Geospatial Information Authority of Japan (GSI) has nationwide monitored surface changes by conducting interferometric SAR (InSAR) analysis. L-band InSAR is advantageous for measuring surface displacements in Japan because high coherence can be achieved even in mountainous areas with vegetation and steep slope. This is why Advanced Land Observing Satellite-2 (ALOS-2) satellite, whose operation started in 2014 by Japan Aerospace Exploration Agency (JAXA), is suitable for our ground deformation monitoring. We have processed the ALOS-2 data to monitor crustal deformation in Japan, and have detected surface displacements associated with earthquakes, volcanic activities, and land subsidence.
Volcano monitoring is one of our main missions. There are 111 active volcanoes in Japan, and hazardous eruptions have repeatedly occurred historically; such as the 2014 Ontake eruption in Japan that led to 58 casualties. Therefore, it is desired to continuously monitor volcanic activities and detect anomalous signals to mitigate volcanic hazard. InSAR is a powerful tool to monitor volcanic activity, while it often suffers from atmosphere-related noises etc. and likely prevents from detecting anomalies. It is crucial for volcano observation to overcome the shortcoming of conventional InSAR–based monitoring.
In order to observe spatio-temporal development of surface displacements that slowly proceed over medium to long term on volcanoes, GSI has developed a software for InSAR time series analysis (GSITSA) in which small baseline subset (SBAS) algorithm is implemented. The analysis can improve a signal-to-noise ratio by substantially reducing atmospheric noises, and thus enables us to monitor temporal evolution of slow deformation. With the background, we have conducted the InSAR time series analysis using more than 20 SAR images acquired for the period from 2014 to 2021. The time series analyses have successfully unveiled surface displacements with a few cm/year or slower at several active volcanoes, which can be hardly identified by conventional InSAR analysis. The striking point is that the time series analysis using L-band data of ALOS-2, unlike C-band, can map surface displacement over volcanoes due to high coherence even in mountainous areas. The information on spatio-temporal variation of surface displacements can be a good indicator to know a degree of volcanic activity, and thus the ALOS-2 InSAR time series analysis is about to be one of major tools for evaluation of volcanic activity in Japan. In this presentation, we mainly report our volcano monitoring by ALOS-2 InSAR time series analysis, and further demonstrates effectiveness of L-band in comparison to C-band.

The lack of regular, area-wide observations of snow mass (snow water equivalent, SWE) is a main gap in monitoring of the global cryosphere. Repeat-pass differential SAR interferometry (RP-InSAR) offers a well-defined approach for mapping SWE at high spatial resolution by measuring the path delay of the radar signal propagating through a dry snowpack. In C-band and lower radar frequencies the absorption and scattering losses in dry seasonal snow are small so that the change in the InSAR phase delay of the signal reflected from the snow/ground interface is directly related to the snow mass that accumulated during the repeat-pass time span. A critical issue for routine application of RP-InSAR is the temporal decorrelation caused by changes in the complex backscatter signal. In C-band comparatively moderate snowfall amounts may already cause complete decorrelation whereas this effect is of less concern in L-band. The use of L-band RP-InSAR for SWE mapping has by now been limited to case studies because of the lack of regular repeat observations over snow-covered land surfaces. In anticipation of the enhanced repeat observation capabilities of upcoming L-band SAR systems, such as ROSE-L, we conducted field campaigns in two Alpine test sites in order to consolidate the SP-InSAR retrieval tools and evaluate the product performance. The use of C-band RP-InSAR for SWE retrievals was as well studied. Among the topics addressed are the effects of snowfall intensity on temporal decorrelation, the impacts of topography and land cover type on the RP-InSAR phase and resulting SWE, and possible effects of snow structural properties on the observed signal.
