Conference Agenda

The performance of low frequency Synthetic Aperture Radar (SAR) is con- strained by trans-ionospheric propagation because the dispersive nature of the ionosphere. In Interferometric SAR (InSAR) the ionospheric signature is trans- lated into shifts in azimuth due to differential phase gradients, phase ramps in range, ionospheric phase screens and decorrelation due to Faraday rotation (FR). All these degrade the quality of the interferometric products [1]. Not to mention the defocusing present in the single images due to the fast changing ionosphric electron density irregularities. In the framework of the new Biomass mission (full polarimetric P-band operation) different algorithms have been pro- posed for the polarimetric calibration and phase correction: the approaches are based in the Bickel and Bates estimation of the FR (as a bypass for phase cor- rection) [2], Mapdrift Autofocus (MDA) or a combination of both [3]. We are proposing an extension of the autofocus that incorporates information from in- terferometric pairs to enhance the phase estimation stability and resolution for better calibration of the single images, and at the same time is consistent with the interferogram (which we know has high resolution).

Good sensitivity of the FR for phase correction is not always warranted; there is the accuracy of the FR angle due to the Signal-to-Noise Ratio (SNR), the latitude-sensitivity dependence of the FR (lower sensitivity towards the electromagnetic equator) and the large scaling factor between FR and phase error (with the associated noise scaling). The accuracy of this scaling factor depends on the uncertainty in the determination of the ionospheric height and corresponding piercing point geomagnetic field [2].

The development of the MDA is an effort to directly apply phase corrections and the retrieval of higher resolution phase screens, but its performance on the other hand depends on the contrast of the image and Signal to Clutter Ratio (SCR) [4] as well as the quality of the cross-correlation peaks between azimuth sub-looks. The MDA is sensitive to the second derivative of the variations along the azimuth direction [5], so errors in the estimation of this second derivative will propagate as random walks during the integration. This integration can be bounded with a Weighted Least Squares (WLS) in which the FR information (when available and reliable) is included but even then further external infor- mation can be desired. Here is where we believe the interferometric autofocus can provide better phase estimation.

None of these methods work at full resolution, which is limited by the filtering of the FR and block processing of the MDA (that also acts as a block averaging filter). Towards the geomagnetic equator or in low SNR scenarios, small error in the FR angle can require large averaging filters. Similarly, when the contrast in

the image is not good enough, larger MDA blocks are needed. In any case, the spectrum of the originally disturbing phase screen is cut by a band-pass filter and the high frequency component goes lost. This high frequency component corresponds to fast varying phase screen structures which are left behind as calibration errors and seen as undesired phase patterns in the interferograms. By better bounding the MDA integration step and cancelling random walks, smaller blocks that correspond to a larger band-pass cut-off frequency can be taken.

An autofocus algorithm together with an error assessment based on the spectral analysis of the calibration errors will be presented. First results con- taining the corrected images and corresponding phase screens obtained with the Biomass End-to-End Performance Simulator (BEEPS) [6] will be shown.

References

1. [1] Franz J Meyer and Jeremy Nicoll. The impact of the ionosphere on interfero- metric sar processing. In IGARSS 2008-2008 IEEE International Geoscience and Remote Sensing Symposium, volume 2, pages II–391. IEEE, 2008.

2. [2] Jun Su Kim, Konstantinos P Papathanassiou, Rolf Scheiber, and Shaun Quegan. Correcting distortion of polarimetric sar data induced by iono- spheric scintillation. IEEE Transactions on Geoscience and Remote Sensing, 53(12):6319–6335, 2015.

3. [3] Valeria Gracheva, Jun Su Kim, Pau Prats-Iraola, Rolf Scheiber, and Marc Rodriguez-Cassola. Combined estimation of ionospheric effects in sar images exploiting faraday rotation and autofocus. IEEE Geoscience and Remote Sensing Letters, 19:1–5, 2021.

4. [4] Richard Bamler and Michael Eineder. Accuracy of differential shift esti- mation by correlation and split-bandwidth interferometry for wideband and delta-k sar systems. IEEE Geoscience and Remote Sensing Letters, 2(2):151– 155, 2005.

5. [5] Walter G Carrara Ron S Goodman and Ronald M Majewski. Spotlight synthetic aperture radar signal processing algorithms. Artech House, pages 245–285, 1995.

