ALOS-Indonesia POLinSAR Experiment (AIPEX): First Result*

Transcription

1 ALOS-Indonesia POLinSAR Experiment (AIPEX): First Result* Mahmud Raimadoya(1), Ludmila Zakharova(2), Bambang Trisasongko(1), Nurwadjedi(3) (1) Bogor Agricultural University (IPB), P.O. Box 2049, Bogor 16020, INDONESIA, (2) IRE Russian Academy of Sciences (RAS), Vvedensky sq., 1, Fryazino , RUSSIA, (3) Bakosurtanal, Jalan Raya Jakarta Bogor Km46, Cibinong, INDONESIA,

2 Overview 1. Introduction 2. Test Site 3. Study Approach 4. SAR-CAL 1. Faraday Rotation 2. Polarimetric Calibration 5. POL-InSAR 6. POL-SAR 7. Conclusions

3 1. Introduction Joint research Indonesia Russia - Japan : JAXA/ALOS Research Announcement 2 (RA2-402) Further improvement of ESA/Envisat AOE-869: Envisat-Indonesia Radar Biomass Experiment (EIRBEX), SAR-Cal/(Val), POL-SAR, POL-InSAR, POLSARPRO C-Band > L-Band; Dual-POL > Quad-POL 1 Test Site > 4 Test Sites (ALOS Systematic Observation Strategy + PLR/FBD Imaging Geometry) > ITS-A (Sumatera), ITS-B (Borneo), ITS- C (Papua), ITS-D (Java) Plantation Forest > Natural Forest + Rice Field SAR-Cal/Val : Institute of Radio-engineering and Electronics (IRE), Russian Academy of Sciences (RAS) POL-SAR, POL-InSAR : Radar Analysis Working Group, Bogor Agricultural University (IPB) National Mapping Coordination Agency (Bakosurtanal)

4 2. Test Sites

5 3. Study Approach SAR-CAL exercise was implemented as double check from user side to ensure PLR image calibration is OK for further analysis POL-InSAR analysis was implemented after SAR-CAL, and harmonization between ALOS strategic image acquisition, PLR imaging geometry and test site field data Upon identification of PLR site for NATURAL FOREST, this site will be used for 2009 PLR acquisition POL-SAR was concentrated for RICE.

6 4. SAR-CAL Scattering Matrix (M) [1] M M HH VH M M HV VV = δ 2 cos Ω sin Ω S S cos Ω sin Ω 1 δ jϕ HH HV 3 (, θ ) e δ 1 f1 sin Ω cos Ω SVH SVV sin Ω cos Ω δ 4 f 2 A r 1 M matrix notation M = Ae Φ R T R F SR F T Faraday rotation for one-way propagation (rad) [2] Ω = 4 ε 0 µ π 0 2 e m 3 2 f 2 * s F ( ϕ, λ ) * cos( θ ) * N () s ds *10 θ = 2 f 4 * F ( ϕ, λ )* cos( Ω) * TEC / cos( α ) (Supposing the magnetic field is constant in the range of altitudes km around maximum of electron concentration in ionosphere and inserting numerical values of constants used)

7 4. SAR-CAL

8 4. SAR-CAL

9 4. SAR-CAL (ITS-A)

10 4. SAR-CAL Faraday rotation of the scattering matrices values are corrupted (w/o equipment distortion matrices) M M M HV VH M HH = = VV S S = = VH S HV S HH VV cos Ω S ( S HH + SVV ) ( S + S ) HH 2 cos 2 VV Ω S VV sin 2 Ω sin Ω cos Ω sin Ω cos Ω HH sin 2 Ω Both co-polarized components become lower than nominal values as well as cross-polarized components now are not equal. As many calibration techniques are based on the target reciprocity (MHV=MVH), in such a conditions they cannot be applied for the estimation of the distortion matrices and subsequent correction of the measurements. On the other hand, Faraday rotation angle may be estimated correctly and removed if we have polarimetric SAR observations with instrument distortions removed (the calibration of the data is done preliminary).

11 4. SAR-CAL The approach selected by JAXA is based on the fact the PALSAR equipment may be calibrated if SAR measurements are made in Faraday rotation free conditions. For calibration puposes JAXA placed corner reflectors in Amazon rainforest area, near equator, because of observation geometry the Faraday effects are very small. The distortion matrices measured in the above mentioned approach are provided in the leader file and presented in Table 1 below.

12 4. SAR-CAL Table 1 (Leader File) Matrix element Re Im t t t t r r r r Quegan calibration algorithm [3] was used to evaluate cross-talk terms and calibration targets were used for the estimation of imbalance. These distortion matrices on transmit and receive were used to compensate the corrupting influence of equipment. All the polarimetric data delivered to users by JAXA are corrected for the instrument inaccuracies.

13 4.1 Faraday Rotation First we will estimate the level of Faraday rotation using model information about ionosphere properties for the location of Indonesia Riau test site and observation time. TEC values may be evaluated using International Reference Ionosphere model IRI2007 from site. Information about magnetic field (field intensity components, in nanoteslas) was provided by Earth geomagnetic field model WMM2005 obtained from The direction of SAR observation was calculated from ALOS state vector in SAR Leader file and PALSAR look angle. The models and observation parameters along with Faraday rotation estimation are presented in the Table 2 below. It should be noted that magnetic field components Fx, Fy and Fz here are respectively northward, eastward and vertical components.