In winter 2019-2020 the studies on SAR SWE retrievals focussed on L-band and C-band RP-InSAR applications in the Upper Engadin region in Switzerland, based on repeat-pass SAR time series acquired by ALOS PALSAR-2, respectively Sentinel-1. ALOS PALSAR-2 strip map mode data of different tracks were acquired in 14, respectively 28 day repeat intervals, starting from snow free conditions in October 2019 until the beginning of the main melting period in March 2020. On days of the PALSAR overflights snow and soil parameters were measured at several locations in snow pits and along transects. Continuous time series on the temporal evolution of snow accumulation, melt events and meteorological parameters are available from automatic weather stations. In case of dry snow, the coherence of the PALSAR data is preserved also in case of snowfall, although intensive snowfall events cause a substantial decrease in coherence. SWE retrievals based on 14-day PALSAR data show for dry snow cases good agreement with insitu snow measurements. However, a continuous time series throughout winter could not be obtained because the 2019/20 PALSAR data set suffers from lack of continuity, comprising data from 4 different tracks, and also because on some acquisition dates the coherence was affected by transient melt. Limitations in the spatial coverage are imposed by the steep topography causing large gaps due to layover and foreshortening, calling for the acquisition of near-coincident data from ascending and descending passes. The Sentinel-1 6-day RP-InSAR data show suitable coherence during stable dry snow conditions also in case of light snowfall, but melt events and moderate and intensive snowfall causes complete decorrelation.
During March 2021 an experimental airborne campaign was carried out in the high Alpine test site Wörgetal/Kühtai near Innsbruck, addressing two complementary approaches for SWE measurements: (i) exploring the measurement concept of a geostationary C-band SAR mission for retrieval of dense SWE time series; (ii) consolidating the assessment of the RP-InSAR based SWE retrieval method and performance in support of mission preparation for ROSE-L and Sentinel-1 NG. The activities were performed by DLR and ENVEO within the ESA project SARSimHT-NG. Here we provide a first account on the studies related to objective (ii), as the airborne data enable the direct comparison of dual frequency (C- and L-band) polarimetric measurements. Within the period 2nd to 19th March 2021 multiple C- and L-Band SAR data were acquired F-SAR flights on 7 days spanning two snow fall events of about 10 cm and 40 cm mean fresh snow depth. On days of the F-SAR overflights vertical profiles of physical snow parameters were measured in snow pits at different locations as well as snow depths along transects.
In the presentation we report on the impact of snowfall and other environmental parameters on C- and L-band InSAR coherence over time periods ranging from several hours to days, and report on the performance of the InSAR SWE retrievals performed over the test sites Wörgetal/Kühtai and Engadin. The campaign activities and the presented results are of relevance for preparing snow monitoring activities with current and upcoming L-Band SAR missions including SAOCOM A/B, NISAR and especially the Copernicus Expansion Mission ROSE-L. Furthermore, the investigations are of relevance for exploring the combined use of C- and L-Band SAR data for monitoring main snowpack parameters.

Rice is a staple cereal crop in Asia, and the continent accounts for about 90% of global rice production and consumption. Rice-planted area map is important parameter to estimate rice production for food security or economy, and also to quantify the carbon, water cycle or methane emission via paddy fields. Rice is mainly cultivated in the rainy season, Synthetic Aperture Radar (SAR) is therefore a robust tool because it penetrates cloud cover. Recently, machine learning has been widely used in many land cover related researches and distinct results were reported, however, limitation is that it needs a large amount of training data, it is normally time and cost consuming task. In this research, we utilized the combination of unsupervised and supervised classification to efficiently produce the training data. Training data were generated from the k-means classification results for the sampled regions, then a random forest classifier was applied to ALOS-2 PALSAR-2 ScanSAR data to identify rice-planted area in Southeast Asian countries including Cambodia, Laos PDR, Thailand, and Vietnam. It is also difficult to identify rice-planted area in this region since there are high variations in rice phenology. In order to compensate for the variations, we used time-series metrics of calculated from SAR data. Classification models were fine-tuned for the target countries, and most of the models had an accuracy of 0.9 or better. Independent verification through visual interpretation using very high resolution images (VHRs) on Google Earth also showed a high degree of consistency with the classification results. The developed paddy field maps showed high accuracy in most countries and regions. In addition to these verifications, resulted map was compared with other rice planted area maps developed by Vietnam National Space Center(VNSC) and CNES/CESBiO under the collaboration of the VNSC’s 2019 Committee on Earth Observation Satellite (CEOS) chair initiative on rice monitoring. VNSC developed rice map for the Vietnam using time series Sentinal-1 data, and CNES/CESBiO developed rice map also using time series Sentinal-1 for four countries including Cambodia, Laos PDR, Thailand, and Vietnam under the ESA GEORICE project. Comparison results showed good consistency between these products by visual interpretation. However, further quantitative comparison or verification using in-situ data and national statistics, as well as application to other regions, seasons and years, is necessary to confirm the effectiveness of proposed methodology.

