LOGICAL

RO DE
EO

PA
ET
IN D I A M

RT M
EN T
GOVERNMENT OF INDIA
NA

E
INDIA METEOROLOGICAL DEPARTMENT

TR
O
TI

NA EN
L C L A TE C
IM

Met. Monograph Climatology No. 22/2005

DEVELOPMENT OF A HIGH RESOLUTION

DAILY GRIDDED RAINFALL DATA
FOR THE INDIAN REGION

M. Rajeevan, Jyoti Bhate, J.D. Kale & B. Lal

National Climate Centre

India Meteorological Department
Pune 411 005, India.

2005
 Summary

To fulfill the demand from the research community, a high resolution (10 X 10 Lat/Long)
gridded daily rainfall data set for the Indian region was developed. For the analysis, daily rainfall
data of 6329 stations were considered. However, there were only 1803 stations with a
minimum 90% data availability during the analysis period (1951-2003). We have used only
those 1803 stations for interpolation in order to minimize the risk of generating temporal
inhomogeneities in the data set due to varying station densities. The geographical area, 6.50
N to 37.50 N, 66.50E to 101.50 E was considered for interpolating the station rainfall data. For
the analysis, we have followed the interpolation method proposed by Shepard (1968). This
method is based on the weights calculated from the distance between the station and the grid
point and also the directional effects. Standard quality controls were performed before carrying
out the interpolation analysis.

Quality of the present gridded rainfall analysis was compared with similar global gridded
rainfall data sets. Comparison revealed that the present gridded rainfall analysis is better in
more realistic representation of spatial rainfall distribution. The global data sets underestimated
the heavy rainfall along the west coast and over Northeast India. However, the inter-annual
variability of southwest monsoon seasonal (June-September) rainfall was found to be similar in
all the data sets.

Using this gridded rainfall data set, statistical analyses of daily rainfall were carried out.
One such analysis was the identification of break periods during the southwest monsoon
season. Break periods were identified as the periods in which the standardized daily rainfall
anomaly averaged over the Central India (210- 270 N, 720 – 850 E) was less than -1.0. The
break periods thus identified for the period 1951-2003 were found to be comparable with the
periods identified by earlier studies. Further, no evidence was found for any significant trends in
number of break or active days during the period 1951-2003.

The present IMD gridded daily rainfall data set is available to the research community
for non-commercial purposes. The data set can be extensively used for several studies on
numerical model validation, intra-seasonal variability and monsoon predictability studies.

1
1. Introduction

Information on spatial and temporal variations of rainfall is very important in

understanding the hydrological balance on a global/regional scale. The distribution of
precipitation is also important for water management for agriculture, power generation, and
drought monitoring. In India, rainfall received during the south-west monsoon season (June to
September) is very crucial for its economy. Real time monitoring of rainfall distribution on daily
basis is required to evaluate the progress and status of monsoon and to initiate necessary
action to control drought/flood situations.

Many modeling groups around the world are involved in simulating the mean monsoon
rainfall and its variability using general circulation models (GCM). These GCM results are to be
verified using observed rainfall data to understand the drawbacks in the model and to improve
its performance. High resolution observed rainfall data are also required to validate
regional/mesoscale models. Daily rainfall analyses are also required to examine and model the
intra-seasonal oscillations like Madden-Julian Oscillation (MJO) over the Indian region and
regional studies on the hydrological cycle and climate variability.

In the recent years, there has been a considerable interest among different research
groups in developing high resolution gridded rainfall data sets. (Huffman et al. 1997, Xie and
Arkin 1997, Dai et a l. 1997, Rudolf et al. 1998, Gruber et al.2000, New et al. 1999, Chen et al.
2002, Adler et al. 2003, Mitra et al, 2003, Beck et al. 2005). Better understanding of the global
and regional hydrologic cycles is at the heart of the World Climate Research Program (WCRP)
and Global Energy and Water Experiment (GEWEX) component. The Global Precipitation
Climatology Project (GPCP) is the WCRP GEWEX project devoted to producing community
analyses of global precipitation. The GPCP is an international project with input data sets and
techniques contributed to the monthly analysis by a number of investigators. Version 1 of the
GPCP monthly data set is described by Huffman et al. (1997) and the Version-2 data set is
described by Adler et al. (2003).

In this report, we discuss the development of a high resolution (10 X 10

Latitude/Longitude) daily gridded rainfall data set for the Indian region for 53 years (1951-2003).
The details of data used, quality control adopted and the methodologies of interpolation are
discussed in this report. The gridded data thus developed have been compared with other
similar global data sets. A few statistical analyses of gridded data also have been made to
demonstrate the use of this data set for monsoon research.

2
2. Rainfall Data and Quality Control

After the major drought of 1877 and the accompanying famine, India Meteorological
Department established a large network of rain gauge stations, which provided a valuable
source of data to analyze the space-time structure of the monsoon rainfall and its variability.
Over India, monsoon variability has profound influence on agricultural sector, which in turn
affects the Indian economy. Using this large data source of rainfall, numerous research papers
have been published addressing the Indian monsoon variability, its teleconnection and
prediction.

