TI
SIN
SID
ZN AS, Á
© ON
NN ESS NN (9) SN (©)
a AS N WN 5 + EN + MESS
AS NN GERS SAN > P ree
2X NW NS c AX SAY RRA Emre
© 46ST TN, 7 © RAIN ° DAD DASS
© | AK N NWS //7] o EIRE SES D ur ROA SS
8 ADS Soon] Loo 0 EERIE o LP P IA SN
CHE % ZH RAS \
2 CESSE) 7 IRN = ESS 2 A AS
' LR NNO = CESSE EN € SG ES
E — SM E E + EE] A
5 esse SOS " £a = SAAT 22777 À
2 SST = 3 EZ)
UNS 5 i25
| |
}
Y eU > UE
= + m +
(a) z (BEZ (b) ^
WW MEAN aS 222
- a ON + a RR + À 7
s AY SN 2» 2 Er 0000772777777 2 NN,
T yy NS Y Nr OU 3 OUI
23 ZZ ä ANNE $0 NN
3 RNIN Nez € NS ENS € SN
E + 24 8 vo RANK gu D
5 SAN NN OZ B QT SOU iz NA
o SCALIS iW d o S o 0)
= ASS =. = NGL
2 i X e
+
= E. y = Es -
= + = + ue +
Figure 3: Co- (top) and cross-polarisation (bottom) signatures for native eucalypts at (a) P-, (b) L- and (c)
C-bands measured by NASA/JPL AirSAR system.
nent is also responsible for the pedestals. In general,
therefore, the higher the pedestal is, the more the dou-
ble bounce scattering and the cross-polarised compo-
nents. The result that a significant part of the double
bounce scattering is decomposed at C-band, seems in
Table 4: Scattering components as a percentage of the
total HH backscattering response for ocean water at
the incidence angle of 35°.
conflict with the common understanding that C-band [Rand | Single % Double % Bragg % Error %
mainly interacts with the crown layer. There is a pos- P Ü 38 62 58
sible reasons for this result. Eucalypts in the site are L 49 43 7 0.0
sparse, and their crown closure is less than 25%. It is C 31 55 14 0.0
possible, therefore, that the double bounce scattering
between the underlying vegetation and the ground sur-
face is also captured. However, for whatever reason,
according to the theory, a significant double bounce It is understood that the Bragg scattering is frequency
scattering component will be decomposed if there is a dependent, and relates to the frequency of both gravity
significant pedestal in the signature. and breaking waves on the ocean surface. As shown in
Figure 4. only P-band captured a strong Bragg scat-
: tering component in this example.
Table 3: Scattering components as a percentage of the 8 R
total HH backscattering response for native eucalypts
at incidence angle of 45°. 4.4 Reconstruction of Polarisation Signa-
tures
Band | Single % Double % Bragg % Error % So far the decomposition of the HH response for var-
P 17 T] G 01 ious types of targets has been shown. The error be-
L 24 67 10 14 tween the measured values and the predicted values
C AT 44 9 0.6 is usually less than 5% in most cases. In fact the
backscattering response of an arbitrary polarisation
ellipse can also be retrieved from polarimetric SAR
data in addition to the linear polarisation of HH, VV
4: 2 and VH. To demonstrate the precision of the proposed
decomposition technique for any polarisation, polari-
Table 4 gives the percentages of the single, double sation signatures using the measured Mueller matrix
and Bragg scattering components decomposed from and the reconstructed Mueller matrix given by (18)
HH polarisations for the ocean water at P- L- and C- are compared. It is found that the error between the
bands. measured polarisation signature and the simulated po-
In this example, P-band predicts a strong Bragg larisation signature is less than 15% for most cases, at
scattering component, but less in the other two bands. all bands. Figures 5-7 show both the co- and cross-
200
International Archives of Photogrammetry and Remote Sensing. Vol. XXXI, Part B7. Vienna 1996