6. [6] Maria J Sanjuan-Ferrer, Pau Prats-Iraola, Marc Rodriguez-Cassola, Mariantonietta Zonno, Muriel Pinheiro, Matteo Nannini, Nestor Yague- Martinez, Javier del Castillo-Mena, Thomas Boerner, Konstantinos P Pap- athanassiou, et al. End-to-end performance simulator for the biomass mis- sion. In EUSAR 2018; 12th European Conference on Synthetic Aperture Radar, pages 1–5. VDE, 2018.

Coastal wetlands are highly productive ecosystems providing important habitat for a wide variety of plants and animals and provide a range of ecosystem services from improving water quality and sequestering carbon. Due to pollution, urban and agricultural development, and sea level rise, wetlands are under environmental stress. There is a pressing need to monitor coastal wetlands’ health and hydrology. Thus far, most observations of hydrodynamic processes within coastal wetlands have been done through deployment of in situ water level gauge stations. While these networks measure water level changes (WLC) with fast temporal sampling, they are spatially sparse.

Spaceborne and airborne synthetic aperture radar interferometry (InSAR) can, on the other hand, characterize large scale water level changes in wetlands. The approach works because of the presence of emergent vegetation which, with water, effectively create corners that reflect microwaves toward the radar instrument (so-called double-bounce effect). We measure the differential phase between images of the same region collected with the same viewing geometry but different time. As such, any water level change occurring between radar acquisitions will change the distance traveled by the microwaves (Fig1).

On a practical level, the sensor frequency, vegetation type and seasonal vegetation changes impact the quality of measurements. However, the impact of changes in target characteristics, which include changes in moisture, wind and atmosphere can significantly decrease repeat-pass InSAR coherence. Phase delays caused by atmospheric effects greatly limit InSAR measurement accuracy and may lead to misunderstanding and/or misinterpretation of the phenomena of interest. Several studies have been conducted to characterize the atmosphere and mitigate its effects on InSAR time-series measurements, either with or without external data [2]-[6]. Often, atmospheric InSAR corrections based on external weather-model data or GPS delay estimations are used to minimize the impact of atmospheric phase delays. However, for airborne InSAR, many of these implementations are not suitable due to the coarse resolution of available models, and the poor spatial coverage of GPS stations. Thus, for airborne InSAR where the wet troposphere presents the main issue, there is no straightforward approach to deal with it or correct the bias introduced by dense cumulus clouds that contains an important amount of water vapor.

In this work, we aim to assess the differences between airborne InSAR and spaceborne InSAR for water level change monitoring in coastal wetlands with emphasis on atmospheric effects identification and corrections. While the InSAR process is the same for airborne and spaceborne SAR, the considerations are different. In fact, atmospheric corrections of spaceborne interferograms, including ionospheric delay correction, using split spectrum algorithm, and tropospheric delay correction using weather models, are different from airborne “atmospheric” corrections. Airborne SAR are affected by the wet troposphere (up to 15 km from ground surface) which includes cumulus clouds. For this study, we conduct interferometric processing on L-band spaceborne SAR acquisitions and L-band airborne SAR acquisitions. For the airborne wet troposphere delay correction, we suggest an approach based on Independent Component Analysis (ICA)[7]-[10].

We use 10 ALOS-2/PALSAR-2 L-band spaceborne acquisitions over wetlands of coastal Louisiana with a temporal baseline of 14 days. We apply time series analysis and generate 9 final water level change maps of coastal Louisiana wetlands from January to February 2019. The processing steps include ionospheric correction with the split spectrum algorithm. We also used the airborne InSAR time series from the Uninhabited Aerial Vehicle Synthetic Aperture Radar (UAVSAR) L-band sensor acquired in the scope of NASA’s Delta-X project over coastal Louisiana.