Henry Blanford, the first head of the India Meteorological Department focussed on the
task of establishing a uniform system of rainfall monitoring and the careful scrutiny of the rainfall
records, received from the rain gauge network. With the introduction of telegraph system, daily
rainfall and also other meteorological observations began to be collected and analysed on daily
basis. India Meteorological Department over the years has maintained very high standards in
monitoring the rainfall and other meteorological parameters over India with great care and
accuracy (Sikka 2003).

A brief historical account and description of the rainfall data collection by the IMD are
given by Walker (1910) and Parthasarathy and Mooley (1978). Hartman and Michelsen (1989)
used the IMD daily rainfall data of 1901-1970, to convert into a gridded data set by grouping the
station data into 10 Lat/Long grid boxes. Using this gridded rainfall data set, Hartman and
Michelsen (1989) and Krishnamurthy and Shukla (2000) studied the intra-seasonal and
interannual variability of rainfall over India.

For the present analysis, we have used the daily rainfall data archived at the National
Data Centre, India Meteorological Department (IMD), Pune. IMD operates about 537
observatories, which measure and report rainfall occurred in the past 24 hours ending 0830
hours Indian Standard Time (0300 UTC). In addition, most of the state governments also
maintain rain-gauges, for real time rainfall monitoring. IMD digitizes, quality controls and
archives these data also. Before archiving the data, IMD makes multi-stage quality control of
observed values. The major source of error is the systematic measuring error which results
from evaporation out of the gauge and aerodynamic effects, when droplets are drifted by the
wind across the gauge funnel. For this analysis, this error was not corrected in the rainfall
records.

3
 We have considered rainfall data for the period 1951-2003 for the present analysis.
Standard quality control is performed before carrying out the analysis. First, station information
(especially location) was verified, wherever the details are available. The precipitation data
themselves are checked for coding or typing errors. Many such errors were identified, which
were corrected by referring to the original manuscripts.

For the period, 1951-2003, IMD has the rainfall records of 6329 stations, with varying
periods. Out of these 6329 stations, 537 stations are the IMD observatory stations, 522 stations
are under the Hydrometeorology programme and 70 are Agromet stations. Remaining stations
are rainfall-reporting stations maintained by state governments. However, only 1803 stations
had a minimum 90% data availability during the analysis period (1951-2003). We have used
only those 1803 stations for which a minimum 90% data are available for the analysis in order
to minimize the risk of generating temporal inhomogeneities in the gridded data due to varying
station densities. The network of stations (1803 stations) considered for this study is shown in
Fig.1. The density of stations is not uniform throughout the country. Density is the highest over
south Peninsula and is very poor over northern plains of India (Uttar Pradesh) and eastern
parts of central India. Fig.2. shows the day to day variation of number of rainfall stations, which
were available for the analysis. On an average, about 1600 station data were available for the
analysis. However, during the recent years, number of stations available for the analysis
dropped significantly. This is due to the delay in digitizing and archiving the manuscripts which
are received at IMD in a delayed mode. Fig.3. shows the mean number of stations (averaged
for the period 1951-2003) available for the analysis in a 10 X 10 square grid. Number of stations
per grid (station density) is very large over south Peninsula. Rain-gauge density is
comparatively less over east central India (Chattisgarh) and northern plains of India.

3. Interpolation Method

Numerical interpolation of irregularly distributed data to a regular N-dimensional array is

usually called “Objective analysis” (OA). Thiebaux and Pedder (1987) grouped the OA
techniques as follows: empirical interpolation, statistical interpolation and function fitting. In
empirical interpolation, array values are computed from a distance-weighted sum of the data.
The weighting function is usually predetermined. Some of the empirical interpolation techniques
are iterative: array values are generated by successive approximation on the basis of errors of
back-interpolation to the original data locations. In statistical interpolation, array values are also
computed from a weighted sum of input data, except that weights are based on statistics of
spatial covariance of the data. In the third technique, mathematical functions or surfaces (for

4
e.g., thin plate splines, Hutchinson, 1998) are fitted to the data. The array values are computed
from these functions.

Bussieres and Hogg (1989) studied the error of spatial interpolation using four different
objective methods. The techniques considered are the empirical techniques of Barnes,
Cressman and Shepard and a Gandian-based statistical technique. For application to the
specific project grid, the statistical optimal interpolation technique displayed the lowest root
mean square errors. This technique and Shepard OA, displayed zero bias and would be useful
for areal average computations. The GPCP used a variant of the spherical-coordinate
adaptation of Shepard’s method (Willmott et al. 1985) to interpolate the station data to regular
grid points. These regular points are then averaged to provide area mean, monthly total
precipitation on 2.5 grid cells. New et al. (1999) used the thin plate splines proposed by
Hutchinson (1998). Mitra et al (2003) used the successive correction method of Cressman
(Krishnamurti et al. 1983).

