PDF file - Caos

Preliminary Comparison of Daily Rainfall from
Satellites and Indian Gauge Data
Hasibur Rahman and Debasis Sengupta
CAOS Technical Report No. 2007AS1
Centre for Atmospheric and Oceanic Sciences
Indian Institute of Science
Bangalore-12
January 2007
Contents
1 Introduction
1
2 Data
2
2.1
2
TRMM 3B42-V5 and 3B42-V6 . . . . . . . . . . . . . . . . . . . . . .
2.2
GPCP
2.3
IMD
. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
4
. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
5
3 Results and Discussion
6
3.1
Spatial distribution of rainfall over land
. . . . . . . . . . . . . . . .
7
3.2
Time series over land
. . . . . . . . . . . . . . . . . . . . . . . . . .
10
3.3
Spatial distribution over the ocean . . . . . . . . . . . . . . . . . . .
11
3.4
Time series over the Ocean
. . . . . . . . . . . . . . . . . . . . . . .
16
3.5
Intraseasonal and interannual variability . . . . . . . . . . . . . . . .
17
4 Conclusions
22
i
List of Figures
1
Study regions
. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
6
2
Summer monsoon rain rate . . . . . . . . . . . . . . . . . . . . . . . .
7
3
All India seasonal rainfall
. . . . . . . . . . . . . . . . . . . . . . . .
8
4
Daily standard deviation of monsoon rain rate . . . . . . . . . . . . .
9
5
Rain rate for 15-18 August, 2002
6
Daily rain rate at east coast of India and Central India
7
Daily rain rate at east coast of India and Central India during 2002
.
13
8
Daily rain rate at east coast of India and Central India during 2003
.
14
9
Daily rain rate over Central India . . . . . . . . . . . . . . . . . . . .
17
10
Daily rain rate over East India . . . . . . . . . . . . . . . . . . . . . .
18
11
Daily rain rate over North India . . . . . . . . . . . . . . . . . . . . .
18
12
Seasonal mean rain rate (1998-2003) over Indian region from 3B42-V6
. . . . . . . . . . . . . . . . . . . .
. . . . . . . .
11
12
and GPCP . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
19
13
Daily rain rate over tropical Indian Ocean
20
14
Intraseasonal rain rate over Central India and north Bay of Bengal
. . . . . . . . . . . . . . .
during 2002 and 2003 . . . . . . . . . . . . . . . . . . . . . . . . . . .
15
Seasonal mean(JJAS) rain rate over Indian region during 2002 and
2003
16
21
. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
Interannual variation of seasonal rain over Central India
ii
. . . . . . .
23
24
List of Tables
1
Data sets
. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
2
All India seasonal (JJAS) total rainfall (cm), 1998-2003.
3
All India seasonal (JJAS) mean rain rate and daily standard deviation
1998-2003 (mm/day)
4
. . . . . . . . . . . . . . . . . . . . . . . . .
10
15
Daily average rain (mm/day) at two grid points during an active
. . . . . . . . . . . . . . . . . . . . . . . . .
16
RMS dierence (mm/day) between daily satellite and IMD rain rate
at two grid points, JJAS 1998-2003
7
8
Daily average rain (mm/day) at two grid points during an active
monsoon period in 2003
6
. . . . . . .
. . . . . . . . . . . . . . . . . . . . . . . . . . .
monsoon period in 2002
5
2
. . . . . . . . . . . . . . . . . . .
22
RMS dierence (mm/day) between daily satellite and IMD rain rate
averaged over the regions CI, EI and NI, JJAS 1998-2003 . . . . . . .
22
Citation
This Technical Report should be cited as : Rahman, H. and D. Sengupta, 2007,
Preliminary comparison of daily rainfall from satellites and Indian gauge data,
Tech. Rep. 2007AS1, 26pp, Indian Institute of Science, Bangalore, India.
On Web
This document is available on-line in PDF format via the World Wide Web from
http://caos.iisc.ernet.in/faculty/dsen.html.
Contact
Authors can be contacted at
[email protected] (H. Rahman) &
[email protected] (D. Sengupta).
iii
1
1
INTRODUCTION
1
Introduction
Rainfall data is crucial in applications such as water management for agriculture
and power, and drought and ood forecasting. Reliable observations of rainfall are
important for climate science because precipitation is a major component of the
earth's water and energy cycles. These cycles are inherently complex, with interaction between land, ocean, atmosphere and cryosphere. The tropical atmosphere gets
three-fourth of its heat energy from the release of latent heat associated with precipitation. Elevated latent heating has strong inuence on surface presuure, surface
winds in the tropics, evaporation and ocean circulation. Rainfall determines river
runo and modies sea surface salinity, upper ocean stratication and mixed layer
depth. Thus rainfall has very important direct and indirect inuence on the distribution of tropical sea surface temperature, atmospheric water vapour, boundary layer
moisture convergence, and convection. Finally, the distribution of water vapour and
clouds modies radiative uxes in the atmosphere, and turbulent exchange of heat
and water vapour with land and ocean.
One of the most challenging problems in climate science is observing and understanding spatial and temporal variations of tropical rainfall. Rainfall has variability on time scales ranging from minutes to hours (diurnal) through intraseasonal
(weeks), to seasonal, inter-annual, and longer. Further, rain is associated with cloud
systems that have complex structure, organised on dierent space scales. Even over
land, it is a dicult atmospheric variable to measure because of its large variability.
Traditionally it has been hard to obtain reliable precipitation information over
the oceans, where rain gauges are not available. In this context the Tropical Rainfall
Measuring Mission (TRMM) satellite has become an important resource, providing
data over the entire tropics since November 1997.
