[S-TT35] Synthetic Aperture Radar and its application

[STT35-02] Evaluation of interferometric SAR coherence in view of high frequency observation by next generation L-band SAR satellite

*Keisho Ito¹, Takeshi Motohka¹, Takeo Tadono¹ (1.Japan Aerospace Exploration Agency)

Keywords:Synthetic aperture radar (SAR), coherence, ALOS-2/PALSAR-2, ALOS-4/PALSAR-3, land cover, landslide

Advanced Land Observing Satellite-4 (ALOS-4), planned as the successor to ALOS-2, is expected to significantly enhance observation frequency along with an expanded observation width. Particularly in Japan, observations are anticipated as frequent as once every two weeks, promising early detection of changes and improved accuracy in time-series analysis due to abundant archive data. Furthermore, the global operation of L-band SAR satellites capable of high-frequency observations, such as the NASA-ISRO SAR (NISAR) mission with a 12-day global land observation cycle, is drawing worldwide attention for the use of high-frequency L-band SAR data. In this context, interferometric SAR coherence (correlation between two SAR observation signals, indicating the degree of interferometric capability), whose primary degradation factor is the interval between interferometric pairs, is expected to benefit various applications due to reduced degradation from high-frequency observations. This is especially anticipated to be a critical information source in improving detection accuracy in fields such as damage assessment and land cover change detection. However, to utilize these applications, it is necessary to investigate the impact of the interval of interferometric pairs and targeted land cover on coherence and develop algorithms based on this understanding. In this study, we investigated the temporal changes in coherence according to the interval of interferometric pairs and land cover using high-frequency L-band SAR observation data of the same level as the next-generation SAR satellites like ALOS-4, exploring its effectiveness and potential for improved accuracy in change detection.
In this research, we used ALOS-2 PALSAR-2 L1.1 Single Look Complex (SLC) data in Stripmap mode with a 3 m resolution for coherence analysis. This data was observed in the Iburi and Ishikari regions of Hokkaido over a period from January 15, 2018, to April 18, 2019. These data were observed at a frequency of once every two weeks, which is identical to the anticipated observation frequency for ALOS-4, although there were exceptions during certain periods. We also used the 'JAXA High Resolution Land-Use and Land-Cover Map (JHR LULC Map)' ver. 21.11 with a 10 m resolution to investigate coherence trends according to nine types of land cover. Similarly, we examined the coherence response to landslides caused by the 2018 Eastern Iburi earthquake and snow cover. We found that coherence significantly decreases for intervals longer than 28 days compared to 14-day intervals across many land covers, including vegetated areas, indicating that high-frequency observations can reduce temporal degradation in vegetated areas and significantly improve coherence. The extent of coherence reduction varied depending on land cover and snow cover, suggesting that data from smaller intervals like 14 days can be a valuable indicator for land cover classification and surface change detection. However, coherence varied greatly depending on the pair used and the observation period, and when detecting landslides using coherence, the results varied significantly depending on the data used. By averaging multiple coherences in the time series as a pre-earthquake dataset, detection accuracy could be substantially improved. This suggests that accumulating archives over time is desirable for the effective use of coherence.