For the present analysis, we have used the interpolation scheme proposed by Shepard
(1968), which is explained in detail below:

It is assumed that a finite number of N of triplets (Xi, Yi, Zi) are given, where Xi and Yi
are the locational coordinates of the data point Di, and Zi is the corresponding data value. Data
point locations may not be coincident. An interpolation function Z= f(x,y) to assign a value to
any location P(x,y) in the plane is sought. This two dimensional interpolation function is to be
smooth (continuous and once differentiable), to pass through the specified points, and to meet
the user’s intuitive expectations about the phenomenon under investigation. A surface based on
a weighted average of the values at the data points, where the weighting was a function of the
distances, to those points, would satisfy the criteria.

Let Zi be the value at data point Di, and d(P,Di) be the Cartesian distance between P
and D, where the P is the reference point, d(P, Di) will be shortened to di. The interpolated
value at P using this first interpolation function is:

∑ d i− u Z i
f (P )= i =1
N if di ≠ 0 for all Di (u>0)
∑i =1
d i
−u

5
 f ( p ) = Zi if di =0 for some Di

Empirical tests showed that higher exponents (u>2) tend to make the surface relatively
flat near all data points, with very steep gradients over small intervals between data points.
Lower exponents produce a surface relatively flat, with short blips to attain the proper values at
data points. An exponent of u=2, not only gives seemingly satisfactory empirical results for
purposes of general surface mapping and description, but also presents the easiest calculation.
In Cartesian coordinates,

(di)-2 = 1/{(x-xi)2+(y-yi)2}

Since the above weighting function means that only nearby data points are significant in
computing any interpolated value, a considerable saving in computation could be effected by
eliminating calculations with distance data points. As their inclusion tended to make the surface
of higher order, extra points of inflection could also be eliminated by interpolating from nearby
points only. To select nearby points, either (1) an arbitrary distance criterion (all data points
within some radius r of the point P) or (2) an arbitrary number criterion (the nearest n data
points) could be employed. The former choice, though computationally easier, allowed the
possibilities that no data points, might be found within the radius r. A combination of the two
criteria combined their advantages. In order to maintain the interpolation work reasonably well,
a minimum of one data point was chosen. A maximum four was established to limit the
complexity and amount of computation required. Further, an initial search radius Dx is
established according to the overall density of data points. We have fixed the search radius Dx
as 2.0 degrees.

For a reference point P, new weighting factors Si= S(di) may be defined from the
following function:

di =0, : Station data used directly

0<di < Dx/3 : Weight based on 1/di
Di
Dx/3 < di < Dx : Weight based on 27( − 1)2 / 4Dx
Dx
di > D x : Station data not used

Where, Dx is the search radius.

In order to improve the interpolation, a direction factor, in addition to a distance factor

was needed in defining weightings. Intuitively this represented the “shadowing” of the influence

6
of a data point from P by a nearer one in the same direction. A directional weighting term for
each data point Di near P was defined by

ti =
∑ S j [1 − cos( D i PD j ]
∑ S j

The cosine of the angle DiPDj can be evaluated by the inner product as

{(x-xi) (x-xj) + (y-yi) (y-yj)} / di dj

Since for all angles θ, -1 ≤ cos(θ) ≤ 1, it follows that 0≤ ti ≤2. If other data points Dj are in
roughly the same direction from P as Di, then the (1-cos) factors are near zero and ti is near
zero. If, on the other hand, the other data points are roughly opposite P from DI, then the (1-
cos) factors and hence ti are near 2. Counting direction, a new weighting function is defined:

Wi = (Si)2 X (1 + ti).

The latest version of the interpolation function is

f(P ) =
∑W Z i i
if di ≠ 0 for all Di
∑W i

f ( p ) = Zi if di = 0, for some Di

The starting point of the rectangular grid is 6.50 N and 66.5 0 E. From this point, there
are 35 points towards east and 32 points towards north. A separate binary file (35 X 32 grids)
was created for each year with different name. For the leap year, the binary file contained the
data for 366 days. The size of each data file (per year) is 1.6 Mb.

4. Comparison of IMD data set with other data sets

After completing the rainfall analysis for the period 1951-2003, we have compared the
IMD gridded data set with two global gridded rainfall data sets, the GPCP and VASClimo data
sets. The GPCP is an international project under the WCRP GEWEX programme devoted to
producing community analyses of global precipitation (Huffman et al. 1997, Rudolf et al. 1994,
7
Adler et al. 2003). These data are available at 1o latitude X 10 longitude and 2.50 latitude X 2.50
longitude resolutions. We have considered the 10 latitude X 10 longitude monthly GPCP data for
the period 1986-2003. German Weather Service and Johann Wolfgang Goethe-University,
Frankfurt jointly carried out a climate research project, named Variability Analysis of Surface
Climate Observations (VASClimo), which was started in October 2001. Main objective of this
project is the creation of a new 50-year precipitation climatology for the global land-areas
gridded at three different resolutions (0.50 lat/lon, 10 lat/lon, and 2.5o lat/lon) on the basis of
quality controlled station data.