In combination with infrared
observations from geostationary satellites, useful daily rain products (the 3B42
Adler et al.,
datasets) have been made available by the TRMM Project (
2000).
Another daily satellite-derived rainfall dataset developed under the Global Precip-
◦
◦
itation Climatology Project (GPCP) is available in 1 x 1 gridded format (GPCP
1DD) from late 1996 to present. GPCP 1DD is a merged dataset based on gauge
measurements and satellite estimates of rainfall.
The satellite data used in the
rainfall estimations are Special Sensor Microwave Imager (SSM/I) and Meteosat
◦
◦
IR. India Meteorological Department (IMD) has recently developed a 1 x 1 gridded
daily rainfall dataset (
Rajeevan et al., 2005) based on rain gauge measurements from
1803 stations over Indian land for the period 1951-2003.
In this report we present rst results of a comparison of daily GPCP and 3B42
2
DATA
2
Table 1: Data sets
Spatial
Resolution
DataSet
T emporal
Resolution
Duration
TRMM 3B42 V5
◦
◦
1 x1
Daily (0-23Z)
GPCP
◦
◦
1 x1
Daily (22:30-22:30Z)
IMD
◦
◦
1 x1
Daily (03Z-03Z)
TRMM 3B42 V6
◦
◦
1 x1
Daily (0-23Z)
June-September 1998-2003
rainfall with IMD daily rainfall over Indian land. We examine the seasonal mean
summer monsoon (1 June-30 September) rainfall and its daily time variability, spatial patterns of seasonal mean and daily standard deviation, and briey look at
simple measures of intraseasonal and interannual variability over selected regions.
Comparison between the two satellite rainfall products is also presented for selected
oceanic regions, where there is no gauge data. The datasets used in this study are
shown in Table 1. A description of the data is given in Section 2. Results of the
satellite-gauge rainfall comparison are presented in Section 3, with a brief discussion.
2
2.1
Data
TRMM 3B42-V5 and 3B42-V6
TRMM is National Aeronautics and Space Administration's (NASA) rst mission
dedicated to observing and understanding tropical precipitation and its relation with
global climate.
Launched in November 1997, TRMM provides a unique platform
for measuring rainfall from space using a passive sensor TRMM Microwave Imager
(TMI), an active Precipitation Radar (PR) operating at 13.6 GHz, and a visible and
infrared scanner (VIRS) radiometer. TMI is a multi-channel/dual polarized (except
in 22 GHz) microwave radiometer (10, 18, 22, 37 and 85 GHz), which provides
data related to rainfall rates over the tropical oceans besides sea surface temperature (SST), sea surface wind speed (SSW), total water vapour (TWV) and cloud
liquid water content (CLW). TRMM algorithm 3B-42 provides adjusted 24-hour cu-
2
DATA
3
mulative estimates of rain using merged microwave and infrared (IR) precipitation
information (
Adler et al., 2000).
The TRMM adjusted Geostationary Observational
Environmental Satellite (GOES) precipitation index (GPI) (AGPI) is produced by
using cases of (nearly) coincident TRMM combined instrument (TCI) using the combined TMI and PR algorithm (
Haddad et al., 1997) and VIRS IR data to compute a
time and space varying IR-rain rate relationship that matches (i.e. is adjusted" to)
the TCI IR rain rate. This relation is used to calibrate IR estimates from geosynchronous satellite IR data to form the 3B42 product.
Global estimates are made
by adjusting the geosynchronous satellite Precipitation Index (GPI) to the TRMM
estimates. The monthly TRMM and merged estimate is produced by merging the
AGPI with information from rain gauges. The gauge analysis used in this proce-
Rudolf , 1993).
dure is from the GPCP (
et al. (1997).
The merger is computed following
Human
The 3B42 algorithm provides daily precipitation and root mean square
◦
◦
◦
(RMS) error estimates at 1 x 1 latitude/longitude grids in the TRMM domain 40 N
◦
to 40 S (
Human et al.,
◦
◦
2001) for V5 and at 0.25 x 0.25 latitude/longitude grids
◦
◦
over 50 N to 50 S for V6.
Some validation of TRMM 3B42-V5 data has been done using IMD rain gauge
data by
Narayanan et al. (2005) over Indian land.
Their main nding is that 3B42-
V5 does not pick up small (< 1mm) and very high (> 80 mm per day) daily average
rainfall.
Thus, the daily variance (day-to-day variations within the season) esti-
mated by 3B42-V5 is poor compared to the gauge data. The reasons may be related
to deciencies in the IR estimates. However at pentad (ve-day) time scale the correspondence between the two datasets improves, and intraseasonal and interannual
variations are reasonable. The correlation coecient over all of India on the monthly
scale is high (∼0.92) in comparison to 5-day (∼0.89) and daily (∼0.79) time scale.
Recently
Chokngamwong and Chiu
(2005) have validated 3B42 data using rain
gauge data from more than one hundred gauges over Thailand. Their results show
that 5-year (1998-2002) daily average rainfall for gauge, 3B42-V5 and 3B42-V6 are
4.73, 5.62 and 4.58 mm/day respectively. The bias and root mean square deviation
(RMSD) for V5 are 0.88 mm and 9.71 mm whereas for V6 it is -0.15 mm and 9.60
mm respectively.
Scatter plots of daily gauge data vs.
3B42 data show that V6
correlates better with gauge (0.44) than V5 (0.37). The distribution of daily 3B42V6 rain rate is quite similar to gauge while V5 has more rain in the range 5-20
mm/day.
The V6 TRMM algorithm shows improvement over V5 in terms of the
bias, RMS dierence, and mean absolute dierence.
