Optimization of the Coherence Estimation Window Size Aiming at Growing Stock Volume Retrieval in Siberian Forest
Stelmaszczuk-Gorska, Martyna; Thiel, Christian; Schmullius, Christiane
Friedrich-Schiller University, Jena, GERMANY
The basis for the Interferometric Synthetic Aperture Radar (InSAR) system is observation of a target by the radar antenna from two slightly different positions and or at different times (Bamler & Hartl, 1998). The main output of InSAR processing is an interferogram. It consists of magnitude (correlation between images - coherence) and phase (interferometric phase). It is formed by multiplication of one SAR image with the complex conjugate of a second image (Hagberg et al. 1995). Before the interferogram is generated the co-registration of the images must be done to an accuracy of less than 0.1 pixel size (Askne et. al 1997). Additionally, because the images are acquired from slightly different angles, the data must be filtered in terms of the ground reflectivity. Afterwards the complex coherence is computed. In practice the coherence is obtained using estimators by spatial averaging within a two-dimensional window. Unfortunately, the estimation is biased and tends to overestimate low coherence (Wegmüller & Werner, 1995; Askne et al. 1997; Bamler & Hartl, 1998; Touzi et al. 1999; Lopez-Martinez & Pottier, 2007), which typically represents forest. That is why the larger the estimation window size used for the forest areas the lower the estimation bias and the true coherence value is approached. The complete bias cancelation is only achieved when an infinite number of samples is used. This is not possible in case of real SAR data due to the heterogeneity of large areas. Additionally, with larger window size the resolution gets coarser. Hence, the optimization of coherence estimation window size is a compromise between good spatial resolution and reliable estimation of coherence.
For the investigation of coherence estimation two forest territories in Central Siberia have been used as test sites. These areas belong to the southern taiga sub-zone of the boreal forest and constitute two Russian Siberian Federal districts: Krasnoyarsk Kray and Irkutsk Oblast. Phased Array type L-band Synthetic Aperture Radar (PALSAR) data from the Advanced Land Observing Satellite (ALOS) have been used to investigate the repeat-pass coherence estimation. The PALSAR datasets that have been used are in 1.1 processing level, which is a Single Look Complex (SLC) product. The data have been provided as complex data including phase history in slant range geometry.
The result of the research is the optimization of coherence estimation window size aiming at growing stock volume retrieval in the case of boreal forest in Siberia. First, after the SLC images were co-registered with sub-pixel accuracy, the baseline was estimated, the common band filtering was applied and an interferogram for each pair of ALOS PALSAR data was calculated. Based on the normalized interferogram the coherence was calculated using different dimensions of the coherence estimation window. The obtained results for the different window sizes were compared and analysed. It was observed that when implementing larger window sizes the values for a zero coherence reference area (unfrozen water body at calm conditions) were approaching zero, as it was expected. It was seen that usage of different window sizes did not change much the high coherence values. On the other hand clear differences were observed for the medium and low coherence values. For those values implementation of small averaging windows resulted in the high standard deviation with values greater than the actual coherence. It was observed based on preliminary results that after implementing window size using more than 600 samples the coherence bias was negligible.
This work has been performed within the GIONET project, funded by the European Commission, Marie Curie Programme, Initial Training Networks, Grant Agreement number PITN-GA-2010-264509. The ALOS PALSAR data were provided by Japan Aerospace Exploration Agency (JAXA) within the Kyoto & Carbon Initiative.
Askne J.I.H., Dammert P.B.G., Ulander L.M.H., Smith G., "C-band repeat-pass interferometric SAR observations of the forest", IEEE Transactions on Geoscience and Remote Sensing, 35:1, 25-35, 1997.
Bamler R., Hartl P., "Synthetic aperture radar interferometry", Inverse Problems, 14, R1-R54, 1998.
Hagberg J.O., Ulander L.M.H., Askne J., "Repeat-pass SAR interferometry over forested terrain", IEEE Transactions on Geoscience and Remote Sensing, 33:2, 331-340, 1995.
Lopez-Martinez C., Pottier E., "Coherence estimation in synthetic aperture radar data based on speckle noise modeling", Applied Optics, 48:4, 544-558, 2007.
Touzi R., Lopes A., Bruniquel J., Vachon P.W., "Coherence Estimation for SAR Imagery", IEEE Transactions on Geoscience and Remote Sensing, 37:1, 1999.
Wegmüller U., Werner C.L., "SAR Interferometric Signatures of Forest", IEEE Transactions on Geoscience and Remote Sensing, 33:15, 1995.