Wenyu Gong

Temporal Filtering of InSAR Data Using Statistical Parameters From NWP Models

Finding solutions for the mitigation of atmospheric phase delay patterns from differential synthetic aperture radar interferometry (d-InSAR) observations is currently one of the most active research topics in radar remote sensing. Recently, many studies have analyzed the performance of regional numerical weather prediction (NWP) models for this task; however, despite the significant efforts made to optimize model parameterizations, most of these studies have concluded that current regional NWPs are not able to robustly reproduce the atmospheric phase delay structures that affect SAR interferograms. Despite these previous findings, we have revisited the application of NWPs for atmospheric correction using a different analysis strategy. In contrast to earlier studies, which assessed the quality of NWP-derived phase screen data, we have studied NWPs from a statistical angle by analyzing whether they are able to provide realistic information about the statistical properties of atmospheric phase signals in d-InSAR data. We have determined that NWP forecasts can provide relevant statistical information about the atmospheric phase screen captured in d-InSAR data. Based on this, this study presents a new atmospheric phase filtering approach that is using statistical atmospheric information as a prior in order to optimize the choice of unknown filter parameters. The mathematical concept of the prior-driven filtering approach is outlined, and its implementation is explained. We have determined the performance of this new filter concept and have shown that it comes very close to a filter optimum.

InSAR Detection and Field Evidence for Thermokarst after a Tundra Wildfire, Using ALOS-PALSAR

Thermokarst is the process of ground subsidence caused by either the thawing of ice-rich permafrost or the melting of massive ground ice. The consequences of permafrost degradation associated with thermokarst for surface ecology, landscape evolution, and hydrological processes have been of great scientific interest and social concern. Part of a tundra patch affected by wildfire in northern Alaska (27.5 km2) was investigated here, using remote sensing and in situ surveys to quantify and understand permafrost thaw dynamics after surface disturbances. A two-pass differential InSAR technique using L-band ALOS-PALSAR has been shown capable of capturing thermokarst subsidence triggered by a tundra fire at a spatial resolution of tens of meters, with supporting evidence from field data and optical satellite images. We have introduced a calibration procedure, comparing burned and unburned areas for InSAR subsidence signals, to remove the noise due to seasonal surface movement. In the first year after the fire, an average subsidence rate of 6.2 cm/year (vertical) was measured. Subsidence in the burned area continued over the following two years, with decreased rates. The mean rate of subsidence observed in our interferograms (from 24 July 2008 to 14 September 2010) was 3.3 cm/year, a value comparable to that estimated from field surveys at two plots on average (2.2 cm/year) for the six years after the fire. These results suggest that this InSAR-measured ground subsidence is caused by the development of thermokarst, a thawing process supported by surface change observations from high-resolution optical images and in situ ground level surveys.

Measurement and interpretation of subtle deformation signals at Unimak Island from 2003 to 2010 using weather model‐assisted time series InSAR

A 7 year time series of satellite radar images over Unimak Island, Alaska—site of Westdahl Volcano, Fisher Caldera, and Shishaldin Volcano—was processed using a model-free Persistent Scatterer Interferometry technique assisted by numerical weather prediction model. The deformation-only signals were optimally extracted from atmosphere-contaminated phase records. The reconstructed deformation time series maps are compared with campaign and continuous Global Positioning System (GPS) measurements as well as Small Baseline Subset interferometric synthetic aperture radar (InSAR) results for quality assessment and geophysical interpretation. We observed subtle surface inflation at Westdahl Volcano that can be fit by a Mogi source located at approximately 3.6 km north of Westdahl peak and at depth of about 6.9 km that is consistent with the GPS-estimated depth for the 1998 to 2001 time period. The magma chamber volume change decays during the period of 2003 to 2010. The deformation field over Fisher Caldera is steadily subsiding over time. Its best fit analytical model is a sill source that is about 7.9 km in length, 0.54 km in width, and located at about 5.5 km below sea level underneath the center of Fisher Caldera with strike angle of N52°E. Very little deformation was detected near Shishaldin peak; however, a region approximately 15 km east of Shishaldin, as well as an area at the Tugamak range at about 30 km northwest of Shishaldin, shows evidence for movement toward the satellite, with a temporal signature correlated with the 2004 Shishaldin eruption. The cause of these movements is unknown.

