md 
  
  
oking 
cision 
ges — 
| then 
> the 
bolic- 
usual 
i-pass 
; from 
loped. 
robust 
netric 
sr by 
ar and 
'oduct 
on for 
ooded 
ithern 
> river 
nd its 
uired. 
serve 
| thus 
elf is 
amps, 
    
which can be flooded seasonally. Especially, changes of the 
water level are an important factor for the unique fauna (fishes, 
amphibians, waterbirds) and flora of the nature reserve. These 
fluctuating and mostly falling water levels result from 
inappropriate farming practices and drainage impacts. 
Furthermore, falling water levels also affect human activities on 
the river Rhine, which is an important shipping route. Besides 
the monitoring of water levels, general changes of land use are 
also important because of the land consumption at the expense 
of natural areas (RAMSAR, 2012). 
The time series of TerraSAR-X data cover the period of half a 
year from autumn 2011 to spring 2012. Out of the database two 
acquisitions are chosen in order to demonstrate the new 
methodology: 2011-11-17 and 2011-12-09. Both images are 
taken in the dual-co-pol strip map mode. 
2. APPLICATION 
The new SAR change detection method roughly described 
above now is applied on the mentioned data sets acquired by 
TerraSAR-X over the Testsite “Upper Rhine”. The complex 
data is decomposed into Kennaugh elements, multilooked and 
geocoded with a pixel spacing of 4m, which corresponds to a 
look factor of about 4.3 nominal looks. The geocoded 
Kennaugh elements are combined to the sum and the difference 
of the image pair. In the first case the result is the mean 
reflectance for both acquisitions, in the ladder case the temporal 
changes between both acquisition times are emphasized. 
Finally all images are enhanced by the help of the pyramidal 
multilooking with an enhancement factor of -5dB — ie. a 
threshold of nearly 32% is introduced to the statistical 
modelling — and normalized in relation to the total intensity. 
2. Single Images 
Figs. 2 and 3 show quicklooks (QL) of the single image 
Kennaugh elements. The colours are composed as follows: 
R: KO — total intensity 
G: K3 - difference double bounce and surface scattering 
B: K4 — difference HH and VV intensities 
  
Figure 2: 2011-11-17 Figure 3: 2011-12-09 
Kennaugh elements QL Kennaugh elements QL 
International Archives of the Photogrammetry, Remote Sensing and Spatial Information Sciences, Volume XXXIX-B7, 2012 
XXII ISPRS Congress, 25 August — 01 September 2012, Melbourne, Australia 
               
The TANH-scaled elements are brought to an 8bit image format 
in order to give a first impression of the image content. 
2.2 Sum Images 
If the scattering information of both images is summed up, the 
mean reflectance is achieved, see quicklook in Fig. 4. Stable 
objects become sharper while the influence of non-stable 
objects is reduced. Fig. 5 shows the total intensity of the sum 
image. The polarimetric layers are depicted in Figs. 6 and 7. 
The colour coding starts from dark blue at -10dB, passes lighter 
colours — even gets transparent around zero — and reaches dark 
red for values around +10dB. The blue colours in Fig. 6 mark a 
dominant surface scattering in the natural land whereas double- 
bounce scattering dominates over man-made objects. 
  
Figure 5: Kennaugh 0 
total intensity 
Figure 4: QL of Mean 
Kennaugh elements 
   
eus. 
* 
s 
4 
  
  
rei 
   
tm, 
$ 
* 
  
  
  
  
  
  
  
  
     
  
  
  
OR E 
E o: OE 
MA 
es 
ps 
Figure 7: Kennaugh 4 
HH - VV intensities 
Figure 6: Kennaugh 3 
Double bounce — surface