More details of this new rainfall climatology are available in Beck et al. (2005). To
compare the IMD analysis with the VASClimo data set, we have considered the VASClimo data
of 1951-2000. Since both these global data are available on monthly time scale, we have added
IMD daily gridded rainfall data into monthly total before comparing with the global data sets.

Fig. 4.a shows the difference between the present analysis (IMD) and GPCP data set
and Fig. 4.b shows the correlation coefficient between the IMD analysis and GPCP analysis.
Fig.5.a shows the difference between the present analysis (IMD) and VASClimo data set and
Fig 5.b shows the correlation coefficient between the IMD analysis and VASClimo data set.

The differences between IMD and GPCP data are positive along the west coast,
implying that GPCP data set underestimates the heavy rainfall amounts along the west coast.
Similarly over eastern parts of central India, GPCP data under estimates the heavy rainfall
amounts. Over most parts of central India, the differences are of the order of 50mm only. The
correlations between GPCP and IMD rainfall data exceed 0.6 over central and NW India.

Over the most parts of India, the differences between the IMD data and VASClimo data
are of the order of 50mm only. However, along the west coast of India, IMD rainfall values are
more than the VASClimo values. The correlations between VASClimo and IMD rainfall data are
very large (exceeding even 0.6) over central and NW parts of India. Over Gujarat and central
Peninsula, correlations exceed even 0.8.

We have further compared the inter-annual variations of rainfall among the data sets.
For this purpose, area weighted rainfall for the southwest monsoon (June-September) season
was calculated with all the three data sets. However, we have excluded NE parts of India for
calculating the area weighted rainfall. The seasonal mean and standard deviation of rainfall of
three data sets are given below:

8
 Rainfall Product Mean Rainfall Coefficient of variation
IMD 836.3 mm 11.6 %
(1951-2003)
GPCP 747.8 mm 12.1%
(1986-2003)
VASClimo 844.0 mm 12.1%
(1951-2000)

GPCP data show smaller mean rainfall compared to IMD and VASClimo data sets. But
all the three data set shows similar coefficient of variation, i.e.12%. Fig. 6 shows the interannual
variation of southwest monsoon seasonal (June-September) rainfall calculated from IMD
analysis as well as GPCP data set. The similar plot for the VASClimo data set is shown in Fig.
7. There is an excellent similarity in the inter-annual rainfall variation among the different data
sets with all major drought and excess years were well captured by all the three data sets. The
correlation coefficient between the IMD and GPCP data sets is 0.94 (for the period 1986-2003)
and the same between the IMD and VASClimo data sets is 0.95 (for the period 1951-2000).

5. Analysis of Daily rainfall Data

The gridded daily rainfall data, presented in this report will be useful for many
applications. Some of the possible applications are validation of general circulation and
numerical weather prediction models and studies on intra-seasonal variability like active and
break cycles. In the past, similar gridded data sets (1901-1970) were used to examine the intra-
seasonal variability (Hartmann and Michelsen 1989, Krishnamurthy and Shukla, 2003) of the
Indian summer monsoon.

In this section, some results of a similar analysis are presented. Daily climatological
mean (mm/day) of rainfall for selected days (5 June, 15 June, 1 July, 31 July, 31 August and 20
September) are shown in Fig. 8. The climatological mean monsoon onset over Kerala is 1 June
with a standard deviation of 7 days. By 5 June, monsoon covers southwestern and
northeastern parts of India. Along the west coast and over NE India, rainfall exceeds 20
mm/day. By 15 June, monsoon advances further northwards along the west coast of India and
by 1 July, monsoon covers a major portion of the country except northwest parts of India.
Heavy rainfall (exceeding 20mm/day) is reported along the west coast and northeast India.
Over central parts of India, rainfall of the order of 10 mm/day is observed. On 31 August, good
rainfall amounts (of the order of 8-10 mm/day) are observed throughout the country. By last

9
week of September, monsoon is in the withdrawal stage. Rainfall is mostly confined over east
central India and south peninsula.

The daily climatological maps of rainfall for the entire year were also examined. As in
Krishnamurthy and Shukla (2000), we have considered four representative regions of India
(Southwest, Southeast, Central and Northwest) as shown in Fig.9. The daily mean
climatological rainfall charts in these four regions are shown in Fig.10. Rainfall amounts are the
highest in the southwest region, wherein July rainfall is of the order of 20 mm/day. Daily rainfall
variations averaged over all India is similar to the rainfall variations over the central India. Over
the country as a whole, rainfall peaks in July, which is of the order of 8 mm/ day. Rainfall
amounts are the lowest over NW India, where the southwest monsoon arrives late and
withdraws early. Over Southeast India, rainfall extends beyond September due to the onset of
northeast monsoon rainfall.

Fig.11 shows the spatial variation of southwest monsoon seasonal (June-September)

mean rainfall (calculated for the period 1951-2003) and Fig.12. shows the standard deviation of
southwest monsoon seasonal (June-September) rainfall during the same period. Rainfall is
maximum along the west coast and NE India. Another maximum rainfall zone is observed over
Chattisgarh and neighbourhood.