Adeyewa and Nakamura
(2003) have shown that TRMM PR data overestimates
rain in the tropical rain forest region of Africa when compared with Global Pre-
2
DATA
4
Rudolf , 1993).
cipitation Climatology Centre (GPCC) rain gauge data (
The 3B43
product, which is the TRMM merged analysis on monthly scale, has the closest
agreement with rain gauge data.
Nicholson et al.
(2003a), using rain gauge data
from 515 stations over North Africa shows 3-4% bias for GPCC or GPCP with
reference to seasonal rainfall elds (1988-1994).
Nicholson et al. (2003b) nd excel-
lent agreement of TRMM-adjusted GOES precipitation Index (AGPI) and TRMM
merged rainfall analysis with high density (920 stations) gauge data over west Africa
on monthly to seasonal time scale. The RMSD of both satellite-derived products is
∼0.6
mm/day at seasonal scale and 1 mm/day at monthly resolution. The bias of
AGPI is only 0.2 mm/day whereas the TRMM-merged product shows no bias over
◦
◦
West Africa. The 1 x 1 latitude/longitude product also shows excellent agreement
at the seasonal scale and good agreement at monthly scale.
◦
◦
The main dierence between V5 and V6 is resolution - 3B42-V5 is on a 1 x 1 grid
◦
◦
and covers the global tropics (40 S-40 N latitude), whereas the V6 product is on
◦
◦
◦
◦
a 0.25 x 0.25 grid and covers 50 S-50 N latitude. Both the datasets are available
◦
◦
with daily time resolution. In this report we use daily 1 x 1 3B42-V5 data and daily
◦
◦
3B42-V6 data regridded to 1 x 1 .
2.2
GPCP
The Global Precipitation Climatology Project (GPCP) was established by the World
Climate Research Program (WCRP) to address the problem of quantifying the
distribution of precipitation around the globe over many years.
GPCP has pro-
moted the development of an analysis procedure for blending various estimates
together to produce the necessary global gridded precipitation elds.
Human et al.,
release GPCP produced Version 1 data product (
In its rst
1997) which is
◦
◦
a globally complete, on monthly time scale at 2.5 x 2.5 latitude/longitude resolution.
It uses Special Sensor Microwave Imager (SSM/I) microwave observations
and geo-IR estimates. Gauge information is also included in its nal step. GPCP
1DD daily data is a companion to the GPCP Version 2 satellite-gauge (SG) combination (
Adler et al.,
2003).
The main components of the GPCP datasets are
IR and Microwave observations.
In this rst release, the 1DD uses the best"
quasi-global observational estimators of underlying statistics to adjust quasi-global
observational datasets that have desirable time/space coverage.
Specically, Spe-
◦
◦
cial Sensor Microwave Imager (SSM/I; 0.5 x0.5 by orbit, GPROF algorithm) provides the fractional occurrence of precipitation, and GPCP Version 2 SG combi-
◦
◦
nation (2.5 x2.5 monthly) provides monthly accumulation of precipitation to algo-
2
DATA
5
rithms applied to geosynchronous-orbit IR (geo-IR) brightness temperature (Tb)
◦
◦
◦
◦
histograms (1 x 1 in the band 40 N-40 S, 3-hourly), low-orbit IR (leo-IR) GOES
Precipitation Index (GPI; same time/space grid as geo-IR), TIROS Operational
◦
◦
Vertical Sounder (TOVS; 1 x 1 on daily nodes, Susskind algorithm,(
Suskind et al.,
◦
◦
1997)), and Atmospheric Infrared Sounder (AIRS; 1 x 1 on daily nodes, Susskind
algorithm).
The GPCP 1DD product is derived only from precipitation estimates based on
Hu-
satellite data; no rain gauge information is included in this product directly (
man et al.,
2001).
The monthly totals accumulated from the daily precipitation
elds are only scaled to t the monthly totals of GPCP's combined precipitation
product (Version 2), which includes GPCC rain gauge analyses. The GPCP Combined Product Version 2 is a near real-time product, and includes monthly precipitation data from global telecommunication system (GTS) stations, i.e. synoptic
weather stations and climate stations - there are about 80 stations for India (Dr.
Udo Schneider, personal communication, 2006). The TRMM merged product and
GPCP analysis both use very similar procedures with dierent initial inputs, thus
simplifying the intercomparison.
2.3
IMD
◦
◦
A high resolution (1 x 1 ) gridded daily rainfall data from the IMD (
Rajeevan et al.,
2005) is used to validate GPCP and TRMM merged analysis 3B42 daily rainfall
products over Indian land. IMD uses the
Shepard
(1968) interpolation technique for
gridding data from individual stations, while GPCC uses the
Willmott et al. (1985)
method for interpolation. The IMD product uses gauge data from 1803 stations to
estimate accumulated rainfall in the 24 hours ending 0830 IST (0300 Z). The basic
dierence between GPCP and IMD data is that in GPCP, IR and Microwave data
are used for rainfall estimation with the help of rain gauge calibration.
Another
dierence is that the GPCP data gives daily accumulated rainfall starting at 22:30
Z of previous day to 22:30 Z of the day named.
Recently
Rajeevan et al.
(2006) have compared other global datasets with 53
years (1951-2003) of IMD gridded rainfall data from gauges, and identied active
and break periods during the southwest monsoon season. Comparison on monthly
scale with the VASClimo dataset (
Beck et al., 2005) shows dierences of the order
of 50 mm over most of India. However, along the west coast of India IMD rainfall
values are higher than VASClimo values.
