First Terrasar-x Interferometry Evaluation

The German radar satellite TerraSAR-X was launched in June 2007 [1] and is currently ending its commissioning phase. We anticipate quite different interferometric application scenarios compared to ERS1/2 and ASAR due to the X-band frequency, the short orbital repeat cycles of 11 days, the high range resolution and the spotlight mode of this sensor. During the commissioning phase we have scheduled a number of acquisitions over selected test sites with different characteristics to get an early quick look of TerraSAR-X's interferometric capabilities and to assess the phase quality of the sensor and DLR’s processor system [2]. Our first results are quite encouraging and the technical parameters of the system are as specified. Many spectacular image details let us expect that the high resolution will demand a different view on SAR interferometry and allow new applications in urban environments. In our paper we show interferograms and images of different test sites, coherence measurements and a first assessment of the interferometric properties. We will give hints to future scientific users on data selection and data processing. The results are of high relevance for the TanDEM-X mission scheduled for 2009, when a second compatible SAR-sensor will be launched for a joint 3 year bistatic interferometric formation flight. 1 SENSOR AND INSAR PROCESSING DETAILS The TerraSAR-X satellite has been launched on 15 of June 2007. It provides high resolution and short wavelength SAR imagery at a repeat cycle of only 11 days [3] allowing new interferometric applications. Most interesting for interferometry are the strip map, the spotlight and the high resolution spot light mode of the sensor. The tables 1-3 list some characteristic parameters of the different acquisition modes in single polarization. Table 1: Strip map mode data range bandwidth 150 MHz scene coverage azimuth 50 km scene coverage range 30 km azimuth resolution 3.3 m slant range resolution 1.2 m Table 2: Spotlight mode data range bandwidth 150 MHz scene coverage azimuth 10 km scene coverage range 10 km azimuth resolution 2 m slant range resolution 1.2 m Table 3: High Resolution Spotlight mode data range bandwidth 150 MHz / 300 MHz scene coverage azimuth 5 km scene coverage range 10 km / 6 10 km azimuth resolution 1.1 m slant range resolution 1.2 m / 0.6 m Different algorithms of the interferometric processing need to be adapted in actual processing chains in order to cope with the high resolution and the spotlight acquisition principle. E.g. the coregistration procedures used with the current sensors are not sufficient any more. Typically, a misregistration in the order of a 10 of a sample can be tolerated otherwise the interferometric coherence is reduced [4]. For the sensors ERS and Envisat ASAR the topography induced pixel shift can be modelled by low order polynomials for the whole scene. Fig. 1 and Fig. 2 visualize the sensitivity of the coregistration regarding topography and the error introduced by a polynomial modelling of the mutual shifts. The high resolution of the sensor TerraSAR-X results in a severe impact of the topography on the mutual shift of the master and slave samples [5]. Fig. 3 and Fig. 4 visualize that the topography induced mutual pixel shift and that their polynomial modelling results in a misregistration in the order of one sample leading to severe loss of coherence and can therefore not be tolerated any more. DLR’s interferometric system GENESIS uses a coregistration based on the observation geometry [6]. The mutual shift is estimated for each pixel utilizing a DEM and the precise orbit information. Fig. 8 provides an example for the performance of the algorithm compared to a conventional coregistration. Fig. 1: Topography induced sample shift over range for the sensor ERS; the green line is the optimal polynomial modelling of the shifts Fig. 2: Error caused by polynomial modelling of the shifts for ERS (difference between black and green line in Fig. 1). The deviation in the order of 10 sample can be tolerated. The spotlight acquisition is characterized by an azimuth varying Doppler frequency which is caused by the steering of the radar beam. Fig. 5 visualizes a typical Doppler spectrum. Algorithms like the resampling and spectral shift filtering need to be adapted to cope with such spectra. The TerraSAR-X product provides the Doppler frequency polynomials describing the raw data [7]. However, the interferometric processing is based on focussed SSC data and the azimuth time annotation RAW t of the Doppler polynomials needs to be corrected to zero Doppler coordinates using the FM-rate: