MSMR operational products are provided with a statistical
technique using radiative transfer simulations (Gohil et al., 2000).
Difference in MSMR operational GPDs with TMI or SSM/I may
arise due to difference in their operating frequencies, noise figures
and algorithms. The basic features of retrieval algorithm for TMI
and MSMR are shown in table 1, which shows the difference in
basic approach as well as the channels used. Table 2 shows the
theoretical accuracy of MSMR operational retrieval algorithms.
Table 1: Salient features of retrieval algorithms
Features MSMR Algorithm Wentz Algorithm
1. Type Statistical technique Minimization
using Radiative approach between
Transfer Simulations | measured and
(Gohil et al., 2000) simulated brightness
temperatures (Wentz,
2. Channels 1997).
for-
IWV | 18,21 (H & V) 22 (V), 37(V & H)
OWS | 6.1018, 21 (H & V) 22 (V), 37(V & H)
10,18, 21(V &FD ] ^ -———
CLW | 18,21 (V & H)
SST 1.6,10,13, 21(VR IH) | ——
3. Other Climate SST and Average SST and
Inputs in incidence angle incidence angle
Retrieval
Algorithms
Table-2: Theoretical Retrieval Accuracy of MSMR GPDs (Gohil
et al., 2000)
Parameter Tropics Midlat. Polar
SST (K) (grid: 150 km) 1.52 1.92 1.90
OWS (ms™) (grid: 150 km) 1.63 1.59 1.51
OWS (ms™) (grid: 75 km) 2.10 2.00 1.91
IWV (g cm™)' (grid: any) 0.20 0.18 0.15
CLW (mg cm” ) (grid: any) 13.0 11.0 9.0
Table 3: Comparison of MSMR GPDs with insitu observations
(Sharma et al., 2002)
Para- | No. of Rmsd Bias Rmsd after
meter | Points removal of
bias
SST 153 1.49 K 0.98 K 1.13 K
OWS | 162 1.62 ms” 1.62 ms” 1.80 ms”
IWV 16 053gcm^ | 039gcm? | 032g cm?
Table 4: Intercomparison of colocated MSMR and TMI GPDs
within 1 hr. of temporal difference (Varma et al., 2002a)
Para- | Grid No.of | R bias S.D. rms
meter | (km) pts. of diff. | diff.
50 27543 | 0.96 | -0.23 0.40 0.46
IWS 1:75 26626 | 0.96 | -0.22 0.40 0.46
150 10884 | 0.96 | -0.21 0.37 0.43
OWS | 75 24953 | 0.66 | -1.81 2.31 2.93
150 10254 | 0.73 | -1.78 2.01 2.68
50 25923 | 0.52 | 4.42 12.91 13.64
CLW | 75 25077 | 0.48 | 4.57 13.82 14.56
150 10219 | 0.44 | 5.39 10.60 11.89
IAPRS & SIS, Vol.34, Part 7, "Resource and Environmental Monitoring", Hyderabad, India,2002
Sharma et al. (2002) and Varma et al. (2002a) presented the
validation and intercomparison of operationally available
geophysical parameters with insitu and other satellites. They
found a very close agreement of IWV from MSMR with insitu
and other satellites (TMI & SSM/I). The OWS and SST from
MSMR compare reasonably well. The comparison of CLW from
MSMR was not found in a reasonable agreement with other
satellites. Comparison of MSMR GPs with insitu is shown in
Table 3 and that with TMI is shown in Table 4.
Furthermore, Varma et al. (2002b and 2002c) and Gairola et al.
(2000) presented sensitivity of MSMR channels for rainfall, and
presented an algorithm for the retrieval of rainfall. Pokhrel et al.
(2002) used the rainfall from MSMR to study variability of S-W
monsoon over Indian oceanic region.
In general, there are many approaches for the retrieval of GPs
from satellite measurements. One of them is empirical approach
in which a statistical relationship is established between
concurrent satellite measurements and insitu observations. This
approach is not general in nature and is specific to a sensor,
satellite, region and period. However, it takes care of all sensor
related errors. It cannot be applied in the initial phase of satellite
operations. On the other hand, simulation techniques are more
robust and are independent of these aspects except that they need
fine-tuning to account for satellite, sensor and other errors. In
either of these retrieval techniques, statistical inversion is an
essential component where a few approaches, like statistical
regression technique (Gohil et al., 2000), neural network (Gairola
et al, 2002), principal component analysis and iterative
minimization could be used
3. DATA AND ANALYSIS
In the present study, near concurrent observations of MSMR and
TMI of July 1999 are used and statistical relationships are
established between MSMR brightness temperatures and TMI
derived GPDs. As the north-south coverage of TMI is restricted to
X40? latitudes, the algorithms thus obtained will best represent
this range of latitudes. Due to high variability of CLW over space
and time, to establish empirical relation for CLW, we have
selected colocated observations from two sensors (MSMR &
TMI) in very close proximity of within 10 km and 5 minutes.
However, the other parameters which show less variability, in
order to allow larger number of colocated valid points for more
stable statistical relationship, while maintaining the spatial
difference of 5 km, the maximum temporal difference was relaxed
to 10 minutes. As presence of CLW and rain effects the retrieval
of IWV, OWS and SST, for retrieval of these GPs, we have
selected colocated observations with low amount of CLW in the
atmosphere. For, development of retrieval algorithm for IWV and
OWS, the CLW values up to 15 mg cm” are allowed, whereas for
retrieval of SST, CLW values up to 10 mg cm” are allowed. Out
of various forms selected for multiple regression between
colocated MSMR brightness temperatures and TMI GPDs,
following general form is selected and used for retrieval of all
GPs for their best statistical results.
3
3 3
Bra Ya: infas0 - ny). b, in (20 - ni). Xue infaso > (v E nu)
1 i + = 1 I 1
=
i=1 i=1 1
— (1)
Whe
10G
Equa
and .
table
Tabl
Coef
Mom ls | im ib a Sq A
The «
18 ar
requi
10 G
coeff
are g
Tabl
Parat