On inter-annual time scales, all major
drought and excess years are captured by the VASClimo dataset. The correlation
3
RESULTS AND DISCUSSION
6
Figure 1: The region of study. Dierent regions are shown in outline: CI-Central
India, EI-Eastern India, NI-North India WC-West Coast of India, MC-Myanmar
Coast, BoB-Bay of Bengal, and EIO-Equatorial Indian Ocean.
coecient for the period 1951-1995 between IMD and VASClimo data is 0.97.
3
Results and Discussion
As mentioned above the time intervals used for estimating accumulated daily rain
by the IMD and 3B42 datasets are dierent (Table 1).
In order to co-locate the
time, IMD data has been shifted in time by one day. The dierent regions used for
comparison are shown in Figure 1. Three representative areas over Indian land i.e.
Central India (CI), Eastern India (EI) and North India (NI) are chosen based on the
consideration that daily variability of rain is more or less spatially uniform within
Goswami et al., 2006).
each of these regions (
We have also compared GPCP and 3B42-V6 rainfall data over oceanic regions.
Four representative boxes have been chosen (Figure 1). Two are in coastal regions
along the Myanmar coast and the west coast of India where there is intense precip-
Xie et al., 2006).
itation during the Indian summer monsoon (
The other two boxes
are chosen to capture oceanic rain associated with the intertropical convergence zone
(ITCZ) in the Bay of Bengal (BoB) and equatorial Indian Ocean (EIO).
3
RESULTS AND DISCUSSION
7
Figure 2: Monsoon season (June-September) 1998-2003 mean rain rate (mm/day).
The average rain rate spatially averaged over India (mm/day) is shown at top right
in each panel.
3.1
Spatial distribution of rainfall over land
The seasonal (June-September; JJAS) 1998-2003 mean rainfall over India for IMD,
GPCP and 3B42-V5 and V6 is shown in Figure 2.
We have used the IMD grid
to mask out oceanic regions from the satellite products. The GPCP and 3B42-V5
products do not show the observed high rainfall over the west coast of India, the
northeast, or the Himalayan foothills.
The high rainfall over the Western Ghats
and the Himalayan foothills is attributed to orography; it is captured by seasonal
Xie et al., 2006).
TRMM PR measurements and SSM/I data (
3B42-V6 captures the
pattern of orographic rain seen in the IMD data, but consistently underestimates
the rainfall amount. This product also underestimates rain over the Gangetic plains
3
RESULTS AND DISCUSSION
Figure 3:
8
All India seasonal (JJAS) total rainfall in cm, 1998-2003, from IMD
(black), GPCP (red) and 3B42-V6 (blue).
Table 2: All India seasonal (JJAS) total rainfall (cm), 1998-2003.
Y ear
IM D
GP CP
3B42 − V 6
1998
94.75
77.00
78.41
1999
83.44
69.48
70.59
2000
78.81
69.60
69.71
2001
81.94
72.37
75.48
2002
71.04
60.57
63.47
2003
94.20
77.00
81.60
Mean
84.03
71.00
73.21
Standard Deviation
9.15
6.11
6.58
(see Figure 2). The six-year seasonal mean 3B42-V6 rain is 6.7 mm/day, compared
to the IMD value of 7.8 mm/day. Note that the seasonal mean rainfall for the 1951-
Rajeevan et al., 2005).
2003 IMD data is 7.7 mm/day (
An interesting feature is the
region of suppressed rain east of the eastern Ghats in peninsular India. This rain
shadow region is present in 3B42-V6 data, but not in 3B42-V5 and GPCP.
The interannual variation of seasonal mean All-India rainfall is shown in Figure
3 for 1998-2003. The phase of interannual variability is captured correctly by the
GPCP and 3B42-V6, but both products underestimate the total seasonal rain by
7-18 cm in dierent years.
The seasonal mean for individual years for 1998-2003
and its six-year average, as well as the six-year seasonal standard deviation (SD)
are shown in Table 2. The coecient of variation (i.e SD divided by mean) of IMD
seasonal rainfall over this six year period is 10 %, close to the 11% for the 1951-2003
Rajeevan et al.,
IMD data (
2005).
The GPCP and 3B42-V6 variability are lower
than the IMD values, but 3B42-V6 has somewhat higher year-to-year variability
3
RESULTS AND DISCUSSION
9
Figure 4: 1998-2003 Seasonal (June-September) daily standard deviation (mm/day).
than GPCP.
The mean as well as daily standard deviation (computed with respect to the
six-year JJAS mean) of JJAS all India rain during 1998-2003 is shown in Table 3.
The spatial distribution of daily SD of IMD, GPCP and 3B42 rain during JJAS
1998-2003 are shown in Figure 4. Although the large-scale spatial distributions of
GPCP and 3B42-V6 SD might be considered reasonable, the large SD in the IMD
data along the west coast of India is captured only by the 3B42-V6 data. Averaged
over all of India, the daily SD of GPCP and 3B42 V6 rain are within 20 % of the
IMD daily SD (11.4 mm/day) (see Table 3).
A monsoon low-pressure system gave intense rain over central India in midAugust 2002. Figure 5 shows the 15-18 August average rain for IMD, GPCP, 3B42V5 and V6. The maximum rain rate in the IMD data is about 93 mm/day at the
3
RESULTS AND DISCUSSION
10
Table 3: All India seasonal (JJAS) mean rain rate and daily standard deviation
1998-2003 (mm/day)
Y ear
M ean
Std.
IMD
7.8
11.4
GPCP
6.3
10.0
3B42-V5
7.1
8.4
3B42-V6
6.7
9.6
centre of a circular rain patch. GPCP and 3B42-V5 grossly underestimate the peak
rain rates, but the maximum rain in 3B42-V6 is about 76 mm/day, closer to the
IMD value.
