Vol. 85, No.5, 3 February 2004

Updated Repeat Orbit Interferometry Package Released

Paul A. Rosen, Scott Hensley, and Gilles Peltzer, Jet Propulsion Laboratory, Pasadena, Calif.; and Mark Simons, Seismological Laboratory, California Institute of Technology, Pasadena

---

Copyright 2004 American Geophysical Union

ROI_PAC V2.3, a Repeat Orbit Interferometry package that allows researchers in the area of topography and surface change to apply Interferometric Synthetic Aperture Radar (InSAR) methods, is now freely available to the community. InSAR is the synthesis of conventional SAR techniques and interferometry techniques that have been developed over several decades in radio astronomy and radar remote sensing. In recent years, it has opened entirely new application areas for radar in the Earth system sciences, including topographic mapping and geodesy [e.g., Thompson et al., 1986; Massonnet and Feigl, 1998; Rosen et al., 2000].

ROI_PAC, developed primarily to work with European Remote Sensing (ERS) satellite radar data, currently supports ERS-1, ERS-2, and Japanese Earth Resources Satellite (JERS) radar data, and is configurable to work with "strip-mode" data from all existing satellite radar instruments. The method of SAR interferometry combines two full-resolution complex radar images into an "interferogram," the point-wise product of one image and the complex conjugate of the other. The phase of the interferogram encodes the topography and topographic change over the time spanned by the pair of images. NASA has sponsored the development of InSAR techniques and applications through its Topography and Surface Change and Solid Earth and Natural Hazards Programs, as well as radar projects such as SIR-C and the Shuttle Radar Topography Mission.

The first release of ROI_PAC (V1.0) was made quietly in 2000, and roughly 30 groups in the academic and research community currently use it. In response to a growing number of InSAR users worldwide, and after some substantial efforts to streamline the processing for general users, the software is ready for wider distribution to the geosciences community.

ROI_PAC uses raw radar data, ancillary information from telemetry and navigation solutions, and digital elevation models (DEM; externally provided or interferometrically derived) to produce a variety of derived data products, including the full-resolution images themselves, interferograms, phase images measured as principal value and continuously "unwrapped" DEMs, and error estimates. Each of the products is available in its natural radar coordinate system and geo-referenced to a DEM. The software computes the interferometric baseline---that is, the orbital separation of the satellite at the observation times---from the provided navigation solutions, and then refines the estimate to the millimeter level of precision using the DEM provided for reference. A deformation model can also be supplied to the package to remove large-scale phase effects that might bias the baseline estimate. The inclusion of a model is particularly important when studying broad deformation zones that have well-characterized secular rates. To remove the topographic signature from an interferogram, ROI_PAC simulates an interferogram from the orbit data and the DEM, and subtracts the phase from the measured interferogram, leaving just the deformation phase.

Technical Structure

ROI_PAC implements its fundamental algorithms in C and Fortran 90 and drives each executable module with a Perl control script. Higher-level Perl scripts control logical groups of functions. At the highest level, there are three Perl "recipes"---one to generate a geocoded deformation image from raw SAR data files, orbit information, and a DEM; another that generates a DEM from the SAR and orbit data; and a third that produces a geocoded deformation image using an additional set of radar data files as the source of topography. ROI_PAC supports the use of orbit data from ESA (PRC orbits) and Delft University in The Netherlands (http://www.deos.tudelft.nl/ers/precorbs) and relies on software from Delft for conditioning their orbit data. The high-level scripts control parameters that default to typical ERS settings, but the scripts accept an optional parameter file to override nearly all defaults.

ROI_PAC runs on Silicon Graphics Incorporated (SGI), Sun, Linux and Mac OS X platforms. On SGI and Sun machines, the code is designed to work with the companies' native compilers and math libraries. On Linux and Mac, ROI_PAC requires a commercial Fortran 90 compiler. ROI_PAC can be built on all platforms with the Fast Fourier Transform package (http://www.fftw.org) to perform operations.

Processing Algorithms

The ROI_PAC SAR processor uses a straightforward range-Doppler algorithm. The raw data are segmented into overlapped patches, and each patch is compressed in range and azimuth centered at the measured Doppler centroid. The centroid is chosen to be appropriate for a given interferometric pair, and the resolution is adjusted to filter energy out of the common spectral band in range and azimuth. The centroid is allowed to vary in range, but remains fixed for all patches.