Methods of InSAR atmosphere correction for volcano activity monitoring

When a Synthetic Aperture Radar (SAR) signal propagates through the atmosphere on its path to and from the sensor, it is inevitably affected by atmospheric effects. In particular, the applicability and accuracy of Interferometric SAR (InSAR) techniques for volcano monitoring is limited by atmospheric path delays. Therefore, atmospheric correction of interferograms is required to improve the performance of InSAR for detecting volcanic activity, especially in order to advance its ability to detect subtle pre-eruptive changes in deformation dynamics. In this paper, we focus on InSAR tropospheric mitigation methods and their performance in volcano deformation monitoring. Our study areas include Okmok volcano and Unimak Island located in the eastern Aleutians, AK. We explore two methods to mitigate atmospheric artifacts, namely the numerical weather model simulation and the atmospheric filtering using Persistent Scatterer processing. We investigate the capability of the proposed methods, and investigate their limitations and advantages when applied to determine volcanic processes.

Performance analysis of atmospheric correction in InSAR data based on the Weather Research and Forecasting Model (WRF)

The influence of the turbulent atmosphere is seen as the main performance limitation for high-quality Interferometric Synthetic Aperture Radar (InSAR) techniques in ground deformation monitoring applications. Atmospheric correction using numerical weather prediction (NWP) models is widely seen as a promising emerging technology for mitigation of atmospheric signals. First results showed promising capabilities for correction of stratified delay yet have revealed limited performance for modeling and mitigating turbulent atmospheric water vapor signals from SAR [1, 2]. This paper presents an integration of InSAR observations with predictions from the high-resolution Weather Research and Forecasting Model (WRF). Special focus is put on investigating improvements in the weather model parameterization to achieve enhanced performance in atmospheric correction. First, a statistical analysis of the quality of absolute delay predictions is presented based on a comparison of vertically integrated WRF delays with radiosonde measurements. Second, the performance of WRF for atmospheric correction of InSAR data is analyzed by comparing WRF phase delay maps to SAR interferograms and analyzing structure functions and variances of the residual atmospheric delay signal. Here, significant improvements could be achieved through modifications of the WRF model parameterization, which are highlighted in Section 3.2. From our study, we conclude that the performance of latest generation high-resolution NWPs can be significantly improved if the setup and parameterization of the model domain is optimized.

Optimized filter design for irregular acquired data stack in persistent scatterers synthetic aperture radar interferometry

After more than a decade of continuous improvements, Persistent SAR Interferometry (PSI) has become an accepted tool for long-term monitoring of geophysical phenomena such as volcanoes, earthquakes, and city deformation. To extract surface deformation from PSI observations, different phase component are separated based on their physical or stochastic properties. Atmospheric signals, one of the main error contributions, are usually mitigated using a low-pass filter in time. This approach provides unsatisfactory results if the temporal sampling of the surface deformation signal is close to Nyquist, and if there are sampling gaps in the time series. Hence, in many cases, a more sophisticated way to mitigate atmospheric signals is required. This paper proposes an adaptive filter design that is utilizing knowledge of the stochastic properties of the atmosphere at SAR acquisition times. The concept of the filter is presented and its performance is tested on simulated signals. In these studies, the adaptive filter design performed favorably and we are expecting this approach to be of major benefit for PSI processing. In current work, the performance of the adaptive filter is tested on real data. For this study, the pre-known statistics of atmospheric delay is provided by Numerical Weather Forecast Models.