The present rainfall data set has been used to identify the active–break periods during
the southwest monsoon season. Long intense breaks are often associated with poor monsoon
seasons, and they have a large impact on rainfed agriculture (Gadgil and Joseph 2003).
Traditionally monsoon breaks have been identified at IMD on the basis of surface pressure and
wind patterns over the Indian region. The traditional breaks as followed by the India
Meteorological Department have been documented by Ramamurthy (1969) and De et al.
(1998). Recently, Gadgil and Joseph (2003) have examined the active and break periods
(1901-1989) using only rainfall data over the monsoon zone area covering the central parts of
India.

In the present analysis, the active and break periods during the southwest monsoon
season have been identified in the following way. The area averaged daily rainfall time series
for each year from 1951-2003 has been prepared by simply taking arithmetic mean of all rainfall
at all grid points over the central India (210 - 27oN, 720-850 E). For each calendar day, the
climatological mean and standard deviation of rainfall were calculated using the data of 1951-

10
2003. Then for each year, the area averaged daily rainfall time series has been converted to
standardized rainfall anomaly time series, by subtracting the daily rainfall time series from the
climatological mean and then dividing by daily standard deviation. The standardized rainfall
anomaly time series for the year 1988 (an excess monsoon year) and 2002 (a deficient
monsoon year) are shown in Fig. 13. The break period has been identified as the period during
which the standardized rainfall anomaly is less than –1.0, provided it is maintained
consecutively for 3 days or more. For 2002, the break periods have been identified as 6 July to
17 July and 23 July to 31 July.

We have examined the active and break periods for other years also based on the
above criteria. We have listed the break days during July and August months only, provided it
lasts consecutively for 3 days or more. The break and active periods during the period 1951-
2003 are given in Table-1. In the same table, break days as defined by Ramamurthy (1969), De
et al. (1998) and Gadgil and Joseph (2003) are also shown for comparison. The periods
identified in this study are comparable with others, especially with Gadgil and Joseph (2003).

Recently, Joseph and Simon (2005) reported that duration of break (weak) monsoon
spells in a monsoon season has increased by 30% during the period 1950-2002. Number of
days with daily average rainfall less than 8 mm/day (break or weak monsoon spells) increased
and number of days with daily average rainfall more than 12 mm/day (active monsoon spells)
decreased during the same period. These are alarm findings for a country whose food
production and economy depend heavily on monsoon rainfall.

To confirm the findings of Joseph and Simon (2005) and to further explore the issue, we
have made an analysis with 53 years of IMD daily gridded rainfall data. Using the standardized
daily rainfall anomaly averaged over the central India (210 – 270 N, 720 to 850 E), number break
and active days during the period June to September were calculated for each year for the
period 1951-2003. The time series of number of break and active days for the period 1951-
2003 are shown in Fig. 14 and 15 respectively. The time series of break days has a very high
negative correlation (-0.86) with southwest monsoon seasonal (June to September) rainfall.
With the time series of active days, the corresponding correlation is +0.62. In the study of
Joseph and Simon (2005), the number of break and active days were identified using a different
criterion as mentioned in the above paragraph for the period 1950-2002. Their study revealed
weaker correlations of -0.58 and 0.54 respectively for number of break and active days. Thus
the time series of break and active days prepared in this study is a more representative
measure of monsoon activity during the season. However, Figs. 14 and 15 do not show any

11
significant trend either in number of break days or number of active days, which is contrary to
the results of Joseph and Simon (2005).

6. Conclusions

In this report, we discussed the results of the development of a high resolution (10 X 10
Lat/Long) daily gridded rainfall data. We have considered the period 1951-2003 for the
analysis. We have considered in all 6329 stations. However, only 1803 stations had a minimum
90% data availability during the analysis period (1951-2003). We have used only those 1803
stations for which a minimum 90% data are available for the analysis in order to minimize the
risk of generating temporal inhomogeneities in the gridded data due to varying station densities.
We have considered the Shepard (1968) method with directional effects for the interpolation to
10 X 10 Latitude / Longitude regular grids. Before the interpolation, we have carried out certain
quality controls of the data. The gridded rainfall data set, thus developed, was compared with
other similar global gridded rainfall data sets. The present IMD analysis shows more accurate
representation of rainfall over the Indian region, especially along the west-coast and central
India. The correlation coefficients between the IMD rainfall time series and other global data
sets are more than 0.90. The present rainfall data set is used to identify the active and break
periods during the southwest monsoon season using an objective criterion. The active and
break periods thus identified were found comparable to the periods identified by earlier studies.
Analysis of break days does not show any significant trends.

It is believed that the present IMD gridded rainfall data set will be extensively used for
many applications like validation of climate and numerical weather prediction models and also
for studies on intra-seasonal variability and monsoon predictability studies. We shall be further
updating the present rainfall analysis from 1901 onwards, so that more than 100 years of
gridded daily rainfall data are available for the research community.

