POLARIMETRIC SAR CALIBRATION VALIDATION FOR ALOS/PALSAR IN MONGOLIA

Motoyuki SATO(1), Koichi IRIBE(1), D. ANARSAIKHAN(2)

(1)Center for Northeast Asian Studies, Tohoku University, Sendai 980-8576, Japan, Email : sato@cneas.tohoku.ac.jp

(2)Institute of Informatics and RS, Mongolian Academy of Sciences, av.Enkhtaivan-54B, Ulaanbaatar-51, Mongolia, E-mail: amar64@arvis.ac.mn

ABSTRACT The Japanese remote sensing satellite ALOS/PALSAR was launched successfully in January 2006. We are carrying out polarimetric calibration of the PALSAR sensor, which has the full-polarimetric function of L-band SAR. We set corner reflectors in Ulaanbaatar in May and August 2006, and the acquired data was used to confirm the polarimetric calibration coefficients, which will be used for the data products. In addition, the acquired polarimetric SAR data will provide new possibility in remote sensing. Since this is the world first satellite borne full-polarimetric SAR system. Full-polarimetric SAR data acquired over Mongolia show some interesting features. Temporal changes can be clearly detected, and also the data can be used for estimation of terrain slope. 1. INTRODUCTION JAXA (Japan Aerospace Exploration Agency) launched the earth observation satellite ALOS “Daichi”, which is equipped with three independent sensors in January 2006. One of the sensors is “PALSAR” which is a full polarimetric SAR, which is the world first full polarimetric SAR system operated regularly on a space craft for a long term. PALSAR operates in L-band (1.27GHz) and is capable to observe the ground surface condition accurately compared to other SAR operating at higher frequency band, namely C-band or X-band. Therefore, we have a hope that ALSO/PALSAR will be used for various environmental observations. The ALOS data products are now available in public domain, after sensor calibration. We jointed polarimetric calibration validation missions for ALOS/PALSAR and have deployed corner reflectors. We selected Ulaan Baatar as one of the calibration sites, because we think ALOS data will be quite usefully used for environment studies in Mongolia. Mongolia has various different ground and environmental conditions. I south, Gobi dessert shows bear soil, and in the middle, we have flat grassland, and in north, it is a typical boreal forest area. In this paper, we describe the calibration operation of the ALOS/PALSAR and show some first polarimetric SAR images in Mongolia acquired by ALOS/PALSAR. Then we also show some fundamental analysis using the polarimetric data for environmental evaluation.

2. POLARIMETRIC CALIBRATION 2.1 Test site The first data acquisition for polarimetric calibration was conducted on 25 May, 2006. We sat corner reflectors in a flat area inside the Chingis Khaan International airport, which is located west of Ulaan Baatar city (N47.836,E106.770). We selected this site, because the topography is flat. The site is selected along the run way of the airport as shown in Fig.1. The ground surface was covered by grass, having about 30cm height. At the same time, we conducted Ground Penetrating Radar (GPR) measurement around the test site, in order to measure the ground moisture condition. The second ALOS/PALSAR data acquisition was carried out on 25 August 2006. In this time, we settled the corner reflectors in grass field close to the international air port. The height of the grass was less than 30cm, and was sparse in this test site. The flight path of this data acquisition is almost the same as that in May 2006. 2.2 Corner Reflectors We prepared four different types of corner reflectors, which include a dihedral and three different sizes of trihedral corner reflectors as shown in Table.1.

Chingiskhaan AirportChingiskhaan Airport

Fig. 1 Test site for the corner reflectors in Chingis Khaan International Airport, Ulaan Baatar.

Table 1. Specifications of corner reflectors

In the data acquisition in May, we settled 4 corner reflectors (CR) in this test site. At this time, we did not use the large trihedral corner reflector. Four reflectors were aligned along a straight line parallel to the run way, and each CR was separated 100m away from the other. In August, we used 5 corner reflectors. In this time, corner reflectors were settled in square position, having about 100m separation to each other.