NASA’s Delta-X airborne mission promises to deliver hydrodynamic and ecological models that can be used to assess the resilience and vulnerability of the various parts the Mississippi River delta. One of the Delta-X instrument is UAVSAR’s L-band Synthetic Aperture Radar. During the 2021 Delta-X campaigns, UAVSAR collected repeat-pass Interferometric data to measure (WLC) in wetlands. There were 5 separate UAVSAR flights during the Fall and Spring of 2022. UAVSAR flew in a so-called ‘race-track’ pattern over the West Terrebonne and Atchafalaya basins at an altitude of 12.5 km, repeating measurements every 20 to 40 minutes during each approximately 5 hours flight. After pre-processing the SLC acquisitions using ISCE2^[1] and applying Small BAsline Subset (SBAS) time series analysis using Mintpy^[2], the final WLC UAVSAR- L3 time series products were produced and published for public access on the ORNL DAAC [1]. We found the atmospheric effects to be significant, in particular in the presence of dense cloud cover and potential rain events. Our approach, to identify and reduce the bias introduced by clouds layer, uses a multi-step framework: applying ICA to a stack of unwrapped interferograms, generating independent components, and applying a segmentation algorithm to separate the present information in each axis of the ICA output to isolate the atmospheric features. Finally, we compare the InSAR WLC measurements retrieved from ALOS time series and UAVSAR time series with in situ gauges from the Coastwide Reference Monitoring System (CRMS) stations.

The results also show the potential of using ICA for clouds features identification in UAVSAR airborne time series of WLC. To validate our results, we compared our ICA algorithm output masks, identified as the atmospheric dense cloud layer, against NOAA NEXt-Generation RADar (NEXRAD) ground weather radar. The latter is a high-resolution S-band Doppler weather radar. The National Centers for Environmental Information (NCEI) provides access to archived NEXRAD Level-II data which consist of reflectivity maps. Preliminary results show good correlation between features of high-water vapor content on NEXRAD data and the extracted atmospheric masks. Our algorithm provides an alternative solution to automatically detect atmospheric phase delays introduced by Wet Troposphere layer for airborne InSAR. Moreover, the ICA approach does not require in situ data or models. Our study can serve as a lookup table to what to expect from airborne and spaceborne InSAR and their potential for global monitoring of coastal wetland hydrology.

References:

[1] Jones, C., T. Oliver-cabrera, M. Simard, and Y. Lou. 2022. Delta-X: UAVSAR Level 3 Geocoded InSAR Derived Water Level Changes, LA, USA, 2021. ORNL DAAC, Oak Ridge, Tennessee, USA, doi: 10.3334/ORNLDAAC/2058.

[2] Z. Li, J. Muller, P. Cross, P. Albert, J. Fischer, R. Bennartz, Assessment of the potential of MERIS near — infrared water vapor products to correct ASAR interferometric measurements, International Journal of Remote Sensing, 27 (2006), pp. 349-365, 10.1080/01431160500307342

[3] J. Löfgren, F. Björndahl, A. Moore, F. Webb, E. Fielding, E. Fishbein, Tropospheric correction for InSAR using interpolated ECMWF data and GPS Zenith Total Delay from the Southern California Integrated GPS Network, Geoscience and Remote Sensing Symposium (IGARSS), 2010 IEEE International (2010), pp. 4503-4506, 10.1109/IGARSS.2010.5649888

[4] F. Onn, H. Zebker, Correction for interferometric synthetic aperture radar atmospheric phase artifacts using time series of zenith wet delay observations from a GPS network, Journal of Geophysical Research, 111 (2006), 10.1029/2005JB004012

[5] P. Berardino, G. Fornaro, R. Lanari and E. Sansosti, "A new algorithm for surface deformation monitoring based on small baseline differential SAR interferograms," in IEEE Transactions on Geoscience and Remote Sensing, vol. 40, no. 11, pp. 2375-2383, Nov. 2002, doi: 10.1109/TGRS.2002.803792.

[6] A. Ferretti, C. Prati and F. Rocca, "Permanent scatterers in SAR interferometry," in IEEE Transactions on Geoscience and Remote Sensing, vol. 39, no. 1, pp. 8-20, Jan. 2001, doi: 10.1109/36.898661.

[7] Cohen‐Waeber, J., et al. "Spatiotemporal patterns of precipitation‐modulated landslide deformation from independent component analysis of InSAR time series." Geophysical Research Letters 45.4 (2018): 1878-1887, doi: 10.1002/2017GL075950

[8] Zhu, K.; Zhang, X.; Sun, Q.; Wang, H.; Hu, J. Characterizing Spatiotemporal Patterns of Land Deformation in the Santa Ana Basin, Los Angeles, from InSAR Time Series and Independent Component Analysis. Remote Sens. 2022, 14, 2624. doi:10.3390/rs14112624.