14 4.1 Faraday Rotation Table 2. Ionosphere, Earth magnetic field, geometry parameters and Faraday rotation estimations. Date Latitude, degrees Longitude, degrees UTC, hrs TEC, m -2 Fx, nt Fy, nt Fz, nt Ω, deg 2θ, deg 07/05/ * /06/ * From here we can see that the two-way Faraday rotation angle is remarkably small and should not distort the normalized radar cross section (NRCS) or σ0 measurements significantly. As it is known, if the data are calibrated accurately, the level of Faraday rotation may be estimated from polarimetric data themselves (Table 3). Date Latitude, degrees Longitude, degrees UTC, hrs 2θ, deg 07/05/ /06/

15 4.1 Faraday Rotation PLR (L)

16 4.1 Faraday Rotation PLR (R)

17 4.2 POL-CAL Faraday rotation effect is very low in the analyzed PALSAR data (Indonesia), so we may try to check quality of polarimetric data using some popular techniques like as Quegan algorithm. Some popular free software packages, available in Internet, may be used for calibration purposes. For example, a number of polarimetric calibration techniques was implemented in POLSARPRO software. Among them are Quegan, Papathanassiou and Ainsworth algorithms. We have chosen Quegan algorithm and tried to estimate distortion matrices components in a way, which was done in POLSARPRO. The software was modified to generate range profiles of the scattering matrices elements as well as range distribution of surface scattering matrices elements. We used for our analysis the scene from The data was first change to uncalibrated (the distortion matrices from the Table 1 were restored in the data). 800 lines from the beginning of the scene were processed; the evaluation results were averaged so the range profiles were generated. As Quegan algorithm provides relative calibration, the r11 and t11 terms were set to be 1. The range plots of amplitude values for cross-talk terms on receive (r12,r21) and transmit (t12,t21) as well as imbalance on transmit t22 are plotted below.

18 4.2 Range Profile (R12, R21)

19 4.2 Range Profile (T12,T21,T22)

20 4.2. Range Profile Compare average of the terms from plots above with JAXA values (Table 1). The comparison results are presented in the Table 4. The values obtained are close to JAXA values, that means that the calibration numbers extracted of the PALSAR data for the test site are correct. At the same time, JAXA distorting matrices, which were used to calibrate PALSAR data, are applicable for calibration of PALSAR data obtained over Indonesia. There is no need to uncalibrate PALSAR data and to recalibrate them again.

21 4.2. Range Profile Table 4. Estimated average amplitudes of PALSAR distortion matrices elements. Matrix element Averaged range values JAXA value R R T T T

22 4.2. Sigma Nought The last calibration test is analysis of range distribution of σ0. In the case of homogeneous surface like forests, especially rain forests, this distribution is useful for antenna pattern measurements. After the antenna pattern was taken into account during the image generation, the remaining variations may serve as an indication of calibration problems.

23 4.2. Sigma Nought The 3 figures 8-10 below display the dependence of σ0 (db) for HH, HV(=VH), and VV with respect to range bin. Range values represent entire range line of PALSAR data in a given scene. There is no obvious trend in range for cross-polarized components. There is subtle decrease of σ0 in co-polarized components. It is within the specifications declared in [1], where 1 db absolute calibration error is acceptable for vegetation mapping and monitoring, when biomass density may be mapped with 25% accuracy and Leaf are Index within range 0-2.

24 5. POL-InSAR Test Site for POL-InSAR requires the harmonization among ALOS image acquisition strategy, PLR imaging geometry, availability of Field Data Test Site in Papua (ITS-C) was selected as POL-InSAR site POL-InSAR analysis for PLR was completed at interferogram stage. Further analysis will require next version of POLSARPRO (V.4) POL-InSAR analysis for FBD still in data order stage

25

26

27 6. POL-SAR

28 6. POL-SAR

29 6. POL-SAR

30 6. POL-SAR

31 6. POL-SAR (PSM)

32 6. POL-SAR (LARGE)

33 6. POL-SAR (LARGE)

34 6. POL-SAR (MEDIUM)

35 6. POL-SAR (MEDIUM)

36 6. POL-SAR (SMALL)

37 6. POL-SAR (SMALL)

38 6. POL-SAR (SCANSAR)

39 6. POL-SAR (PLR)

40 6. POL-SAR (FBD)

41 7. Conclusions In general, JAXA calibration quality of PALSAR data acquired over Indonesia test site is good for vegetation studies. Level of Faraday rotation in these PALSAR images is small for the goals of AIPEX (no corrupting effect). POL-InSAR analysis awaits POLSARPRO v.4 Indonesia Super Sites for PLR are identified to be requested for next 2009 cycle of PLR acquisition.

42 ACKNOWLEDGEMENT THE POLSARPRO TEAM 50 YEARS IND-JPN (2008)