The International Archives of the Photogrammetry, Remote Sensing and Spatial Information Sciences. Vol. XXXVII. Part Bl. Beijing 2008 
153 
In the two times of acquisition, the wind speed was low and 
approximately equal. This fact caused nearly same horizontal 
gradient in air pressure (Mobasheri, 2006). In addition, no 
extraordinary Ionospheric phenomenon was detected. So the 
change gradient of Ionospheric parameters could assume the 
same. 
In order to above discussion, water vapor and liquid water are 
the remained factors to be considered. But as a test, the 
differential map of air pressure will also be considered. 
Figure 1 illustrates a part of the formed interferogram. As it 
could be seen (in A), a brief subsidence signal exists in this area. 
GPS and geological data also detect the deformation in the 
centre of this image. But InSAR shows deformation signals in 
surrounding points of the subsidence area which is not detected 
by GPS and geological surveys. This inhomogeneous and 
somehow wave-shaped signal which has around 4 centimeter 
disagreement with GPS points could be formed due to the 
atmosphere. There is around of a few centimeter disagreement 
between GPS and InSAR in subsidence area even after the 
flattening of the InSAR image (Figure 2). This is the motivation 
of the test of atmospheric correction strategies. 
Figurel: Raw interferogram 
Figure2: Flattened Interferogram 
Cloudy pixels cause a very low reflectance in the 14 th band of 
MERIS and this results in an approximately zero-estimation in 
water vapor map. On the other hand, over cloud vapor 
estimations are not acceptable due to the high influential under 
clouds water vapor (Mobasheri, 2006). Interpolation could be an 
appropriate solution for this problem (Li et al., 2005). 
A part of cloudy water vapor scene, acquired in 12 th of 
September could be seen with the corresponding cloud extracted 
map in Figure3. 
Figure3: Cloudy water vapor map (Left), Cloud Map (Right) 
Overlaying the cloud layer and water vapor map shows that the 
carried out algorithm for cloud extraction is not so much 
sensitive in mixed pixels. Water vapor amount of the edge of 
clouds was low either. Hence a pixel buffer applied to the cloud 
layer to avoid the underestimation caused by these wrong pixels 
in interpolation process. 
In this step at first, processed cloud layer of MERIS data was 
overlayed to the Water vapor layer. Then a 15 pixels 
neighborhood was utilized to interpolate the central cloudy pixel 
by non-polluted pixels using weighted distance algorithm (Li, 
2005). As central parts of large clouds or inside of the congested 
clouds may not obey the rule of smooth changes in water vapor 
amount, cumulus clouds and a more than 40 percent cloudy 
15x15 neighborhood window were masked. The masked points 
of a single image had to be neglected in other images either. 
Water vapor differential error map was formed in this stage. 
Forming of this map was done by subtracting zenith wet delays. 
Zenith wet delay was calculated for each pixel and according to 
the Eq.2 as below (Bevis et. al., 1996): 
ZWD wv =U~ x PWV 
(2) 
Where ZWD is the zenith wet delay and could be derived 
from Eq. 3 (Hanssen, 2001). 
rr 1 =KrV,^v 
(3) 
Where T m is the mean temperature of the column containing 
the water vapor and K coefficients are used as the (Smith and 
Weintraub, 1953) 
k x = 11.6 KhPa~ x 
k' 2 = 23. IKhPa-' 
k 3 =3.75x10 5 K 2 hPa~' 
After that the remained interpolated cloudy pixels were taken 
into account. Additional error was estimated, using Table 1 
(Hanssen, 2001): 
Cloud Type 
LWC [g/m 3 l 
SRD [mm/Km] 
Stratiforms 
0.05-0.25 
0.1-0.4 
Small cumulus 
0.5 
0.7 
Ice Clouds 
<0.1 
<0.1 
Tablel: Liquid Water content and Slant range delay of clouds