[9] Maubant, Louise, Erwan Pathier, Simon Daout, Mathilde Radiguet, M‐P. Doin, Ekaterina Kazachkina, Vladimir Kostoglodov, Nathalie Cotte, and Andrea Walpersdorf. "Independent component analysis and parametric approach for source separation in InSAR time series at regional scale: application to the 2017–2018 Slow Slip Event in Guerrero (Mexico)." Journal of Geophysical Research: Solid Earth 125, no. 3 (2020): e2019JB018187. Doi: 10.1029/2019JB018187.

[10] Gaddes, M. E., A. Hooper, M. Bagnardi, H. Inman, and F. Albino. "Blind signal separation methods for InSAR: The potential to automatically detect and monitor signals of volcanic deformation." Journal of Geophysical Research: Solid Earth 123, no. 11 (2018): 10-226. 10.1029/2018JB016210.

^[1] Interferometric synthetic aperture radar Scientific Computing Environment (ISCE): https://github.com/isce-framework/isce2

^[2] The Miami INsar Time-series software in PYthon: https://github.com/insarlab/MintPy

Several authors have reported the results of the beneficial impacts of assimilating InSAR meteorology products when predicting the tridimensional moisture structure as well as the location and timing of precipitations (e.g., Pichelli et al. (2015) among the first works). Mateus et al. (2018) significantly improved the forecast of two consecutive deep convective storms, demonstrating the value of InSAR data in severe weather. Unlike the Adra occurrences, poorly forecasted without InSAR data assimilation, Lagasio et al. (2019) and Pierdicca et al. (2020) combined Sentinel-1 products and GNSS-derived data in two severe events of precipitation in Italy, achieving slight increases in the forecast skill. An InSAR dataset consisting of 51 interferograms was assimilated by Miranda et al. (2019) southwest of the Appalachian Mountains, which resulted in a significant overall improvement in precipitation climatology. (Mateus et al., 2021) continuously ingest InSAR PWV maps (one every 12 hours) over Iberia for 12 days, restricting the model's initial moisture field, and resulting in better specific humidity profiles and more accurate forecasts. More recently, Mateus and Miranda (2022) assimilated 2.5 years of InSAR PWV maps generated from Sentinel-1 images acquired near Santa Cruz de la Sierra, Bolivia, to assess the quality of the water vapor field at the core of the South American Low-level Jet. They mostly conclude that InSAR has the potential to limit systematic biases in water vapor measurement, having a positive or neutral impact on the precipitation forecast.

In this work, we present the results of an application of InSAR meteorology to improve the description of the 3D s vertical distribution of the water vapor in the atmosphere both at the footprint of the Sentinel-1 images used to generate the InSAR meteorology products assimilated in the NWM and in other geographical regions reached by the water vapor flow anomalies. The main contribution of InSAR meteorology is to help to detect the water vapour anomalies not correctly modelled by the NWMs, using the high spatial resolution and large coverage of the Sentinel-1 images. Furthermore, InSAR meterology provides a means to validate the forecasted spatial propagation of the water vapor provided by the NWM after the assimilation of InSAR products. Lagrangian trajectories are computed and used to follow the water vapor mixing ratio anomalies around to the steering level, starting from the footprint of Sentinel-1 images assimilated. The vertical distribution of water vapor anomalies is also studied along each Lagrangian trajectory. The problem of temporal decay of InSAR information within the NWM model is also studied. The main output of this work is to show the potential and perspective use of InSAR meteorology within the Destination Earth (DestineE) initiative. The joined use of high resolution NWM (such as WRF) and the next large availability and redundancy of C- and L-band interferometric SAR missions (besides the current Sentinel-1 A&B and SAOCOM missions and the next Sentinel-1 C&D, N.G., ROSE-L, ALOS-4, NISAR), provides an example of the digital model of Earth that could support the complex task of anticipating extreme weather events.

References:

Lagasio, M., Pulvirenti, L., Parodi, A., Boni, G., Pierdicca, N., Venuti, G., Realini, E., Tagliaferro, G., Barindelli, S., Rommen, B., 2019. Effect of the ingestion in the WRF model of different Sentinel-derived and GNSS-derived products: analysis of the forecasts of a high impact weather event. Eur J Remote Sens 52, 16–33.