3.2
Time series over land
In order to examine how the satellite based products capture time variability we
◦
◦
◦
◦
consider two 1 x1 boxes, one centered at 17.5 N, 82.5 E near the east coast of
◦
◦
India, and another at 21.5 N, 77.5 E in Central India. Figure 6a shows daily time
series for JJAS of 1998-2003 at the east coast grid point. The mean of IMD, GPCP,
3B42-V5 and 3B42-V6 rain rates are 4.8, 6.6, 7.5 and 5.6 mm/day respectively, and
the SD are 9.9, 9.9, 8.1 and 9.6 mm/day. The mean and SD of 3B42-V6 are closest
to IMD observations, although all satellite products overestimate the mean. Note
that this grid point is in a localized rain shadow region (Figure 2). Figure 6b shows
time series at the Central India location, which lies in the monsoon trough zone. All
satellite products underestimate the SD by 30 % or more. Figures 7a and 8a show
the daily rain rate at these two grid points in JJAS of 2002 and 2003. Heavy rainfall
events are in general underestimated in 3B42 and GPCP data. The absence of rain
for most of July 2002 is noteworthy. There are occasions of spurious heavy rain in
the satellite data. This can be seen more clearly in Figures 7b and 8b respectively,
which show two active spells in 2002 and 2003 respectively. These results suggest
that although the satellite products are reasonably good on large spatial scales, at
individual grid points the agreement with gauge data is far from perfect (see Tables
4 and 5).
Time series averaged over three boxes over India (Figure 1) are shown in Figures
9, 10 and 11 respectively. Figure 9 shows rain averaged over Central India (74.5-
◦
◦
86.5 E, 16.5-26.5 N) for JJAS 1998-2003.
The six-year mean and daily standard
deviation are mentioned in the upper corner of each panel; the red line is the mean
3
RESULTS AND DISCUSSION
11
Figure 5: Average rain rate (mm/day) for 15-18 August 2002.
of all six years. The major active spells are picked up by GPCP and 3B42-V6 data
in all years but GPCP tends to overestimate peak rainfall. The variability of 3B42
◦
rain is close to that of IMD rain over the smaller East India box (22.5-26.5 N, 84.5◦
◦
◦
88.5 E) (Figure 10). Time series over the North India box (28.5-32.5 N, 75.5-80.5 E)
are shown in Figure 11; over this region the mean and SD of 3B42-V6 are close to
the IMD values. 3B42-V6 data may be the most suitable for study of variability on
daily and longer time scales.
3.3
Spatial distribution over the ocean
The eastern Bay of Bengal has the heaviest precipitation where the southwest monsoon impinges on the coastal mountains of Myanmar.
Xie et al. (2006), using TRMM
3
RESULTS AND DISCUSSION
12
Figure 6: Daily rain rate (mm/day) at a single grid point (a) near the east coast of
India and (b) in Central India, June-September of 1998-2003.
3
RESULTS AND DISCUSSION
13
Figure 7: Daily rain rate at the same two grid points as in Figure 6, but for (a)
June-September 2002 and (b) 15 Aug-5 September 2002.
3
RESULTS AND DISCUSSION
14
Figure 8: Daily rain rate at the same two grid points as in Figure 6, but for (a)
June-September 2003 and (b) 1 Jul - 20 Jul 2003.
3
RESULTS AND DISCUSSION
15
Table 4: Daily average rain (mm/day) at two grid points during an active monsoon
period in 2002
◦
◦
77.5 E & 21.5 N
◦
◦
82.5 E & 17.5 N
Date
3B42 − V 6 IM D
GP CP
3B42 − V 6
IM D
GP CP
15-AUG-2002
15.06
31.10
13.29
0.01
0.12
0.00
16-AUG-2002
53.68
2.33
24.37
2.87
0.00
0.01
17-AUG-2002
57.13
34.95
58.32
0.01
0.00
0.36
18-AUG-2002
13.82
30.06
44.08
1.26
0.01
0.00
19-AUG-2002
1.44
0.01
0.00
0.77
0.00
0.00
20-AUG-2002
0.41
3.48
0.12
3.16
9.17
0.91
21-AUG-2002
0.02
0.00
0.00
4.32
27.58
5.72
22-AUG-2002
37.22
13.09
16.62
3.58
28.46
19.11
23-AUG-2002
38.41
36.23
21.37
0.17
2.10
2.76
24-AUG-2002
41.71
23.04
18.72
0.84
0.00
0.00
25-AUG-2002
7.60
0.00
0.02
1.05
0.43
1.00
26-AUG-2002
0.00
0.00
0.00
0.00
5.28
1.93
27-AUG-2002
0.00
0.00
0.00
0.08
21.65
20.41
28-AUG-2002
4.28
4.36
7.83
1.28
11.69
33.71
29-AUG-2002
16.61
0.00
1.06
3.37
13.68
1.77
30-AUG-2002
19.23
20.49
40.40
20.59
11.98
2.02
31-AUG-2002
27.21
24.22
17.03
0.03
12.63
1.32
01-SEP-2002
7.07
28.54
34.49
3.56
2.78
3.85
02-SEP-2002
6.53
9.13
15.89
33.27
0.99
4.79
03-SEP-2002
10.72
7.87
0.73
8.57
1.30
6.84
04-SEP-2002
16.00
9.40
14.47
0.36
2.28
1.32
05-SEP-2002
61.25
46.76
42.69
8.34
0.10
3.00
PR and other satellite rainfall estimates together with satellite winds and a regional
climate model, show that rain bands related to orography are not just localized
phenomena but form the core of basin-scale convection over the tropical Indian
Ocean and South China Sea.
Figure 12 shows the 1998-2003 JJAS mean rainfall
over the Indian Ocean from GPCP (contour) and 3BV42-V6 (shaded) data.
