The Impact of Temporal Decorrelation Over Forest Terrain in Polarimetric SAR Interferometry
Seung-Kuk Lee(1), Florian Kugler(1), Kostas Papathanassiou(1) and Irena Hajnsek(1)
(1) German Aerospace Center (DLR), P.O. Box 1116, 82234 Wessling, Germany
Polarimetric Synthetic Aperture Radar Interferometry (Pol-InSAR) is a recently developed radar technique based on the coherent combination of radar polarimetry and SAR interferometry. The potential of Pol-InSAR techniques in forest parameter estimation is based on the ability to separate volume from surface scattering contributions and to recover the vertical distribution of scatterers in mixed (volume) scattering scenarios. While Pol-InSAR techniques are today well established, a critical issue in the case of repeat-pass spaceborne measurements is temporal decorrelation, caused by changes within the scene occurring in the time between the acquisitions. Indeed, temporal decorrelation has been identified as the most critical factor for a successful implementation of Pol-InSAR parameter inversion techniques in terms of conventional space-borne repeat-pass InSAR scenarios. Similar to any other system induced decorrelation contribution, temporal decorrelation reduces the performance of Pol-InSAR techniques by biasing the volume decorrelation contribution used for parameter inversion. This leads to an increased standard deviation of the InSAR phase - for the same number of looks - and introduces a bias in the parameter estimates.
The amount of temporal decorrelation depends on the structure of the scatterer related to the used radar frequency and on the environmental processes occurring in the time during and between the interferometric acquisitions. Temporal changes occur - in general - within the scene in stochastic, spatial and temporal patterns and cannot be modeled with the required accuracy without detailed information about the (environmental) conditions over time between the two observations. While there is a relative good understanding of decorrelation rates for temporal baselines on the order of 30-45 days at C- and L-band provided by the data of the ERS and JERS missions, as well as for baselines on the order of 20-40min provided by airborne repeat-pass data sets (at C-, L- and P-band) there is poor understanding of the decorrelation levels expected at temporal baselines on the order of hours and days. In this paper first we analyse repeat-pass Pol-InSAR data acquired in the frame of dedicated experiments in order to quantify temporal decorrelation for temporal baselines in the order of hours to days up to 2 and 4 weeks at L- and P-band for two different forest types: temperate forests and boreal forests.
In June 2008 DLR’s E-SAR (Experimental Airborne SAR) collected fully polarimetric and interferometric SAR data over the temperate forest in Traunstein, Germany to investigate temporal baselines in the order of days and weeks .For tThis campaign (TempSAR) data have been acquired on six days in L- and X-band in the period from 07/06/2008 to 20/06/2008. DLR’s E-SAR system carried out three campaigns over Remningstorp forest in early March, early April and early May 2007. During these three dates data acquisition at L- and P-band in a repeat pass fully polarimetric mode were performed. We further assess the impact of the estimated temporal decorrelation levels on the performance of Pol-InSAR inversion techniques. Finally we attempt to propose a decorrelation model able to describe the estimated temporal decorelation behaviour observed at the different frequencies and use it to draw some conclusions about different acquisition scenarios.