'sented the
available
lites. They
with insitu
SST from
CLW from
with other
shown in
irola et al.
Unfall, and
khrel et al.
ity of S-W
ial of GPs
|] approach
| between
ions. This
a sensor,
all sensor
of satellite
are more
they need
errors. In
sion is an
statistical
k (Gairola
iterative
[SMR and
ships are
and TMI
stricted to
represent
)ver space
we have
MSMR &
| minutes.
ability, in
for more
le spatial
as relaxed
e retrieval
we have
JW in the
IWV and
hereas for
wed. Out
between
II GPDs,
val of all
0 (v = nu)
l 1
— (1)
IAPRS & SIS, Vol.34, Part 7, "Resource and Environmental Monitoring", Hyderabad, India,2002
Where, P is one of the GPs (CLW / IWV / OWS / SST). Tb, - Tb
10 GHz, Tb, = Tb 18 GHz, and Tb;= Tb 21 GHz. Suffix to Tb, V
& H, represents Vertical and Horizontal polarized Tb values.
Equation gives CLW in mg cm?, IWV in g cm, OWS in ms’,
and SST in K. The value of regression coefficients is given in
table 5.
Table 5: Value of regression coefficients for retrieval of GPs.
Coefficient CLW IWV OWS SST
ao -237.7 42.31 3345.0 3345.0
a; 0.0 4.28 270.0 270.0
a, -296.9 -1.71 -30.56 -30.56
az 119.4 -2.23 -2.08 -2.08
bi 0.0 -8.19 -569.4 -569.4
b; 242.3 14.73 42.63 42.63
bs -133.0 —7.39 24.20 24.20
€ 0.0 -12.89 -494.3 -494.3
Cy 494.7 9.00 96.85 96.85
C3 -396.2 -3.38 57.62 57.62
The equation 1 given above is involving terms with Tb from 10,
18 and 21 GHz channels However, the retrieval of CLW does not
require 10 GHz brightness temperature and hence terms involving
10 GHz Tbs are forced to zero by the corresponding regression
coefficients. The other statistical details of the multiple regression
are given in table 6.
Table 6: Statistical details of colocated MSMR and TMI data
Parameter No. Value of Parameter R Error
of Min. Max. of
Points Estimation
CLW 452 0 156.4 0.82 8.75
mg cm? mg cm? mg cm?
IWV 769 0.7 6.2 0.97 0.31
g cm? g cm? g cm?
OWS 769 0.54 18.4 0.78 1.86
ms! ms”! ms”!
SST 642 13.24 31.62 0.89 1.81
K K K
In the table 6, the second column represents the number of valid
data sets for establishing multiple regression. The third column
provides the minimum and maximum values of respective GP
(from TMI) in colocated data set. The fourth column gives the
value of multiple correlation between GP and different terms in
respective retrieval equation. The fifth, and the last column,
provides error of estimation as estimated from the colocated data
set. It can be seen that this data set represent large dynamic range
of parameters in the given latitudinal range (+ 40°).
MSMR provides global coverage in 2 days, so we used the
algorithms for GPs to generate 2-day averaged and monthly fields
and provided them with corresponding SSM/I and MSMR-
operational GP fields on same scales for comparison. In the
succeeding sections of this paper, we will refer the GPs derived
from the equation 1 as MSMR empirical GPs and those derived
using operational algorithm (Gohil et al, 2000) as MSMR
operational GPs.
4. RESULTS AND DISCUSSION
As can be seen from table 2 and 5 that accuracy of all the
parameters is well comparable. There is however marginal
deterioration of accuracy in case of retrieval of IWV and
significant improvement in case of retrieval of CLW using
empirical algorithm. It may be recalled that the CLW provided as
operational product was not found in reasonable agreement with
other satellite. So the improvement in the retrieval of CLW is a
significant achievement of the present study. It may further be
seen from table 3 and 4, that achieved accuracies of MSMR
operational IWV and CLW was worse than theoretical accuracy
when compared with TMI and/or insitu measurements; the
accuracy of operational OWS was better with insitu, but worse
with TMI; and the accuracy of operational SST was found better
with insitu. With all this in consideration, the accuracies provided
in table 5 are very encouraging.
MSMR - Cloud Liquid Wat
1-2 August 1
> (mg/cme2)
MSMR (Operational) - Cloud Liquid Woler (mq/cree2)
1-2 August 1999
5 19 15 20 25 30 3H 40 45 50
Fjg-1: Global distribution of averaged CLW from MSMR empirical
algorithm (upper panel), SSM/I (middle panel) and MSMR-
operational algorithm (lower panel) for 1-2 Aug. 1999.
Using present empirical algorithm, we have further plotted
average CLW, IWV, OWS and SST for 2-days of August 1999 (1
—2 August). Empirical algorithms are statistically tuned to the
actual measurements (by regression) and hence they often defy
the natural limits of the variability of the parameter. For that
reason, we have taken dynamic range of CLW from 0-500 mg cm
2 IWV from 0 — 8 g cm, OWS from 0 — 35 m cm”, and SST
from 14-32 K for our area of interest (global within +40°
latitudes). These limits of the GPs are based on limits of the
microwave retrieval of the respective parameters and from
climatology of the study area. The microwave measurements of
IWV, OWS and SST are affected by the presence of clouds and