physical equation (Bevis et al, 1992) based on the parameters of LA. This calculation is performed for each of the two different periods of the InSAR pair, and the InSAR atmospheric delay is obtained by taking the difference between two periods. Finally, InSAR atmospheric delay distribution was subtracted from InSAR image to produce corrected InSAR image. The same calculation is performed for MSM.
There was a slight difference in correction effect between LA and MSM. In the Izu Peninsula, the mean standard deviation decreased from 1.70 cm to 1.03 cm by LA correction, and to 1.12 cm by MSM correction. In the Kagawa, the mean standard deviation decreased from 2.27 cm to 2.06 cm by LA. On the other hand, it increased to 2.76 cm by MSM. In this study, there were many scenes in which displacement worsened as a result of correction in both LA and MSM. Especially in the Chichibu, the standard deviation increased in all scenes after correction. In addition, phase-elevation correlation and semivariograms were used in the evaluation. In the case of the Izu Peninsula, the correction was effective in almost all scenes, and the LA correction was slightly better than the MSM correction. In the case of the Yonezawa, Chichibu, and Kagawa, where displacement worsened in many scenes, correction by LA was less worsened than that by MSM. On the other hand, in the case of Mt. Ontake, the correction by MSM was less effective than that by LA. In the case of Osaka, there was no significant difference in the correction effect between LA and MSM.
In this study, atmospheric delay correction by LA was performed on InSAR images generated from ALOS-2 observations. The delay correction by MSM was also performed for comparison. The study areas are set at Yonezawa (Path-Frame:18-2850), Chichibu (Path-Frame:19-2890), Mt.Ontake (Path-Frame:20-2890), Osaka (Path-Frame:20-2920), Kagawa (Path-Frame:21-2930), Izu Peninsula (Path-Frame:126-690). First, InSAR images of these regions were created using RINC (Ozawa et al, 2016). Ionospheric delay correction using the Split-Spectrum method (Gomba et al, 2015) was applied. The SAR images were selected at a time when there was no effect of ground motion, and InSAR images were created in short term pairs so that only the effect of atmospheric delay appeared. Since the period of the prepared LA was from January 2018 to December 2020, SAR images from this period were used. Next, The SAR atmospheric delay is obtained by calculating the line-of-sight component of the atmospheric delay in the zenith direction computing using the physical equation (Bevis et al, 1992) based on the parameters of LA. This calculation is performed for each of the two different periods of the InSAR pair, and the InSAR atmospheric delay is obtained by taking the difference between two periods. Finally, InSAR atmospheric delay distribution was subtracted from InSAR image to produce corrected InSAR image. The same calculation is performed for MSM.
There was a slight difference in correction effect between LA and MSM. In the Izu Peninsula, the mean standard deviation decreased from 1.70 cm to 1.03 cm by LA correction, and to 1.12 cm by MSM correction. In the Kagawa, the mean standard deviation decreased from 2.27 cm to 2.06 cm by LA. On the other hand, it increased to 2.76 cm by MSM. In this study, there were many scenes in which displacement worsened as a result of correction in both LA and MSM. Especially in the Chichibu, the standard deviation increased in all scenes after correction. In addition, phase-elevation correlation and semivariograms were used in the evaluation. In the case of the Izu Peninsula, the correction was effective in almost all scenes, and the LA correction was slightly better than the MSM correction. In the case of the Yonezawa, Chichibu, and Kagawa, where displacement worsened in many scenes, correction by LA was less worsened than that by MSM. On the other hand, in the case of Mt. Ontake, the correction by MSM was less effective than that by LA. In the case of Osaka, there was no significant difference in the correction effect between LA and MSM.