Agricultural production consumes the largest share of the world's water resources, using up to 90% of available freshwater. As demand for agricultural products is estimated to increases in the future, the accompanying intensification of production will put even more pressure on available water resources. This development makes the sector even more vulnerable to the increasing impacts of climate change on hydrological conditions and demands an even more efficient water use. Detailed knowledge of the spatial and temporal state of soil moisture, which is a key parameter in plant nutrition, agricultural production and environmental research, can help addressing these challenges. While in-situ measurement methods can be used for local monitoring, they are unsuitable for regional and even global application. In this regard, high resolution surface soil moisture data for regional and local monitoring (down to precision farming level) are still lacking at a global scale. Here, the increasing spatial and temporal resolution of current and future Synthetic Aperture Radar (SAR) satellite missions (e.g. Sentinel-1, ALOS-2/4, NISAR, ROSE-L) can help to overcome this problem. Nevertheless, the SAR missions come with individual limitations regarding soil moisture estimation. While the C-band SAR mission Sentinel-1 provides high temporal and spatial resolution, it has reduced sensitivity to soil characteristics under certain vegetation coverage due to its short wavelength. On the other hand, L-band SAR missions like ALOS-2 better penetrate covering vegetation, while its temporal and spatial resolution is reduced compared to C-band SAR missions.
Using low pass filtering as well as vegetational detrending, we developed an algorithm using high-resolution Sentinel-1 SAR timeseries for estimating soil moisture based on a change detection approach, so called alpha approximation (Balenzano et al. 2011). Tested and validated over the Rur catchment, comprising a diverse cropping structure and located in the federal state of North-Rhine Westphalia in the West of Germany, it showed encouraging results with a mean R² of 0.46 and an unbiased RMSE (uRMSE) of 5.84 %. Nevertheless, especially during the growing season, the estimated soil moisture shows a bias compared to the in-situ measured soil moisture for some crops. The reason for this is the changing sensitivity of the C-band backscattering signal to a uniform change in soil moisture occurring during the growing period. To overcome this problem, L-band ALOS-2 data is used, being less affected by vegetation cover and more sensitive to changes in soil moisture. By combining both timeseries, the high spatial and temporal resolution of C-band Sentinel-1 is used to fill the gaps between the sparse ALOS-2 recordings, while benefiting from the lower vegetation sensitivity of L-band. In this regard, this study evaluates different methods for assembling both microwave frequency bands, e.g. using L-band ALOS-2 soil moisture estimations as boundary conditions for C-band Sentinel-1 soil moisture estimation or matching changes in C-band backscattering signal to the co-located changes of L-band backscattering signal. The resulting algorithm for soil moisture estimation will be integrated within an automated workflow, using a cloud-computing platform (e.g. Google Earth Engine, CODE-DE). This enables fast processing without the use of local computational infrastructure. It will be validated over an Apulian test site, reflecting the diverse agricultural landscape of the Mediterranean region.