Mateus, P., Miranda, P.M.A., 2022. Using InSAR Data to Improve the Water Vapor Distribution Downstream of the Core of the South American Low-Level Jet. Journal of Geophysical Research: Atmospheres 127, e2021JD036111.

Mateus, P., Miranda, P.M.A., Nico, G., Catalao, J., 2021. Continuous Multitrack Assimilation of Sentinel-1 Precipitable Water Vapor Maps for Numerical Weather Prediction: How Far Can We Go With Current InSAR Data? Journal of Geophysical Research: Atmospheres 126, e2020JD034171.

Mateus, P., Miranda, P.M.A., Nico, G., Catalão, J., Pinto, P., Tomé, R., 2018. Assimilating InSAR Maps of Water Vapor to Improve Heavy Rainfall Forecasts: A Case Study With Two Successive Storms. Journal of Geophysical Research: Atmospheres 123, 3341–3355.

Miranda, P.M.A., Mateus, P., Nico, G., Catalão, J., Tomé, R., Nogueira, M., 2019. InSAR Meteorology: High-Resolution Geodetic Data Can Increase Atmospheric Predictability. Geophys Res Lett 46, 2949–2955.

Pichelli, E., Ferretti, R., Cimini, D., Panegrossi, G., Perissin, D., Pierdicca, N., Rocca, F., Rommen, B., 2015. InSAR Water Vapor Data Assimilation into Mesoscale Model MM5: Technique and Pilot Study. IEEE J Sel Top Appl Earth Obs Remote Sens 8, 3859–3875.

Pierdicca, N., Maiello, I., Sansosti, E., Venuti, G., Barindelli, S., Ferretti, R., Gatti, A., Manzo, M., Monti-Guarnieri, A.V., Murgia, F., Realini, E., Verde, S., 2020. Excess Path Delays from Sentinel Interferometry to Improve Weather Forecasts. IEEE J Sel Top Appl Earth Obs Remote Sens 13, 3213–3228.

Interferometric Synthetic Aperture Radar (InSAR) measurements often suffer from errors caused by atmospheric delays. To reduce these errors, two main classes of methods are typically used: Methods based on external information and methods using data-driven techniques. The former class of methods relies on external data such as GNSS-derived tropospheric models, meteorological data, and atmospheric model outputs, but these usually have lower spatial resolution than required for many InSAR applications. In contrast, data-driven methods directly use the InSAR data and generally separately address the stratified and turbulent components of the atmospheric delays. One issue with such separated error reduction is that it may result in biased estimates of the atmospheric delays due to the interdependence of these two components. Furthermore, InSAR observations are also affected by long-wavelength ionospheric disturbances and orbital errors, making it challenging to obtain reliable InSAR displacements.

To address these issues, we propose a new data-driven method that simultaneously models and mitigates the turbulent and stratified delays by leveraging their spatiotemporal characteristics as a priori information. In this method, which we call DetrendInSAR, the turbulent delays are modeled as a spatially slow-changing process that can be fitted by position-related polynomials within a small area (e.g., 1 km x 1 km), while the stratified delay can be linearly fitted with the local terrain height. These a priori information are combined to establish a solvable mathematical model for the delays based on a novel pixel-by-pixel window-based modeling strategy. Since the ionospheric disturbances and orbital errors show slow-changing spatial patterns within a small area, these two error components can also be accounted in the DetrendInSAR modeling process. Moreover, the displacement signals in the InSAR observations are assumed to be a temporally smooth process, providing additional constraints for distinguishing between displacements and turbulent delays in the DetrendInSAR modeling process. We validate the DetrendInSAR method using both simulated datasets and an actual 16-month-long Sentinel-1 SAR time series of the postseismic deformation after the 22 May 2021 Maduo earthquake, China. The results are compared to those of a standard data-driven strategy that fits a ramp and a terrain-related linear function over the whole image based on far-field signals and suppresses the turbulent delays by temporally averaging adjacent SAR-image acquisitions. By taking 3D GNSS displacement time series as the benchmark, we find that the DetrendInSAR results are more accurate compared with the standard data-driven strategy. Furthermore, from both ascending and descending orbit data (and derived east and vertical displacements), the logarithmic decay of the postseismic deformation after the Maduo earthquake is illuminated, with poroelastic rebound significantly contributing to the near-field postseismic deformation, in addition to afterslip reported in earlier studies.