Type Size (m)

RCS2(dBm )

Number

(1) Dihedral 2 38.7 1

(2) Small

Trihedral 1.4 18.8 2

(3) Middle

Trihedral 2 24.9 1

(4) Large

Trihedral 2.5 34.7 1

(a) Dihedral corner reflector

(b) Small trihedral corner reflector

(c) Large trihedral corner reflector

Fig. 2 Corner reflectors

(a) HH+VV (Single bounce)

(b) HH-VV (Double bounce)

(c) HV (Volume scattering)

Figure 3 Pauli decomposed polarimetric

SAR image at the calibration site

3. POLARIMETRIC ANALYSIS 3.1 Polarimetric SAR Informatioon Fig.3 show the polarimetric SAR images acquired in May, which are represented using the Pauli decomposition. The single bounce, the double bounce and the volume scattering components are shown in Figs. (a), (b) and (c) respectively. We can find the location of 3 dihedral corner reflectors in Fig.3 (a), which is the single bounce component of the polarimetric SAR image. The dihedral corner reflector can be seen in Fig.3 (b), which is the double-bounce component of the polarimetric SAR image. The background scattering is stronger than we have expected, and we think it is due to the shallow off-nadir angle 21.5° which was used in this data acquisition. 3.2 Evaluation of Polarimetric Scattering from CRs We could determine the location of the corner reflectors from the polarimetric SAR images. Then we picked up the polarimetric scattering coefficients from the SAR images in order to confirm the polarimetric characteristics of the SAR images. The polarimetric SAR images we have obtained from JAXA were already calibrated using the polarimetric compensation matrix, and the coefficients shown in the flowing sections are the original data we received from JAXA. The theoretical scattering matrix of a trihedral and dihedral corner reflectors are given as:

1 00 1T⎡ ⎤

= ⎢ ⎥⎣ ⎦

S (Trihedral) (1)

1 00 1D⎡ ⎤

= ⎢ ⎥−⎣ ⎦S (Dihedral) (2)

The corresponding polarimetric signature of the trihedral and dihedral corner reflectors are shown in Fig.4. The measured scattering matrix of the corner reflectors are:

1 -0.3385 - 0.0161i-0.2437 + 0.0175i -1.0358 + 0.0598i⎡ ⎤

= ⎢ ⎥⎣ ⎦

S

(Dihedral corner reflector)

1 -0.0054 - 0.0742i-0.0544 + 0.1092i 1.0315 + 0.0314i⎡ ⎤

= ⎢ ⎥⎣ ⎦

S

(Small trihedral corner reflector)

1 0.0703 - 0.0198i-0.0602 + 0.0963i 1.2227 + 0.0235i⎡ ⎤

= ⎢ ⎥⎣ ⎦

S

(Small trihedral corner reflector)

1 0.1222 + 0.0482i-0.1156 - 0.0530i 1.1496 - 0.0159i⎡ ⎤

= ⎢ ⎥⎣ ⎦

S

(Large trihedral corner reflector)

The corresponding polarimetric signature of the trihedral and dihedral corner reflectors are shown in Figs. 4 and 5. We could confirm that in all the corner reflectors, the measured scattering matrix correspond well to the theoretical ones. Therefore, we think that the polarimetric calibration procedure by JAX is almost satisfactory. We obtained the Multi-Look (4 Look) image fro the single look images and the scattering matrix of the corner reflectors were determined as follows:

1 -0.3174 + 0.0054i-0.2421 - 0.0308i -0.9966 - 0.0645i⎡ ⎤

= ⎢ ⎥⎣ ⎦

S

(Dihedral corner reflector)

1 0.0549 - 0.1282i-0.0521 + 0.0714i 1.1376 + 0.2428i⎡ ⎤

= ⎢ ⎥⎣ ⎦

S

(Small trihedral corner reflector)

1 -0.0599 + 0.0018i-0.1652 + 0.1558i 1.0440 - 0.5829i⎡ ⎤

= ⎢ ⎥⎣ ⎦

S

(Small trihedral corner reflector)

1 0.0807 - 0.0164i-0.0828 + 0.0233i 1.2915 - 0.1245i⎡ ⎤

= ⎢ ⎥⎣ ⎦

S

(Large trihedral corner reflector)

We do not find significant differences, but it seems that the data quality is slightly worse compared to the single look data sets. It may be due to the limitation of the resolution of the images. The nominal resolution of the ALOS/PALSAR in full polarimetric mode is 20x30m, and it is larger than the physical size of the corner reflectors. Therefore, in the multi-look data, the polarimetric characteristics pf the corner reflectors were lost.