The
3B42-V6 rainfall o the Myanmar coast reaches 18-22 mm/day, whereas the GPCP
maximum is 14-16 mm/day; the maximum rain in GPCP is shifted northward relative to 3B42-V6. However over the southeast Indian Ocean GPCP underestimates
◦
◦
rain compared to 3B42-V6. In the latitude band 0 -10 S, GPCP rain is generally 2
mm/day lower than 3B42-V6 rain (Figure 12). As noted earlier, 3B42-V6 rain is a
better representation over the west coast of India and the foothills of the Himalayas.
3
RESULTS AND DISCUSSION
16
Table 5: Daily average rain (mm/day) at two grid points during an active monsoon
period in 2003
◦
◦
77.5 E & 21.5 N
◦
◦
82.5 E & 17.5 N
Date
3.4
3B42 − V 6 IM D
GP CP
3B42 − V 6
3.61
11.52
2.32
0.73
8.04
9.23
19.26
18.27
25.92
30.24
34.42
1.09
28.09
34.13
0.71
2.61
4.33
31.47
13.58
9.47
9.42
12.41
23.95
13.57
14.26
44.52
0.60
16.24
1.84
0.00
0.00
28.46
0.00
0.20
1.26
0.00
0.10
0.07
0.00
0.11
1.20
11.36
8.92
10-JUL-2003
0.00
0.00
0.85
0.47
9.36
5.97
11-JUL-2003
0.00
1.48
0.42
1.61
0.00
0.00
12-JUL-2003
0.85
10.35
6.72
0.00
6.36
0.84
13-JUL-2003
6.31
3.24
0.00
0.56
7.90
16.89
14-JUL-2003
0.00
36.54
11.27
0.04
49.23
19.06
15-JUL-2003
4.29
2.28
7.37
60.22
26.84
41.14
16-JUL-2003
14.80
2.67
6.73
8.51
0.24
13.31
17-JUL-2003
8.57
0.00
0.53
7.63
18.61
29.10
18-JUL-2003
0.98
4.69
13.00
8.59
3.47
6.54
19-JUL-2003
16.43
12.91
5.72
0.64
0.00
0.09
20-JUL-2003
15.24
20.99
4.06
3.06
2.49
3.82
IM D
GP CP
01-JUL-2003
0.00
0.00
6.87
02-JUL-2003
29.51
1.15
1.43
03-JUL-2003
1.85
25.75
04-JUL-2003
23.77
42.25
05-JUL-2003
9.29
06-JUL-2003
07-JUL-2003
08-JUL-2003
09-JUL-2003
Time series over the Ocean
We next compare GPCP and 3B42 rainfall over four boxes in the Bay of Bengal,
eastern Arabian Sea and Equatorial Indian Ocean (Figure 1).
Figure 13a shows
◦
◦
◦
◦
daily rain over the Bay of Bengal (BoB; 11.5 -15.5 N, 80.5 -92.5 E; upper panel) and
◦
◦
◦
equatorial Indian Ocean (EIO; 4.5 S-0.5 S, 80.5-92.5 E; lower panel). Rain events in
the two datasets are in phase, but heavy rain events show larger values in 3B42-V6
than GPCP; the daily SD of 3B42-V6 is somewhat higher than GPCP. This could
be because of the blending of high spatial resolution PR data in 3B42-V6 data. The
qualitative results are the same as over BoB. A tendency for sustained high rainfall
events in the two regions to occur out of phase is visible to the eye. To compare the
heavy rain o the Myanmar coast (MC) and the west coast of India (WCI), we have
chosen rhombus-shaped boxes. Figure 13b shows the JJAS 19998-2003 time series
3
RESULTS AND DISCUSSION
17
Figure 9: Daily rain rate (mm/day) averaged over Central India for JJAS of 19982003. The six-year mean and daily standard deviation (mm/day) for each dataset
is mentioned in the respective panels.
for MC (upper panel) and WCI (lower panel). For both regions the mean and SD
of 3B42-V6 rain are higher than GPCP. Although the algorithm is similar for the
two products, high resolution PR data goes into the 3B42-V6 estimate- this might
be responsible for better representation of high rainfall events. A notable dierence
between the two oshore regions is the occurrence of sustained low rainfall periods
in WCI but not MC. For example, note the periods longer than a week with
<
5
mm/day WCI rain in August and September 1999; July and September 2000; July
and September 2002, and August and September 2003.
3.5
Intraseasonal and interannual variability
We suppress high frequency variability by using a simple seven-day running mean,
◦
◦
◦
◦
to illustrate intraseasonal variability over Indian land (75 - 82 E, 22 - 26 N) and
North Bay of Bengal (NBB; 86
◦
- 93
◦
E, 16
◦
- 20
◦
N) (Figure 14). The box over land
3
RESULTS AND DISCUSSION
18
Figure 10: Daily rain rate (mm/day) averaged over East India for JJAS of 19982003. The six-year mean and daily standard deviation (mm/day) for each dataset
is mentioned.
Figure 11: Daily rain rate (mm/day) averaged over North India for JJAS of 19982003. The six-year mean and daily standard deviation (mm/day) for each dataset
is mentioned.
3
RESULTS AND DISCUSSION
Figure 12:
19
1998-2003 June-September mean rain rate (mm/day) over the north
Indian Ocean and south Asia from 3B42-V6 and GPCP.
(not shown in Fig 1) has been chosen at more or less the same latitude as the NBB
box, and to have the same spatial extent. The rain has pronounced intraseasonal
oscillations (ISO) with periods of 10 days and longer over NBB and the Indian land
box in 2002.
