Tropical forest tree height retrieval with TanDEM-X: Algorithm development and accuracy analysis
Antropov, Oleg; Rauste, Yrjö; Häme, Tuomas
VTT Technical research Centre of Finland, FINLAND

This paper details work performed on forest tree height retrieval from interferometric TanDEM-X data in the project ReCover funded by European Comission in Framework Programme 7.

The main objective of the TanDEM-X mission is to produce a consistent DEM of the Earth which is homogeneous in quality and unprecedented in accuracy, from the bistatic X-Band SAR interferometric acquisitions. The mission provides single-pass multi-polarization interferometric data acquired from space, which allows the acquisition of data on global-scale with the minimized disturbing effect of temporal decorrelation. The acquisition geometry (effective baseline, incidence angle, along and across track separation etc.) also increases the complexity of the SAR system characteristics, resulting in complex SAR instrument with different sensitivities to different imaging target related interferometric observables. One of possible applications relevant for tropical forest mapping in terms of the ReCover project is derivation of forest tree height with some potential indicated already in earlier studies (Hajnsek et al., 2009, Kugler et al., 2010).

While longer wavelength is more suitable for forest height mapping (at L- and P-band), some encouraging results have been demonstrated already even at X-band with airborne sensors. However, as difference in interferometric phase centre height location provided by use of different polarisations is relatively small, as well as single-polarization mode being more feasible option for InSAR-based mapping, an external DEM is usually needed for analysis. Then forest tree height can be estimated either 1) via simple DEM differencing (Izzawati et al., 2006, Balzter et al., 2007), or 2) by supplying an external DEM to model-based inversion methods.
Among later techniques the so-called Random Volume over Ground (RVoG) model (Treuhaft & Siqueira, 2000, Papathanassiou & Cloude, 2001) has gained popularity, primarily to its simplicity. One of the RVoG-based inversion algorithms for tree height estimation was developed at the early part of the ReCover project (Praks et al., 2012a). The approach developed uses an idea of correction for penetration depth into forest canopy at X-band, and offered an effective technique for measuring forest extinction over selected areas of relatively sparse forest. The inversion implemented was based on RVoG-based inversion of single-polarization single-pass InSAR data at X-band acquired over boreal forest. The accuracies obtained (RMSE) from airborne interferometric SAR data were approaching 1.5 meters with average tree height of about 20 meters after correction for penetration depth was performed.
The applicability of this approach was further tested over the same area with TanDEM-X dual- and single-polarization Coregistered Single look Slant range Complex (CoSSC) data (see Praks et al., 2012b). The accuracies were not as high compared to airborne InSAR data, with RMSE exceeding 3 meters. Detailed analysis was complicated by several years of discrepancy between the TanDEM-X data (acquired in autumn 2011) and Lidar data (collected during 2008-2009) that was used for cross-validation, but qualitative evaluation confirmed general applicability of this approach.

For the purpose of this study, some experimental TanDEM-X mission data (20 dual- and single-pol CoSSC pairs) were acquired over a test site in Chiapas, Mexico, with acquisitions covering the spring to summer period of 2012. Usual interferometric analysis includes complex coherency calculation followed by analysis of its magnitude and phase. Latter includes earth flattening, phase unwrapping and further analysis of variations of interferometric phase. The detailed analysis at this stage was complicated by absence of accurate reference (auxiliary) digital elevation model and canopy height model, to be obtained later. This limited assessment of results in Mexico at current stage to basic quantitative analysis, with an example of acquired imagery and derived tree height variations shown in Fig. 1. Analysis of interferometric phase centre variations at the borders between forest patches and agricultural areas reveals estimate of tree height of 15-35 meters in the area of interest. Further analysis will be performed at the later stage of the Recover project with reference tree height and/or forest biomass data and reported in the symposium paper.


Fig. 1. Example of TanDEM-X data over test site in Mexico, fragment of imagery in the slant range: (a) InSAR coherence amplitude with interferometric fridges; (b) earth-flattened InSAR phase height over AOI in the slant range geometry; (c) initial analysis of InSAR phase variations (modulus of 1st order difference of earth-flattened InSAR phase height)

References:

BALZTER H., LUCKMAN A., SKINNER L., ROWLAND C., and DAWSON T., 2007, Observations of forest stand top height and mean height from interferometric SAR and LiDAR over a conifer plantation at Thetford Forest, UK. International Journal of Remote Sensing, 28, pp. 1173-1197.
HAJNSEK, I., KUGLER, F., LEE, S. K., and PAPATHANASSIOU, K. P., 2009, Tropical-forest-parameter estimation by means of Pol-InSAR: The INDREX-II campaign, IEEE Transactions on Geoscience and Remote Sensing, 47, pp. 481–493.
IZZAWATI, WALLINGTON E. D., and WOODHOUSE, I. H., 2006, Forest height retrieval from commercial X-band SAR products. IEEE Trans. Geosci. Remote Sens., 44, pp. 863-870.
KUGLER, F., SAUER, S., LEE, S.-K., PAPATHANASSIOU, K., and HAJNSEK, I, 2010. Potential of TanDEM-X for forest parameter estimation. Proc. EUSAR, pp. 178-181.
PAPATHANASSIOU K. P., and CLOUDE S. R., 2001. Single-baseline polarimetric SAR interferometry. IEEE Trans. Geosci. Remote Sens., 39, pp. 2352-2363.
PRAKS, J., ANTROPOV, O., and HALLIKAINEN, M.T., 2012a, LIDAR-aided SAR interferometry studies in boreal forest: scattering phase center and extinction coefficient at X- and L-Band. IEEE Trans. Geosci. Remote Sens., 50, pp. 3831-3843.
PRAKS J., HALLIKAINEN, M., ANTROPOV, O., and MOLINA, D., 2012b. Boreal forest tree height estimation from interferometric TanDEM-X images. Proc. Intl. Geosci. Remote Sens. Symp., pp. 1262-1265.
TREUHAFT, R. N., and SIQUEIRA, P. R., 2000, The vertical structure of vegetated land surfaces from interferometric and polarimetric radar. Radio Sci., 35, pp. 141-177.