Acknowledgements

We have received encouragement and guidance from many scientists, during the
process of development of this data set. We are grateful to all of them. The completion of this
important project was a result of the encouragement and valuable guidance of Shri.S.R.Kalsi,
ADGM (S), IMD and Dr N. Jayanthi, DDGM(WF) & LACD ADGM (R), IMD Pune. We also had
fruitful discussions with Dr.J. Shukla , Dr. V. Krishnamurthy, Dr. B.N. Goswami, Dr. K. Rupa
Kumar, Dr. Ravi Nanjundiah, we thank them for their guidance. We also would like to thank
Shri._G.S.Prakasa Rao, Director, NDC, IMD Pune for his encouragement and greatly
supporting us during this important project.

12
References

Adler, R.F., Huffman, G.J., and Coauthors, 2003, The version-2 Global Precipitation
Climatology Project (GPCP) Monthly precipitation analysis (1979-present),
J.Hydrometeorology, 4, 1147-1167.
Beck,C., Grieser, J., and Rudolf, B., 2005, A New monthly precipitation climatology for the
Global Land Areas for the Period 1951-2000, Climate Status Report 2004, German
Weather Service, Offenbach, Germany, 10 pp.
Bussieres, N and W.Hogg, 1989, The Objective analysis of daily rainfall by distance weighting
schemes on a mesoscle grid, Atmosphere-Ocean, 3, 521-541.
Chen, M., P.Xie, J.E.Janowiak, and P.A.Arkin, 2002, Global Land Precipitation: A 50 yr monthly
Analysis Based on Gauge Observations, J.HydroMeteorology, 3, 249-266.
Dai, A., Fung, I., and Genio, D.G., 1997, Surface observed global land precipitation variations
during 1900-1998, J.Climate, 10, 2943-2962.
De, U.S., Lele,R.R., and Natu, J.C., 1998, Breaks in southwest monsoon, India Meteorological
Department, Report no 1998/3.
Gadgil, S., and Joseph, P.V., 2003, On breaks of the Indian monsoon, Proc.Indian.Acad.Sci.
(Earth and Planet Sci), 112, 4, 529-558.
Gruber, A., X Su, M.Kanamitsu, and J.Schemm, 2000, The comparison of two merged rain
gauge –satellite precipitation data sets, Bull.Amer.Met.Soc., 81, 2631-2644.
Hartman, D.L., and M.L.Michelsen, 1989, Intraseasonal periodicities in Indian rainfall,
J.Atmos.Sci., 46, 2838-2862.
Huffman, G.J., and Coauthors, 1997, The Global Precipitation Climatology Project (GPCP)
combined precipitation data sets, Bull.Amer.Meteor.Soc., 78, 5-20.
Hutchinson, M.F,1998, Interpolation of rainfall data with thin plate smoothing splines: I two
dimensional smoothing of data with short range correlation. Journal of Geographic
Information and Decision Analysis 2(2): 152-167.
Joseph, P.V., and Anu Simon, 2005, Weakening trend of the southwest monsoon current
through peninsular India from 1950 to the present, Current Science, 89, 4, 687-694.
Krishnamurti, T.N., S.Cocke, R.Pasch, and S.Low-Nam, 1983, Precipitation estimates from rain
gauge and satellite observations, summer MONEX, FSU report 83-7, Department of
Meteorology, The Florida State University, 373 pp.
Krishnamurthy, V., and J.Shukla, 2000, Intra-seasonal and Inter-annual variability of rainfall
over India, J.Climate, 13, 4366-4377.

13
Mitra, A.K., M.Das Gupta, S.V.Singh and T.N.Krishnamurti, 2003, Daily rainfall for the Indian
monsoon region from merged satellite and raingauge values: Large scale analysis from
real time data, J.Hyrdrometeorology, 4, 769-781.
New, M, Hulme, M. and Jones, P.D., 1999, Representing Twentieth-Century Space-Time
Variability. Part I: Development of a 1961-1990 Mean Monthly Terrestrial Climatology,
J.Climate, 1999, 12, 829-856.
Parthasarathy, B., and Mooley,D.A., Some features of a long homogenous series of Indian
summer monsoon rainfall, Mon.Wea.Rev., 106, 771-781.
Ramamurthy, K., 1969, Monsoon of India: Some aspects of the “ Break” in the Indian southwest
monsoon during July and August, Forecasting Manual, 1-57, No. IV, 18.3, India
Met.Dept, Poona, India.
Rudolf, B., H.Hauschild, W.Ruth, and U.Schneider, 1994, Terrestrial precipitation analysis:
Operational method and required density of point measurements, Global Precipitation
and Climate Change, M.Desbois and E.Desalmand, Eds NATO ASI Series I, Vol.26,
Springer-Verlag, 173-186.
Shepard, D., 1968, A two-dimensional interpolation function for irregularly spaced data, Proc.
1968 ACM Nat.Conf, pp 517-524.
Sikka, D.R., 2003, Evaluation of monitoring and forecasting of summer monsoon over India and
a review of monsoon drought of 2002, Proc.Indian Natn.Sci.Acad. 69, 5, 479-504.
Theibaux, H.J., and Pedder, M.A., 1987, Spatial Objective analysis with Applications in
Atmospheric Science, Academic Press, London.
Walker, G.T., 1910, On the meteorological evidence for supposed changes of climate in India,
Indian. Meteor.Memo, 21, Part –I, 1-21.
Willmott, C.J., C.M.Rowe, and W.D.Philpot, 1985, Small-scale climate maps: A sensitivity
analysis of some common assumptions associated with grid-point interpolation and
contouring, Amer.Cartographer, 12, 5-16.
Xie, P., and P.A.Arkin, 1997, Global Precipitation: A 17-year monthly analysis based on gauge
observations, satellite measurements and numerical models outputs,
Bull.Amer.Meteor.Soc., 78, 2539-2558.