However, in 2003, pronounced ISO is observed over NBB, but it is
not prominent over land. The result for 2003, but not 2002, is consistent with the
study of intraseasonal variability by
Hartmann and Michelsen
(1989), who report
◦
little evidence of intraseasonal activity in Indian daily rainfall north of 22 N. Figure
15 shows the JJAS mean rain over the Indian region and Bay of Bengal in two
contrasting years 2002 (a drought monsoon year) and 2003 (normal monsoon year).
The spatial distribution patterns is similar in both datasets.
In 2002 northwest
India has very low rainfall (∼2-4 mm/day) compared to 2003 (6-8 mm/day), or
to the 1998-2003 mean (Figure 2). The main dierences in 2002 and 2003 rainfall
are over central and western India, in agreement with IMD observations. The 2003
rainfall is signicantly larger than 2002 rainfall in the WCI. However, there is no
signicant dierence between 2002 and 2003 over the eastern Bay of Bengal. Overall,
GPCP estimates are lower than 3B42-V6. Figure 16 shows the interannual variation
of seasonal rainfall over Central India (see Figure 1). There is a biennial oscillation
3
RESULTS AND DISCUSSION
20
Figure 13: Daily rain rate (mm/day) averaged over (a) North Bay of Bengal and
Equatorial Indian Ocean, and (b) o the West Coast of India and the Myanmar
Coast (see regions in Fig 1), for JJAS of 1998-2003.
3
RESULTS AND DISCUSSION
21
◦
◦
◦
Figure 14: Seven-day running mean rain rate over Central India ( 75 -82 E,22 ◦
◦
◦
◦
◦
26 N and north Bay of Bengal ( 86 - 93 E, 22 - 26 N for May-September (a)
2002 and (b) 2003.
4
CONCLUSIONS
22
Table 6: RMS dierence (mm/day) between daily satellite and IMD rain rate at
two grid points, JJAS 1998-2003
◦
◦
82.5 E 17.5 N
◦
◦
77.5 E, 21.5 N
6.23
6.72
GPCP
3B42-V5
6.62
6.79
3B42-V6
5.84
5.89
Table 7: RMS dierence (mm/day) between daily satellite and IMD rain rate averaged over the regions CI, EI and NI, JJAS 1998-2003
CI
EI
NI
GPCP
2.72
5.47
3.37
3B42-V5
2.21
4.66
3.49
3B42-V6
2.60
5.08
3.28
in gauge as well as satellite data. The amplitude of this oscillation is
IMD gridded data, whereas it is
∼5
∼10 cm in the
cm in 3B42-V6 and GPCP data. In all years,
the seasonal total GPCP rainfall is closer to the IMD total than 3B42-V6.
4
Conclusions
Daily satellite-based rainfall from GPCP and the TRMM merged product 3B42
(versions 5 and 6) have been validated against daily gridded data from rain gauges
released recently by the India Meteorological Department, for the summer monsoon
seasons of 1998-2003. The GPCP and 3B42-V5 products reproduce only the broadest features of mean monsoon rainfall over India. If one looks at the shape and size
of high and low rainfall regions in any detail, GPCP and 3B42-V5 may be considered inadequate representations. However, the patterns of 3B42-V6 mean monsoon
Xie et al., 2006), are reasonably close
rainfall, including those related to orography (
to the observed patterns from IMD data (Figure 2). This may be because of the use
of TRMM precipitation radar data in 3B42-V6. All satellite products underestimate
seasonal mean all-India rainfall (Table 3). The six-year all-India mean of 3B42-V6
is 14% lower than the IMD mean; in individual years, 3B42-V6 underestimates seasonal mean rainfall by 8% to 18%. The six-year daily standard deviation of 3B42-V6
all-India rain is 16% lower than the IMD value (Table 3); the spatial distribution of
4
CONCLUSIONS
Figure 15:
23
Seasonal mean (June-September) rain rate (mm/day) for GPCP and
3B42-V6 over the Bay of Bengal, far eastern Arabian Sea and Indian region in (a)
2002 and (b) 2003.
daily variability in the 3B42-V6 product is reasonably close to IMD data, except for
the Gangetic plains, the Himalayan foothills and parts of east Central India (Figure
4). Both the mean and daily standard deviation of 3B42-V6 rain are close to the
IMD values over the Central India, East India and North India boxes (Figures 9,
10, 11). At individual grid points, the RMS dierence between daily satellite and
gauge-based rain can be high (Figures 6, 7, 8; Table 6), but over larger regions the
agreement is better (Table 7).
Interannual variability in satellite rain has the right phase but the amplitude is
REFERENCES
24
Figure 16: 1998-2003 interannual variation of seasonal (JJAS) total rainfall (cm)
over the Central India box from IMD (black), GPCP (red) and 3B42-V6 (blue).
generally underestimated, specially over orography (Figures 3 and 16). Two satellite
rain products (GPCP and 3B42-V6) have been compared over the tropical Indian
Ocean.
Although dierences are not large in the open ocean, over orographically
active regions (o the Myanmar coast and the west coast of India) GPCP estimates
are substantially lower than 3B42-V6, specially during episodes of intense rain. This
is likely because of the input from TRMM PR rain in the 3B42-V6 product. The
daily satellite data, particularly the 3B42-V6 product, have great potential for the
study of rainfall variability on synoptic to interannual time scales over the the Indian
monsoon region and the tropical Indian Ocean.
Acknowledgements
We thank V. Venugopal for several useful suggestions, and Jaison Kurian for helping
with Latex and Ferret. This work is supported by DOD. We wish to acknowledge
M. Rajeevan for providing the IMD data, and NASA/GSFC for providing the 3B42
and GPCP data on their ftp site. Software packages Ferret and GrADS have been
used for data analysis and graphics.