The role of weather models in mitigation of tropospheric delay for SAR interfermetry

High resolution numerical weather models have recently raised a great interest in the InSAR community for atmospheric phase screen (APS) mitigation. Following the research carried out in [1], in this study we focus on investigating the sensitivity of WRF (Weather Research and Forecasting) predictions to the model parameter settings which may substantially affect the result of water vapor modeling and to different boundary conditions. We validate the model predictions using atmosphere-only interferograms as well as radiosonde records. Our result shows that the radiosonde records (on average) agree very well with the WRF predictions based on our new model settings. However, in terms of spatio-temporal delay variation, the new settings do not always lead to a better prediction and the correction of atmospheric delay is case dependent. Therefore, we conclude that WRF lacks the reliability to correct the realistic APS in interferograms.

A New Numerical Method for Calculating Extrema of Received Power for Polarimetric SAR

A numerical method called cross-step iteration is proposed to calculate the maximal/minimal received power for polarized imagery based on a targets Kennaugh matrix. This method is much more efficient than the systematic method, which searches for the extrema of received power by varying the polarization ellipse angles of receiving and transmitting polarizations. It is also more advantageous than the Schuler method, which has been adopted by the PolSARPro package, because the cross-step iteration method requires less computation time and can derive both the maximal and minimal received powers, whereas the Schuler method is designed to work out only the maximal received power. The analytical model of received-power optimization indicates that the first eigenvalue of the Kennaugh matrix is the supremum of the maximal received power. The difference between these two parameters reflects the depolarization effect of the targets backscattering, which might be useful for target discrimination.

Measuring Co-seismic Deformation of the Sichuan Earthquake by Satellite Differential INSAR

The Sichuan Earthquake, occurred on May 12, 2008, is the strongest earthquake to hit China since the 1976 Tangshan earthquake. The earthquake had a magnitude of M 8.0, and caused surface deformation greater than 3 meters. This paper presents the research work of measuring the co-seismic deformations of the earthquake with satellite differential interferometric SAR technique. Four L-band SAR images were used to form the interferogram with 2 pre- scenes imaged on Feb 17, 2008 and 2 post- scenes on May 19, 2008. The Digital Elevation Models extracted from 1:50,000-scale national geo-spatial database were used to remove the topographic contribution and form a differential interferogram. The interferogram presents very high coherence in most areas, although the pre- and post- images were acquired with time interval of 92 days. This indicates that the L-band PALSAR sensor is very powerful for interferometry applications. The baseline error is regarded as the main phase error source in the differential interferogram. Due to the difficulties of doing field works immediately after the earthquake, only one deformation measurement recorded by a permanent GPS station is obtained for this research. An approximation method is proposed to eliminate the orbital phase error with one control point. The derived deformation map shows similar spatial pattern and deformation magnitude compared with deformation field generated by seismic inversion method.

Pyroclastic Flow Deposits and InSAR: Analysis of Long-Term Subsidence at Augustine Volcano, Alaska

Deformation of pyroclastic flow deposits begins almost immediately after emplacement, and continues thereafter for months or years. This study analyzes the extent, volume, thickness, and variability in pyroclastic flow deposits (PFDs) on Augustine Volcano from measuring their deformation rates with interferometric synthetic aperture radar (InSAR). To conduct this analysis, we obtained 48 SAR images of Augustine Volcano acquired between 1992 and 2010, spanning its most recent eruption in 2006. The data were processed using d-InSAR time-series analysis to measure the thickness of the Augustine PFDs, as well as their surface deformation behavior. Because much of the 2006 PFDs overlie those from the previous eruption in 1986, geophysical models were derived to decompose deformation contributions from the 1986 deposits underlying the measured 2006 deposits. To accomplish this, we introduce an inversion approach to estimate geophysical parameters for both 1986 and 2006 PFDs. Our analyses estimate the expanded volume of pyroclastic flow material deposited during the 2006 eruption to be 3.3 × 107 m3 ± 0.11 × 107 m3, and that PFDs in the northeastern part of Augustine Island reached a maximum thickness of ~31 m with a mean of ~5 m. Similarly, we estimate the expanded volume of PFDs from the 1986 eruption at 4.6 × 107 m3 ± 0.62 × 107 m3, with a maximum thickness of ~31 m, and a mean of ~7 m.