2:15 PM - 2:30 PM
[SGD01-03] Toward crustal deformation transform combining InSAR time series analysis: Evaluation on extrapolation of InSAR in time
Keywords:Crustal deformation tansformation, POS2JGD, GEONET, InSAR time series analysis
Recently, satellite positioning techniques such as QZSS Centimeter-Level Augmentation Service (CLAS) and precise point positioning (PPP) have been utilized to various fields including autonomous driving. For these application fields, the secular crustal deformation transformation is a key enabler to ensure that positioning solution is consistent with digital maps which are based on the reference epoch (1997-04-01 or 2011-05-24) of Japanese Geodetic Datum 2011. Thus on 2020-03-31, Geospatial Information Authority of Japan published a framework to support the transformation for the positioning applications, POS2JGD (https://positions.gsi.go.jp/cdcs/), to provide gridded models for transform between observation and the reference epochs.
While the present gridded model is created from national GNSS CORS with an average spacing of 20 km, higher spatial resolution will inevitably be required to the gridded model as further development and application of positioning techniques is expected in the near future. Thus Yamashita et al. (2022) proposed a modeling approach to combine GNSS CORS and InSAR Time Series Analysis (InSAR TSA), and implied that it could contribute to improve model spatial resolution. However, Yamashita et al. (2022) evaluated the deformation events after 2015 when Advanced Land Observing Satellite-2 (ALOS-2) started data acquisition. A main challenge to implement the approach into operational gridded model is temporal extrapolation of InSAR TSA result, which can lead to accuracy degradation of transformation into the reference epoch. In this study, we evaluate performance of the approach proposed by Yamashita et al. (2022) in the case that InSAR TSA result is extrapolated in time.
This study focuses on Chiba region in Japan, where rapid subsidence is locally detected by geodetic leveling. We consider displacement between observation epoch (January 2022) and reference epoch (January 2012) derived from geodetic leveling as true value, and estimate transformation error of gridded model derived from GNSS CORS (POS2JGD model), and gridded model derived from GNSS CORS and InSAR TSA (CORS+InSAR model).
In the case of CORS+InSAR model, the displacement is calculated by following 4 steps: (1) apply Kriging method to create the gridded model (5 km spacing) using GNSS CORS just as POS2JGD model, (2) apply InSAR TSA (Schmidt et al., 2003) to SAR images acquired by ALOS-2 from 2015 to 2022, and derive the LOS velocity in ascending/descending orbits to create a gridded model (50 m spacing) of the quasi-vertical velocity, (3) linearly extrapolate the velocity in time, and (4) take the average of two gridded models using weight coefficients based on Kriging variance.
Fig.1a, b, and c shows leveling displacement between observation and reference epochs, transformation error using POS2JGD model, and the error using CORS+InSAR model, respectively. Significant subsidence occurs in longitude of 140.2 – 140.4 deg. and latitude of 35.2 – 35.8 deg. (Fig.1a). POS2JGD model can capture long-wavelength component, but cannot account for short-wavelength component in the area enclosed by red circle on Fig.1. CORS+InSAR model captures short-wavelength component more successfully. While RMS error for POS2JGD model is 3.40 cm, the error for CORS+InSAR model is 2.67 cm. Notably the error of 97 % benchmarks is improved in the area where the subsidence velocity is larger than 1 cm/year. On the other hand, larger error can be seen in the area with smaller velocity shown by blue circle on Fig.1.
Our result shows applying simple linear extrapolation of InSAR TSA velocity still contributes to improvement of model spatial resolution in subsidence area with large velocity.
Reference
Schmidt et al.(2003): Journal of Geophysical Research, 108(B9), 2416.
Yamashita et al.(2022): Towards secular deformation transformation combining national GNSS CORS and InSAR time series analysis, Geod. Soc. Japan.
While the present gridded model is created from national GNSS CORS with an average spacing of 20 km, higher spatial resolution will inevitably be required to the gridded model as further development and application of positioning techniques is expected in the near future. Thus Yamashita et al. (2022) proposed a modeling approach to combine GNSS CORS and InSAR Time Series Analysis (InSAR TSA), and implied that it could contribute to improve model spatial resolution. However, Yamashita et al. (2022) evaluated the deformation events after 2015 when Advanced Land Observing Satellite-2 (ALOS-2) started data acquisition. A main challenge to implement the approach into operational gridded model is temporal extrapolation of InSAR TSA result, which can lead to accuracy degradation of transformation into the reference epoch. In this study, we evaluate performance of the approach proposed by Yamashita et al. (2022) in the case that InSAR TSA result is extrapolated in time.
This study focuses on Chiba region in Japan, where rapid subsidence is locally detected by geodetic leveling. We consider displacement between observation epoch (January 2022) and reference epoch (January 2012) derived from geodetic leveling as true value, and estimate transformation error of gridded model derived from GNSS CORS (POS2JGD model), and gridded model derived from GNSS CORS and InSAR TSA (CORS+InSAR model).
In the case of CORS+InSAR model, the displacement is calculated by following 4 steps: (1) apply Kriging method to create the gridded model (5 km spacing) using GNSS CORS just as POS2JGD model, (2) apply InSAR TSA (Schmidt et al., 2003) to SAR images acquired by ALOS-2 from 2015 to 2022, and derive the LOS velocity in ascending/descending orbits to create a gridded model (50 m spacing) of the quasi-vertical velocity, (3) linearly extrapolate the velocity in time, and (4) take the average of two gridded models using weight coefficients based on Kriging variance.
Fig.1a, b, and c shows leveling displacement between observation and reference epochs, transformation error using POS2JGD model, and the error using CORS+InSAR model, respectively. Significant subsidence occurs in longitude of 140.2 – 140.4 deg. and latitude of 35.2 – 35.8 deg. (Fig.1a). POS2JGD model can capture long-wavelength component, but cannot account for short-wavelength component in the area enclosed by red circle on Fig.1. CORS+InSAR model captures short-wavelength component more successfully. While RMS error for POS2JGD model is 3.40 cm, the error for CORS+InSAR model is 2.67 cm. Notably the error of 97 % benchmarks is improved in the area where the subsidence velocity is larger than 1 cm/year. On the other hand, larger error can be seen in the area with smaller velocity shown by blue circle on Fig.1.
Our result shows applying simple linear extrapolation of InSAR TSA velocity still contributes to improvement of model spatial resolution in subsidence area with large velocity.
Reference
Schmidt et al.(2003): Journal of Geophysical Research, 108(B9), 2416.
Yamashita et al.(2022): Towards secular deformation transformation combining national GNSS CORS and InSAR time series analysis, Geod. Soc. Japan.