14
 Table-1
Break days identified in the present analysis and previous studies
YEAR BREAK DAYS – JULY & AUGUST (1951-1989)
Ramamurthy (1969)
Gadgil and Joseph up to 1967
Present analysis
(2003) De et al. (1998) from
1968 to 1989
1-3 J, 11-13 J, 15-17 J, 9-14 J, 21-24 J,
1951 14-15J, 24-30A
24-29 A 25-30 A

1952 1-3 J, 10-13J, 27-30A 9-12J 9-15J

1953 - 24-26J -

1954 22-29A 18-29J, 21-25A 22-29A

1955 24-25J 22-29J -

1956 23-30A 23-26A 23-30A

1957 28-29J 27-31J, 5-7A -

1958 - 10-14A -

1959 - 16-18A -

1960 20-24J, 30-31A 16-21J 20-23J

1961 - - -

1962 27-28J, 1-2A, 7-8A, 25-26A 18-22A 27-29A

1963 18-19J, 22-23J 10-13J, 17-21J 16-18J, 21-23J

1964 - 14-18J, 28J-3A 1-5 A

1965 7-11J, 4-14A 6-8J, 4-15A 6-14J, 3-14A, 17-19 A

1966 2-12J, 22-31A 2-11J, 23-27A 2-13J, 24-31A

1967 6-15J 7-10J 10-14J

1968 25-31A 25-29A 25-31A

1969 27-31A 17-20A, 25-27A 29-31A

1970 14-19J, 23-26J 12-25J 14-19J, 23-26J

1971 8-10J, 5-6A, 18-19A 17-20A 5-7 A, 18-20 A

1972 19J-3A 17J-3A 11-14J, 19J-3A

1973 24-26J, 30J-1A 23J-1A 24- 26 J

1974 24-26A, 29-31A 30-31A 29-31 A

1975 - 24-28J -

1976 3-4J, 21-22A - 1-4 J

1977 15-19A 15-18A 14-21A

1978 - 16-21J -

1979 2-6J, 15-31A 17-23J, 15-31A 2-7J, 12-31A

15
1980 17-20J, 14-15A 17-20J -

1981 19-20A, 24-31A 26-30J, 23-27A 26-31 A

1982 1-8J - 1-9J , 16-19 J

1983 8-9J, 24-26A 22-25A 7-9J , 14-16 J

1984 - 20-24J 12-14 A

1985 2-3J, 23-25A 22-25A 11-13 A, 24-27A

1986 1-4J, 31J-2A, 22-31A 23-26A, 29-31A 3-5 J, 26-31A

16-17J, 23-24J, 31J-4A, 16-18 J, 30 J- 3A,

1987 11-13A 28J-1A 8-10A, 14-18A

1988 14-17A 5-8J, 13-15A 15-17A

1989 30-31J 10-12J, 29-31J 18-20J, 31J- 3A

BREAK DAYS

(1990-2003)
YEAR Present Analysis

1990 -

1991 2-8J

1992 4-10J

1993 19-23J, 8-14A

1994 -

1995 4-7 J

1996 3-5 J

1997 13-17A

1998 21-26J

1999 1-5 J , 18-20 A, 22-24 A

2000 22-24 J, 2-8A, 24-27A

2001 -

2002 5-16J, 22 J- 31 J

2003 -

16
 Fig.1. Locations of 1803 Rain gauge stations

NO. OF STATIONS PER DAY (1951-2003)

2000

1800

1600

1400
NO. OF STATIONS

1200

1000

800

600

400

200

0
1951
1953
1955
1957
1959
1961
1963
1965
1967
1969
1971
1973
1975
1977
1979
1981
1983
1985
1987
1989
1991
1993
1995
1997
1999
2001
2003

YEAR

Fig.2. Number of stations per day available for the analysis

17
Fig.3. Number of stations available for the analysis in a 10 X 10 square grid for which data are
available for at least 40 years. Total number of stations: 1803.

