IAPRS & SIS, Vol.34, Part 7, “Resource and Environmental Monitoring”. Hyderabad, India.2002 
  
  
The regression equation with derived coefficients for various 
channels is as follows. 
SIC =— 84.124 + 0.53148 * Tb (10V) + 0.424172 * Tb (10H) 
— 0.193778 * Tb (18V) — 0.043911 * Tb (18H) 
In the above, we have used MSMR Tbs from 10 and 18 GHz 
channels. While 10 GHz channel has a better transparency to 
the atmosphere, it has a poorer spatial resolution of 75 kms 
compared to 50 kms of the 18 GHz channel. All relevant 
MSMR and SSM/I data were therefore brought to 75 km 
resolution before attempting the regression analysis. The 
resulting rms regression error is 6.62%. 
Using the above regression equation, we have generated SIC 
images over the Southern Ocean for two representative 
months of Sept. 1999 and Jan. 2000. Fig. 4 shows a 
comparison between MSMR derived and SSM/I observed SIC 
images. This is done to examine the spatial characterization of 
SIC in different sectors of the Southern Ocean. It can be seen 
clearly that MSMR derived SIC images faithfully portray all 
the features observed in SSM/I image shown alongside for 
comparison. 
199$ 
NEMR 
SEPTEMBER, 
  
  
  
  
  
  
  
  
  
Fig. 4 — Comparison between monthly average SIC derived 
from MSMR Tb based algorithms and SSM/I for Sept. 1999 
and Jan. 2000. 
4.2 Development of SIC algorithm from MSMR 
Polarization and Spectral Ratios 
In order to reduce the effect of additive and multiplicative 
errors in MSMR Tbs, we have also followed a different 
approach of using the polarization (PR) and gradient (GR) 
ratios, defined below, following (Gloersen et al., 1992): 
PR =[Tbr(V)- Tbr (H) ]/[Tb¢(V) + Tl (H) ] 
  
412 
where f is the frequency (10 or 18 GHz), Tb is the 
brightness temperature, and V & H - Vertical and 
Horizontal polarizations. And 
GR = [ Tb(18(p)) - Tb (10(p)) ] / [ Tb (18(p)) + Tb (10(p) ] 
where p is the polarization - V or H. Use of PR and GR 
also eliminates the effect of differences between center 
frequency of the 18/19 GHz channels of MSMR and SSM/. 
Spatial averaging effect due to 25 km resolution of SSM/I 
VS. the 50 and 75 km resolutions of 18 and 10 GHz 
channels of MSMR would of course effect the derived SIC 
estimates somewhat. 
For estimating the PR and GR, we have once again made 
use of the 10 and 18 GHz channels of MSMR. The PR and 
GR estimates derived from MSMR data were regressed 
against the simultaneous and coincident SSM/I SIC values. 
This regression was carried out over the areas mentioned 
above (see Fig. 1). The regression equation below 
SIC = 101.6459— 14691.9 * PR (10 GHz) 
+ 14297.2 * PR (18 GHz) + 13958.7 * GR (H) 
— 142163 * GR (V) 
has an rms error of 7.57 95. 
  
  
  
  
JANUARY, 2000 
SWR 
  
  
Fig. 5 — Comparison between monthly average SIC derived 
from MSMR PR/GR based algorithms and SSM/I for Sept. 
1999 and Jan. 2000 
Based on the above PR & GR based algorithm, we have 
generated SIC images using MSMR data for the two 
seasons (Sept. 1999 and Jan. 2000). These are shown in Fig. 
5 alongwith the corresponding SIC images generated using 
SSMA data. These images depicting the SIC over the 
Southern Ocean surrounding the Antarctica from the two 
sensors viz. MSMR and SSM/I show very good feature-by- 
feature spatial correspondence in all the sectors.