References
Adeyewa, Z. D., and Â. Nakamura (2003), Validation of TRMM Radar Rainfall
Data over Major Climatic Regions in Africa,
42 (2), 331347.
Journal of Applied Meteorology,
Adler, R. F., D. T. Bolun, S. Curtis, and E. J. Nelkin (2000), Tropical rainfall dis-
REFERENCES
25
tributions determined using TRMM combined with other satellite and raingauge
information,
Journal of Applied Meteorology., 39, 20072023.
Adler, R. F., G. J. Human, A. Chang, R. Ferraro, P. P. Xie, J. Janowiak, B. Rudolf,
U. Schneider, S. Curtis, D. Bolvin, A. Gruber, J. Susskind, P. Arkin, and E. Nelkin
(2003), The Version-2 Global Precipitation Climatology Project (GPCP) monthly
precipitation analysis (1979-present),
Journal of Hydrometeorology., 4, 11471167.
Beck, C., J. Grieser, and B. Rudolf (2005), A new monthly precipitation climatology
for the global land areas for the period 1951-2000.,
Tech. rep., German Weather
Service, Oenbach,Germany.
Chokngamwong,
land
daily
R.,
gauge
and
rainfall
L.
Chiu
comparison
(2005),
,
TRMM
unpublished
and
Thai-
taken
from
http://ams.confex.com/ams/pdfpapers/103039.pdf.
Goswami, B. N., V. Venugopal, D. Sengupta, M. S. Madhusoosudanan, and P. K.
Xavier (2006), Increasing trend of extreme rain events over India in a warming
environment,
Science, In Press.
Haddad, Z. S., E. A. C. Smith, D. Kummerow, T. Iguchi, M. R. Farrar, S. L. Durden,
M. Alves, and W. S. Olson (1997), The TRMM day-1 radar/radiometer combined
rain-proling algorithm.,
J. Meteor. Soc. Japan, 75, 799809.
Hartmann, D. L., and M. L. Michelsen (1989), Intraseasonal periodicities in Indian
rainfall,
Jour. of Atmos. Sci, 46, 28382862.
Human, G. J., R. F. Adler, P. Arkin, A. Chang, R. Ferraro, A. Gruber, J. Janowiak,
A. McNab, B. Rudolf, and U. Schneider (1997), The Global Precipitation Climatology Project (GPCP) combined precipitation dataset.,
Soc., 78, 520.
Bull. Amer. Meteor.
Human, G. J., R. F. Adler, M. M. Morrissey, D. T. Bolvin, S. Curtis, R. Joyce,
B. McGavock, and J. Susskind (2001), Global Precipitation at One-Degree Daily
Resolution from Multisatellite Observations (2001),
2 (1), 3650.
Journal of Hydrometeorology,
Narayanan, M. S., S. Shah, C. M. Kishtawal, V. Sathiyamoorthy, M. Rajeevan,
and R. H. Kriplani (2005), Validation of TRMM merge daily rainfall with IMD
raingauge analysis over Indian land mass,
Ahmedabad,India.
Tech. rep., Space Applications Centre,
REFERENCES
26
Nicholson, S. E., B. Some, J. McCollum, E. Nelkin, D. Klotter, Y. Berte, B. M.
Diallo, I. Gaye, G. Kpabeba, O. Ndiaye, J. N. Noukpozounkou, M. M. Tanu,
A. Thiam, A. A. Toure, and A. K. Traore (2003a), Validation of TRMM and
Other Rainfall Estimates with a High-Density Gauge Dataset for West Africa. Part
I: Validation of GPCC Rainfall Product and Pre-TRMM Satellite and Blended
Products,
Journal of Applied Meteorology, 42 (10), 13371354.
Nicholson, S. E., B. Some, J. McCollum, E. Nelkin, D. Klotter, Y. Berte, B. M.
Diallo, I. Gaye, G. Kpabeba, O. Ndiaye, J. N. Noukpozounkou, M. M. Tanu,
A. Thiam, A. A. Toure, and A. K. Traore (2003b), Validation of TRMM and
Other Rainfall Estimates with a High-Density Gauge Dataset for West Africa.
Part II: Validation of TRMM Rainfall Products,
42 (10), 13551368.
Journal of Applied Meteorology,
Rajeevan, M., J. D. Kale, and J. Bhate (2005), High-resolution gridded daily rainfall data for Indian monsoon studies,
Tech. Rep. 2,
National Weather Service,
Pune,India.
Rajeevan, M., J. Bhate, J. D. Kale, and B. Lal (2006), High resolution daily gridded
rainfall data for the Indian region: Analysis of break and active monsoon spells,
Curr. Sci., 91 (3), 296306.
Rudolf, B. (1993), Management and analysis of precipitation data on a routine basis, in
precipitation and evaporation,
Proc. Int. WMO/IAHS/ETH Symp. Pre-
cipitation and Evaporation, pp. 6976, Slovak Hydrometeorology Institution,
Bratislava, Slovakia.
Shepard, D. (1968), A two-dimensional interpolation function for irregularity spaced
data, In Proc. 1968 ACM Natl. Conf., pp. 517524, ACM Natl. Conf.
Suskind, J., P. Piraino, L. Irsdell, and A. Mehta (1997), Characteristics of the TOVS
pathnder path a dataset,
Bull. Amer. Meteor.Soc., 78, 14491472.
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, 516.
Xie, S. P., H. Xu, N. H. Saji, and Y. Wang (2006), Role of Narrow Mountains in
Large-Scale Organization of Asian Monsoon Convection,
34203429.
Jour. of Climate, 19,