After the interferogram is formed, it is unwrapped. ROI_PAC has two implementations of the traditional branch-cut algorithm [Goldstein et al., 1988]. One version operates on the entire interferogram and defines very simple rules for creating branch cuts. The other has more flexible methods of filtering and branch cut manipulation. It is also patch-oriented; segmenting the image is faster and helps localize unwrapping errors should they occur.

The geometry that relates the interferometric phase to the baseline is rigorously defined through a spherical coordinate system, where the surface of the sphere best fits the Earth's ellipsoidal shape below the spacecraft. From the unwrapped phase and the DEM, the baseline can be refined from 30-cm orbit data to millimeter precision.

The algorithm for simulating topographic fringes is quite straightforward. For each DEM coordinate, the location on orbit when the coordinate is imaged at the correct Doppler and range is calculated by a zero-finding routine. A record is kept of the mapping from DEM coordinate to range-Doppler coordinate, such that the deformation signal can be easily geocoded through a look-up table interpolation.

Given an unwrapped interferogram and a sufficiently accurate baseline, an accurate DEM can be generated. These topographic products can, in some cases, rival the accuracy of existing DEMs. Several algorithms for resampling the irregularly spaced points mapped from range-Doppler coordinates into the geo-referenced frame are provided, including simple nearest-neighbor point placement, bi-linear interpolation, surface fitting, and convolutional resampling. These algorithms have been used most recently in the processing of Shuttle Radar Topography Mission data. In addition to these main algorithms, ROI_PAC has a number of transformational and filtering utilities that facilitate data manipulation and interpretation.

Availability

ROI_PAC source code is available to the international community for research purposes at no charge from the NASA Open Channel software distribution system (http://www.openchannelfoundation.org). Caltech retains intellectual property rights to the software and licenses the source code and executable software for external use. Commercial enterprises interested in the software for profit-oriented endeavors must negotiate a commercial license with Caltech. Companies doing research for the government can obtain a research license with appropriate documentation of government sponsorship. ROI_PAC has received an official classification of EAR99 for the source code and object code from the U.S. Department of Commerce. An item classified as EAR99 is exportable without any government licensing requirement to foreign nationals of all but a few selected countries.

Future Plans

The developers are heavy users of ROI_PAC and attempt to fix software bugs as they arise. A ROI_PAC Web page at the Open Channel Foundation Web site hosts documentation and links to other pages and newsgroups for discussing bugs. It is hoped that the user community will work actively with the developers to report and repair bugs.

There is no desire to hold ROI_PAC to the standard of a well-maintained commercial package. It is research code for those who need the flexibility and access to all stages of processing for handling the inevitable uncharted territory in scientific exploration. The developers intend to improve the efficiency, modularity, and user-interactivity needed for solving big geophysical problems with large quantities of data provided by an array of international SAR systems. In the immediate future, ROI_PAC will be fitted with data conditioners for Europe's EnviSAT, Japan's Advanced Land Observing System (ALOS), and Canada's Radarsat-2 systems in anticipation of the exciting interferometric data they will provide.

Acknowledgments

ROI_PAC has been developed through the collaboration of many researchers over the past decades. From the initial experiments in spaceborne radar interferometry by R. Goldstein and colleagues [Zebker and Goldstein, 1986], through the initial attempts to tie functionality together in scripts by G. Peltzer and colleagues, ROI_PAC has benefited from the contributions of many people at the Jet Propulsion Laboratory and Caltech. A complete chronology is given with the package distribution. Notable contributors to ROI_PAC are: Paul Rosen, Scott Hensley, Gilles Peltzer, Francois Rogez, Frederic Crampe, Eric Fielding, Mark Simons, Rowena Lohman, and Matt Pritchard.

References

Goldstein, R. M., H. A. Zebker, and C. L. Werner, (1988), Satellite radar interferometry: Two-dimensional phase unwrapping, Radio Sci., 23, 713-720. 

Massonnet, D., and K. L. Feigl, (1988), Radar interferometry and its application to changes in the Earth's surface, Rev. Geophys., 36, 441-500. 

Rosen, P. A., S. Hensley, I. R. Joughin, F. K. Li, S. N. Madsen, E. Rodriguez, and R. M. Goldstein, (2000), Synthetic Aperture Radar Interferometry, Proc. IEEE, 333-382. 

Thompson, A. J. M. Moran, and G. W. Swenson, (1986), Interferometry and Synthesis in Radio Astronomy, 720 pp., Wiley-Interscience, New York. 

Zebker, H. A., and R. M. Goldstein, (1986), Topographic mapping from interferometric SAR observations, J. Geophys. Res., 91, 4993-4999.