1) TanDEM-X data was used to characterize forests in Sweden and South Africa, finding that dual polarization data allows forest height estimation without prior information, achieving r^2=0.86 and RMSE=2.02m in Sweden. 2) Single polarization TanDEM-X data was also found to be sensitive to forest height in Sweden when combined with a DEM, achieving r^2=0.91 and RMSE=1.58m. 3) Temporal decorrelation over 3 seconds was found to impact coherence, and significant differences in radar penetration between acquisitions in December and July were observed and attributed to changing dielectric properties.

1. Forest Characterisation by means of TanDEM-X Pol-InSAR Data First Results & Experiments. Microwaves and Radar Institute (DLR-HR) German Aerospace Center (DLR) Florian Kugler, Astor Torano Caycoya, Irena Hajnsek, Kostas Papathanassiou

2. TanDEM-X Data Acquisition Modes Standard DEM Mode July until October 2010 Temporal baseline: 2-3 sec (20-30Km Across Track separation) one satellite transmits and both satellites receive simultaneously small along-track displacement required for Doppler spectra overlap requires PRF and phase synchronisation Bistatic transmitter alternates between PRF pulses provides three interferograms with two baselines in a single pass enables precise phase synchronisation, calibration & verification both satellites transmit and receive independently susceptible to temporal decorrelation & atmospheric disturbances no PRF and phase synchronisation required (backup solution) Alternating Bistatic Pursuit Monostatic

3. - Forest type: boreal - Tree Species: Pine, Spruce, Birch - Forest Height: ~ 30m - Terrain: Hilly BioSAR II: 2008 Test site: Krycklan, Sweden BioSAR II: Krycklan, Sweden Forest heights: Up to 30m - (Mean ~ 17m) Biomass range: Up to 220t/ha - (Mean ~ 90t/ha) Pine, Spruce, Birch & Mixed stands. Hilly Terrain / Steep local slopes Uniform structure

4. Test Site: Krycklan, Sweden HH VV DEM 38m 41m 35m Height of ambiguity HH / VV HH / HV HH / VV Polarisation 0.15 32° 125 08. August 2010 0.17 0.18 K Z 32° 137 19. August 2010 32° 141 28. July 2010 Incidence angle Baseline [m] Date

5. 28. July / kz=0.18 8. August / kz=0.15 19. August / kz=0.17 Interferometric Coherence: HH

6. Temporal Decorrelation 28. Juli K Z = 0.18 08. August K Z = 0.15 19. August K Z = 0.17 3 seconds may be enough !!! > two acquisitons are affected by temporal decorrelation Interferometric Coherence HV 19. August / kz=0.17 HH HV

7. SNR correction Coherence VV VV HH HH Terra SAR-X TanDEM-X NESZ Polynomials (from xml files)

8. SNR correction Coherence VV VV HH HH Terra SAR-X TanDEM-X NESZ Polynomials (from xml files)

9. 28. July 2010 K Z =0.17 HH VV SNR correction Coherence After calibration: 3% (HH) to 5%(VV) of the coherences > 1. Assuming SNR lower by 1 dB reduces the coherences > 1 to 0.6%.

10. TanDEM-X Data Acquisition Modes Standard DEM Mode one satellite transmits and both satellites receive simultaneously small along-track displacement required for Doppler spectra overlap requires PRF and phase synchronisation Bistatic transmitter alternates between PRF pulses provides three interferograms with two baselines in a single pass enables precise phase synchronisation, calibration & verification both satellites transmit and receive independently susceptible to temporal decorrelation & atmospheric disturbances no PRF and phase synchronisation required (backup solution) Alternating Bistatic Pursuit Monostatic

11. Test Site: Krycklan, Sweden HH Interferometric Coherence HH 125m 69m Height of ambiguity HH HH Polarisation 0.05 19° 39 11. June 2011 0.09 K Z 19° 69 17. December 2010 Incidence angle Baseline [m] Date

12. Test Site: Krycklan, Sweden HH 12.2010 HH 06.2011 125m 69m Height of ambiguity HH HH Polarisation 0.05 19° 39 11. June 2011 0.09 K Z 19° 69 17. December 2010 Incidence angle Baseline [m] Date

13. Test Site: Krycklan, Sweden HH 06.2011 Interferometric Coherence HH Interferometric Coherence HH

14. VV Jul θ =32° r²=0.65 / RMSE = 8.27m HH Jul θ =32° r²=0.54 / RMSE=9.45m HH Dec θ =19° r²=0.62 / RMSE = 11.80m HH Jun θ =19° r²=0.61 / RMSE = 9.77m Larger @ HH than @ VV Larger in Winter than in Summer Penetration Depth @ X-band Less sensitive to incidence angle Number of stands: 216 Phase Center Height Phase Center Height Phase Center Height Phase Center Height

15. Pol-InSAR Phase Difference Pol-InSAR Coherence Region Phase Center Difference Amplitude / Lidar Heights 6m 0m 30 m 0m

16. 2 Layer Scattering Model Volume Coherence Interferometric Coherence Volume Coherence Vertical Wavenumber: G/V Ratio: Volume Height Extinction Topography G/V Ratio Volume Coherence Vertical Wavenumber: G/V Ratio: Volume Height Extinction Topography G/V Ratio

17. 1 Layer Scattering Model (m=0) Volume Coherence Volume Coherence Vertical Wavenumber: G/V Ratio: Volume Height Extinction Topography G/V Ratio Volume Coherence Vertical Wavenumber: Volume Height Extinction Topography Single Channel X-band Inversion Interferometric Coherence

18. Probably Harvested Lidar Data: 2008 Radar Heights Single Pol Inversion Amplitude / Lidar Heights r² = 0.91 - r² = 0.93 (harvested area removed) RMSE = 1.58 Forest Height Estimation: 1 Pol + DEM 30m 0m

19. 2 Layer Scattering Model Volume Coherence Interferometric Coherence Volume Coherence Vertical Wavenumber: G/V Ratio: Volume Height Extinction Topography G/V Ratio Volume Coherence Vertical Wavenumber: G/V Ratio: Volume Height Extinction Topography G/V Ratio

20. Radar Heights dUAL Pol Inversion Amplitude / Lidar Heights Forest Height Estimation: 2 Pol Probably Harvested r²=0.86 / RMSE = 2.02m r²=0.93 / RMSE = 1.44m 30m 0m

21. 3 seconds temporal baseline can decorrelate interferometric coherence SNR limits inversion performance - needs to be corrected Single Pol TanDEM-X data (VV) are sensitive to forest height (with a priori ground phase r²=0.91, RMSE = 1.58) Dual Pol TanDEM-X data (HH, VV) allow forest height estimation without any a priori information – at least for boreal forests as found in Krycklan test site (r²=0.86, RMSE = 2.02) Forest height estimation is possible using TanDEM-X data Significant difference in penetration depth from December to Juli acquisition: Inducesd by: changing dielectric constant Single Pol Forest Height + DEM Dual Pol Forest Height Conclusions

22. Forest Characterisation by means of TanDEM-X Pol-InSAR Data First Results & Experiments. Microwaves and Radar Institute (DLR-HR) German Aerospace Center (DLR) Florian Kugler, Astor Torano Caycoya, Irena Hajnsek, Kostas Papathanassiou

26. Krüger National Park Savanna forest Single trees no closed canopy Tree heights up to 20m

27. Krüger National Park Amplitude HH Coherence HH DEM (200-450m) Dual-pol: / HH-VV, Incidence: 40deg, kz=0.1 0 [m] 6 0 [m] 30