5:15 PM - 7:15 PM
[MTT37-P02] Preliminary analysis of common-mode errors at GNSS stations operated by SoftBank
Keywords:GNSS, Common-mode errors, Noise reduction
Common-mode errors (CMEs) are one of the major components recorded in data from the Global Navigation Satellite System (GNSS). They appear at stations across an entire network in a similar manner (Wdowinski et al., 1997) and often obscure small tectonic signals in GNSS time series. Thus, removing CMEs is essential for extracting small displacements from GNSS data, as many previous studies show (e.g., Mavrommatis et al., 2014; Nishimura et al., 2013; Ueda et al., 2024). Although several studies have explored the characteristics of CMEs in various regions (e.g., Kumar et al., 2020; Tian & Shen, 2016), the number of studies focusing on CMEs specific to Japan is limited (Tabei & Amin, 2002). However, investigating the characteristics of CMEs in a regional network is important because topographic, meteorological, and climatic features should produce regional variations in CMEs. Therefore, in this study, we investigate the characteristics of CMEs in Japan using the ultra-dense GNSS network operated by SoftBank Corp. The advantage of using the network is that it reduces the effect of the local motion at individual GNSS stations and allows us to focus on regional CMEs, which appear in a limited area in the whole GNSS network.
We utilized daily positions at 3,175 GNSS stations from July 1, 2022, to June 30, 2023. We analyzed the GNSS data using GipsyX software (version 2.2) with precise point positioning with ambiguity resolution (Bertiger et al., 2020). The estimated positions were aligned into the International Terrestrial Reference Frame 2020. In calculating CMEs, we first calculated 91-day moving medians and subtracted them as a trend from the raw time series. We also removed outliers whose absolute value was over twice the median absolute deviation of the entire time series. Finally, we stacked the residual time series at stations within an area of 0.5°×0.5°. We repeated this stacking process by sliding the spatial window 0.2° in longitudinal and latitudinal directions and obtained CME time series in each area.
We first focused on the short-term behavior of CMEs. We checked the daily spatial distribution of CMEs and found large negative and regional common-mode signals at the north and vertical components in southern Kyushu on September 5, 2022. On that day, typhoon Hinnamnor was located off the west coast of Kyushu. We speculate that these common-mode signals are related to the typhoon, especially the common-mode signals in the vertical component, which were probably caused by water load due to precipitation (Zhan et al., 2021). We also identified the large eastward and downward common-mode signals on December 18, 2022, on the coast side of the Sea of Japan, whose timing corresponds to significant snowfall (Japan Meteorological Agency, 2025). We suspect these common-mode signals have a relationship with the snow, such as loading by the accumulated snow (Heki, 2001) and pumping of groundwater for snow melting (Sato et al., 2003).
Then, we checked the similarity of the CME time series between the areas. We calculated correlation coefficients between the 1-year CME time series for each area and compared them with the distance between the center of the two areas. We found that the correlations between areas exponentially decay along distance and that correlation coefficients retain 0.7 for horizontal components up to a distance of 1,000 km. On the other hand, the correlations of the vertical component decrease against a distance more rapidly than horizontal components, which reflects the regionality of the vertical component of GNSS data.
Acknowledgments
We appreciate SoftBank Corp. and ALES Corp. for providing raw GNSS data through the framework of the “Consortium to utilize the SoftBank original reference sites for Earth and Space Science”.
We utilized daily positions at 3,175 GNSS stations from July 1, 2022, to June 30, 2023. We analyzed the GNSS data using GipsyX software (version 2.2) with precise point positioning with ambiguity resolution (Bertiger et al., 2020). The estimated positions were aligned into the International Terrestrial Reference Frame 2020. In calculating CMEs, we first calculated 91-day moving medians and subtracted them as a trend from the raw time series. We also removed outliers whose absolute value was over twice the median absolute deviation of the entire time series. Finally, we stacked the residual time series at stations within an area of 0.5°×0.5°. We repeated this stacking process by sliding the spatial window 0.2° in longitudinal and latitudinal directions and obtained CME time series in each area.
We first focused on the short-term behavior of CMEs. We checked the daily spatial distribution of CMEs and found large negative and regional common-mode signals at the north and vertical components in southern Kyushu on September 5, 2022. On that day, typhoon Hinnamnor was located off the west coast of Kyushu. We speculate that these common-mode signals are related to the typhoon, especially the common-mode signals in the vertical component, which were probably caused by water load due to precipitation (Zhan et al., 2021). We also identified the large eastward and downward common-mode signals on December 18, 2022, on the coast side of the Sea of Japan, whose timing corresponds to significant snowfall (Japan Meteorological Agency, 2025). We suspect these common-mode signals have a relationship with the snow, such as loading by the accumulated snow (Heki, 2001) and pumping of groundwater for snow melting (Sato et al., 2003).
Then, we checked the similarity of the CME time series between the areas. We calculated correlation coefficients between the 1-year CME time series for each area and compared them with the distance between the center of the two areas. We found that the correlations between areas exponentially decay along distance and that correlation coefficients retain 0.7 for horizontal components up to a distance of 1,000 km. On the other hand, the correlations of the vertical component decrease against a distance more rapidly than horizontal components, which reflects the regionality of the vertical component of GNSS data.
Acknowledgments
We appreciate SoftBank Corp. and ALES Corp. for providing raw GNSS data through the framework of the “Consortium to utilize the SoftBank original reference sites for Earth and Space Science”.