18
Fig.4.a. The difference (in mm) between the IMD gridded rainfall analysis and GPCP gridded
rainfall for the southwest monsoon season (June to September). Period: 1986-2003

Fig.4.b. Correlation Coefficient between the IMD rainfall data and GPCP rainfall data.
Period of analysis: June to September 1986-2003.

19
 Fig.5.a. The difference (in mm) between the IMD gridded rainfall data and VASClimo gridded
rainfall data for the southwest monsoon season. Period: 1951-2000.

Fig.5.b. Correlation Coefficient between the IMD rainfall data and VASClimo rainfall data during
the southwest monsoon season (June-September). Period of analysis: 1951-2000.

20
 COMPARISON OF IMD DATA FOR ALL INDIA ZONAL AVG. FOR JUNE-SEPT
WITH GPCC ZONAL AVG (1986-2003)
130

IMD DATA
120 GPCC

110

100
%OF AVG

90

80

70

60

50

40
1986

1987

1988

1989

1990

1991

1992

1993

1994

1995

1996

1997

1998

1999

2000

2001

2002

2003
YEAR

Fig. 6. Interannual variation of southwest monsoon season (June- September) rainfall from the
IMD gridded analysis and GPCP analysis. Period: 1986-2003

COMPARISON OF IMD DATA FOR ALL INDIA ZONAL AVG. FOR JUNE-SEPT
WITH VASCLIM ZONAL AVG (1951-2000)
140

IMD DATA
130 VASCLIM

120

110
% OF LPA

100

90

80

70

60
1951
1953
1955
1957
1959
1961
1963
1965
1967
1969
1971
1973
1975
1977
1979
1981
1983
1985
1987
1989
1991
1993
1995
1997
1999
2001

YEAR

Fig. 7. Interannual variation of southwest monsoon season (June- September) rainfall from the
IMD gridded analysis and VASClimo analysis. Period: 1951-2000

21
Fig.8. Daily climatological mean (mm/day) rainfall for selected days (5 June, 15 June, 1 July, 31
July, 31 August and 20 September).

22
 4

Fig. 9 Four representative regions of India (Southeast, Southwest, Central

and Northwest India)

DAILY RAINFALL : CLIMATOLOGY : 5 DAY RUNNING MEAN (1951-2003)

24
INDIA
BOX-1
21 BOX-2
BOX-3
BOX-4
SOUTHWEST
18 INDIA
RAINFALL (MM/DAY)

15

CENTRAL INDIA
12

ALL INDIA
9

SOUTHEAST
6 INDIA

3
NW INDIA

0
Jan Feb Mar Apr May Jun Jul Aug Aug Sep Oct Nov Dec
MONTH

Fig.10. Daily rainfall climatology (mm/day) for the four representative regions

23
Fig.11. Spatial pattern of southwest monsoon seasonal (June to September) mean rainfall,
1951-2003 (Unit: mm/day)

Fig.12. Spatial pattern of standard deviation pf southwest monsoon seasonal (June to

September) rainfall, 1951-2003 (Unit: mm/day)

24
 STANDARDISED ANOMALY FOR DAILY RAINFALL : 1988
3.0

2.0

1.0
STD. ANOMALY

0.0

-1.0

-2.0

-3.0 3-Aug

10-Aug

17-Aug

24-Aug

31-Aug
6-Jul

13-Jul

20-Jul

27-Jul
1-Jun

8-Jun

15-Jun

22-Jun

29-Jun

7-Sep

14-Sep

21-Sep

28-Sep
DATE

STANDARDISED ANOMALY FOR DAILY RAINFALL : 2002

4.0

3.0

2.0
STD. ANOMALY

1.0

0.0

-1.0

-2.0

-3.0
3-Aug

10-Aug

17-Aug

24-Aug

31-Aug
6-Jul

13-Jul

20-Jul

27-Jul
1-Jun

8-Jun

15-Jun

22-Jun

29-Jun

7-Sep

14-Sep

21-Sep

28-Sep

DATE

Fig.13. The standardized rainfall anomaly time series for the year a) 1988 and b) 2002
for the period 1 June to 30 September.

25
 NUMBER OF BREAK DAYS : JUNE - SEPTEMBER
40
y = 0.0782x + 13.19
C.C. With ISMR = - 0.86
R2 = 0.0194
35

30

25
DAYS

20

15

10

0
1951
1953
1955
1957
1959
1961
1963
1965
1967
1969
1971
1973
1975
1977
1979
1981
1983
1985
1987
1989
1991
1993
1995
1997
1999
2001
2003
YEAR

Fig.14. Time series of Break Days during the monsoon season (1951-2003)

NUMBER OF ACTIVE DAYS : JUNE - SEPTEMBER

40
y = 0.03x + 18.605
2 C.C. with ISMR = 0.62
R = 0.0043
35

30

25
DAYS

20

15

10

0
1951
1953
1955
1957
1959
1961
1963
1965
1967
1969
1971
1973
1975
1977
1979
1981
1983
1985
1987
1989
1991
1993
1995
1997
1999
2001
2003

YEAR

Fig.15 Time series of Active Days during the monsoon season (1951-2003)

26