4. APPLICATION 4.1 Environmental studies

Fig. 6 shows the polarimetric SAR image of Ulaan Baatar area acquired in May 2006. The city is dominated by the HH and VV components, while the mountain area is dominated by HV. We can observe the very fine structure of the ground surface topography in this image. By comparing the SAR image in different seasons and different years, we can observe the temporal change of the condition of vegetation, and also the development of the city. At the same time, with the combination of GPR, we are planning to use this data for ground moisture content estimation.

Fig.6 PALSAR image of Ulaan Baatar city

(HH-VV: Red, 2HV: Green, HH+VV: Blue)

Fig.7 PALSAR data acquisition area over Mongolia

(a) Theoretical

(b) Measured Fig. 4 The polarimetric signature of a dihedral corner

reflector

(a) Theoretical

(b) Measured Fig. 5 The polarimetric signature of a trihedral

corner reflector

5. DETECTION OF TEMPORAL CHANGES Fig.7 shows the PALSAR data scenes acquired in May and August. It covers quite wide area of Mongolia including boreal forest, grassland and dessert. May is early spring, and grass is still very young, although there is no snow except high mountains. August is the end of summer, and the height of grass is most high in this season. The conditions of trees in the two seasons are almost the same. Figs. 8 and 9 show PALSAR Pauli decomposed images in north part of Mongolia. Most of the area shown in these images is shallow smooth mountains. We can find small villages in the middle. Changes on the ground use are small in general. However, in the polarimetric image, we can find clear change in agricultural area. Agriculture

in Mongolia is not very popular, but the we can find that the agricultural activities can be clearly detected by PLASAR images. 6. ESTIMATION OF GOUND SLOPE Fig. 10 shows the estimation result of the terrain slope, which was acquired by using the polarization orientation angle shift. The ground surface of the observed area is general smooth slope, having very sparse grass, therefore it is almost bare soil. We believe this condition is quite ideal for many theoretical studies of application of polarimetric SAR.

Scene ID: 2580

May 25, 2006 August 25, 2006

HH-VV2HV

HH+VV

Scene ID: 2580

May 25, 2006 August 25, 2006

HH-VV2HV

HH+VV

Fig.8 Temporal changes observed in PALSAR image

(HH-VV: Red, 2HV: Green, HH+VV: Blue)

7. CONCLUSION In this paper, we demonstrated the first full polarimetric SAR image acquired in Ulaan Baatar are by ALSO/PALSAR. We observed the polarimetric chrematistics of the image, and by analyzing the polarimetric scattering from the trihedral and the dihedral corner reflectors, we could confirm the validity of the polarimetric calibration processed in this datasets. We are also planning to use TerraSAR-X in the same area for the combination of L-and X-band polarimetric SAR study. Acknowledgements Part of this work was supported by JSPS Grant-in-Aid for Scientific Research (S) 18106008. We thank JAXA for providing ALSO/PALSAR data through the

Calibration and Validation of PALSAR program by JAXA. REFERENCES 1. J.S. Lee, D. L. Shuler, and T. L. Ainsworth,

“Polarimetric SAR data compensation for terrain azimuth slope variation,” IEEE Trans. Geosci. Remote Sensing, vol. 38, pp.2153-2163, Sept. 2000.

2. J.S. Lee, D. L. Shuler, and T. L. Ainsworth, E. Krogager, D.Kasilingam and W. M. Boerner, “On the estimation of radar polarization orientation shifts induced by terrain slope,” IEEE Trans. Geosci. Remote Sensing, vol. 40, pp.30-41, Jan. 2002.

3. D. L. Schuler, J. S. Lee, T. L. Ainsworth, and M. R. Grunes, “Terrain topography measurement using multipass polarimetric synthetic aperture radar data,” Radio Sci., vol. 35, pp. 813-832, May-June 2000.

Scene ID: 2590

May 25, 2006 August 25, 2006

HH-VV2HV

HH+VV

Scene ID: 2590

May 25, 2006 August 25, 2006

HH-VV2HV

HH+VV

Fig.9 Temporal changes observed in PALSAR image

HH-VV, 2HV, HH+VV 3 3pixels× 9 9 pixels×Polarization Orientation Angle

HH-VV, 2HV, HH+VV 3 3pixels× 9 9 pixels×Polarization Orientation Angle

Fig.10 Terrain slope estimation using polarization orientation angle shift.