ll is received all 
f spatial rainfall 
‚ which has high 
.nowing its type 
of satellite data, 
thermal infrared 
was applied for 
Godavari basin 
vents have been 
t correlation, the 
S to arrive at an 
[IMATION 
estimation for 
he satellite data 
a catchment or 
minutes of the 
not constrained 
advantages but 
vely expensive. 
data have been 
f rain is cloud. 
that type would 
jns. There are 
used to estimate 
Barret et al). 
ique is required 
veral years and 
s needed by the 
ita, a number of 
at varying time 
1al Infrared and 
At one end of 
ened to produce 
periods e.g. the 
oduce estimates 
r periods of one 
d was designed 
| a global scale. 
Combined MW and IR algorithms using SSM/I radiometric data 
were first oriented to monthly averages over wide areas ( Adler et 
al., 1993) or to global products as it is for the Global Precipitation 
Climatology Project (GPCP) (Huffman et al., 2001). At the other 
end of time- scale there has been number of techniques developed 
for very short period ‘near real-time’ estimation of rainfall at the 
best spatial resolution possible from geostationary satellites (4-8 
km data produced every 30 minutes). These techniques have been 
tried for flash flood warnings, etc. Classification of various 
techniques is according to number of channels, sampling and 
number of predictors. Sampling includes spatial and temporal 
dimensions. All the techniques are empirical in nature. 
Depending on whether a technique is based on geostationary 
satellite data or on polar orbiting satellite data, several important 
parameters governing the amount of rainfall results from clouds 
can be stated as follows: 
- Bright clouds in the visible imagery and clouds with 
cold tops in the IR imagery that are expanding in area 
coverage produce more rainfall than those that are not 
expanding. 
- Decaying clouds produce little or no rainfall. 
- Clouds with cold tops in the IR imagery produce more 
rainfall than those with warmer tops. 
- Clouds with cold tops that are becoming warmer 
produce little or no rainfall. 
- Merging of cumulonimbus clouds increases the rainfall 
rate of merging clouds. 
The principles of most of the current techniques for estimating 
rainfall from the temperature of the cloud tops, are associated 
with relating rainfall with cold clouds assuming that these are the 
tops of active storms. The rate of vertical or horizontal growth of 
clouds or their minimum temperature can give further information 
on the type of cloud and its likelihood of producing rain. 
Techniques incorporating such features are known as the ‘life 
history’ methods. 
2.2 Rainfall estimates from cold cloud statistics 
2.2.1 General: The University of Reading, UK, developed this 
method for using on operational basis. The basic methodology of 
the cold cloud statistics procedures is simple. A regular series of 
thermal infrared (TIR) images of an area is received, pixels with 
apparent temperatures lower than some predetermined threshold 
are classified as “cold cloud’, and their characteristics 
accumulated over some period. The resultant map is converted to 
a rainfall estimate. The procedures adopted as a statistical model, 
which is calibrated through comparisons between the cold cloud 
characteristics and sets of conventional raingauge data. To 
establish the utility of the method, it must subsequently be 
validated, by comparing estimates from some area or period 
distinct from that used for the calibration. 
2.2.2 Space and time considerations: Space and time intervals 
of satellite data are important factor to be considered besides 
spatial resolution. Studies have shown that there is little 
difference between the results obtained from using half-hourly or 
hourly satellite data (Milford and Dugdale, 1989). However, if a 
system is to be dedicated to rainfall estimation, there is no 
disadvantage in acquiring and processing half-hourly data, must 
be used for calibration and validation purposes because the spatial 
variability of rainfall is high and we wish to compare a raingauge 
IAPRS & SIS, Vol.34, Part 7, “Resource and Environmental Monitoring”, Hyderabad, India,2002 
(point) measurement with the rainfall estimate for the 
corresponding satellite pixel. 
2.2.3 Calibration and validation: The objective is to select a 
temperature threshold that will discriminate between clouds, 
which are producing rainfall, and inactive clouds. This threshold, 
which can be applied to convective cloud, is a function of the 
dynamic and physical structure of the storm, which may vary 
widely. Fortunately, over the tropics, these features are usually 
related to the region and season. Having chosen a threshold there 
is still no information on the rainfall distribution beneath an 
individual cloud so one cannot make estimates of the 
instantaneous rate of rainfall over a satellite pixel. Instead, 
estimates must be aggregated over a period of time and/or a large 
area to reduce the influence of the short period, small-scale 
variability of rainfall. The minimum periods over which single 
pixel rainfall estimates are useful are probably about 10 days. 
However, if spatial as well as time aggregation is included daily 
rainfall estimates can be made. This is the approach used for 
catchment rainfall estimation. Calibrations may be achieved either 
in terms of mm of rainfall over the catchment per hour of cold 
cloud duration or, if a rainfall/stream flow model is to be used and 
flow gauge measurements are available, it may be possible to use 
the average clod top temperature as a direct input to the model as 
a surrogate for rainfall. The regressions have to be made for the 
average daily cloud top temperature to the observed daily rainfall. 
These are very similar in terms of mm of rain per day and the 
average cold cloud temperature. The validation can be made by 
repeating the calibration procedure with an independent data set 
for subsequent years/months. Another validation technique is to 
monitor the performance of flow prediction models using satellite 
rainfall estimates as inputs and to compare them to the 
performance using the best available raingauge data. A major 
problem in using raingauge data to calibrate or validate satellite 
estimates of rainfall is the poor representative of area rainfall 
given by raingauge data. 
3.0 STUDY AREA 
Godavari basin has been chosen as study area. Godavari river 
rises near Nasik in Maharashtra at an elevation of 1067 m and 
flows for a length of about 1465 km before outfalling into the Bay 
of Bengal. The principal tributaries of the river are the Parvara, 
the Purna, the Manjra, the Penganga, the Wardha, the Wainganga, 
the Indravati and the Kolab. The catchment area of Godavari 
basin is about 3,12,800 sq.km. The catchment area of the basin is 
bounded on the west by the Western Ghats, on the east by the 
Eastern Ghats and on the north by the Satmala hills. These hill 
ranges play an important role in the distribution of the seasonal 
rainfall in the basin (G Nageswara Rao). The seasonal rainfall is 
very high over the hilly regions of the extreme west and in the 
north and east. Immediately after crossing the Western Ghats, the 
rainfall decreases rapidly and then starts increasing gradually 
towards the east. The north- eastern parts of the basin also 
receives heavy rainfall due to the passage of monsoon 
disturbances from the Bay of Bengal in a northwesterly direction 
across and to the north of the basin. It receives about 85% of its 
annual rainfall during the monsoon season. There are about 32 
raingauges in the basin for rainfall measurement.