Whether Calculation

MONSOON FORECASTING WITH A LIMITED AREA
NUMERICAL WEATHER PREDICTION SYSTEM
SMRC – NO. 11
SAARC METEOROLOGICAL RESEARCH CENTRE

DHAKA, BANGLADESH
___________________________________________________________________________
MONSOON FORECASTING WITH A LIMITED AREA
NUMERICAL WEATHER PREDICTION SYSTEM
(A joint study by SAARC Meteorological Research Centre, Dhaka;

and India Meteorological Department, New Delhi)
SMRC – NO. 11
SAARC METEOROLOGICAL RESEARCH CENTRE

DHAKA, BANGLADESH
2
Author:
K. Prasad
Head, Theoretical Division
SAARC Meteorological Research Centre
Dhaka
Bangladesh
Contributed by:
S.K. Roy Bhowmik

Y.V. Rama Rao
India Meteorological Department
New Delhi
India
Assisted by:
Majajul Alam Sarker

Mizanur Rahman
Romee Afroz
Dhaka
Published by:
Director
Abhawa Bhaban, 7th floor, Agargaon
Dhaka-1207
Bangladesh
Published:
May, 2005
3
FOREWORD
The First Special and the Ninth Meeting of the Governing Board of SMRC held at Dhaka
from 19-22 October, 2003 laid considerable emphasis on numerical weather prediction
(NWP) as one of the key areas of research in SMRC. The Governing Board also
recommended that the research activities of SMRC should focus on problems of common
interest to the SAARC Member States and should be directed towards enhancing their
capability in the field of early warning. The Board highlighted the importance of
collaboration between the research scientists of SMRC and those of the member countries in
pursuing the research goals of SMRC. A collaborative research endeavour with some
international institutions was also encouraged.
Occurrence of heavy rainfall and the resulting large scale damages caused by inundation
and river floods are a common feature during the southwest (summer) monsoon affecting a
large population and are of interest to many countries in the SAARC region. An
understanding of the atmospheric circulation systems responsible for causing these events
and their accurate forecasting are thus crucial for the early warning systems in all the
countries of the Region. SMRC has identified the problem of understanding and prediction
of heavy rainfall events with numerical weather prediction (NWP) techniques as one of the
items of activity under its research programmes, in keeping with the above goals. I am happy
to note that a study on “Monsoon forecasting with a limited area numerical weather
prediction system” has recently been completed by the Theoretical Division of SMRC. The
limited area model, originally developed at the Florida State University, USA, and adopted in
India Meteorological Department (IMD) for NWP research and operational forecasting has
been acquired in SMRC for carrying out further research with the model.
I would like to place on record my appreciation for the useful work undertaken by the
Theoretical Division and compiling the present Report. The Report has been peer-reviewed
by India Meteorological Department, New Delhi, who have commended the work. I am
grateful to India Meteorological Department for their support and cooperation in the activities
of SMRC. It is my sincere hope that further development work on this extremely important
topic will continue in SMRC in collaboration with the concerned institutions, with a view to
benefiting the whole SAARC region.
(Md. Akram Hossain)

Director
May, 2005 SAARC Meteorological Research Centre, Dhaka
4
PREFACE
Application of dynamical models in the tropical weather forecasting has come of age.
Several studies have demonstrated that the numerical weather prediction models – global and
regional – are now capable of handling the tropical weather disturbances quite well. In the
recent past global models have rapidly evolved towards a substantial increase in the
horizontal resolution and emerged as the quasi-static type mesoscale models. While the
global models can provide good forecasts of large scale weather systems and have the
obvious advantage of dispensing with the need for lateral boundary conditions, which are
invariably required for running a limited area model, the regional models have their own
place in the field of NWP modeling. The regional models, through the nesting approach, can
be run on very fine horizontal resolutions of even 5-10 km and with a non-hydrostatic
dynamics, which can handle the high impact and disastrous weather events. Several such
promising limited area models have emerged in recent years.
A limited area tropical numerical weather prediction model based on Florida State
University has been implemented in the SAARC Meteorological Research Centre (SMRC)
Dhaka on a SUNFIRE V480 server for NWP research. Several versions of this model have
been used in tropical prediction experiments, especially for monsoon forecasting, by various
groups. Results of these experiments have amply demonstrated the capability of the FSU
model in forecasting the tropical weather systems effectively. The above model has been
used in the study reported here for diagnostic studies of flood producing heavy rainstorms in
Bangladesh and neighbouring northeastern states of India. Results of a few experiments on
forecasting the movement of monsoon depressions forming in the Bay of Bengal and moving
across India, and the associated heavy rainfall have also been reported. The basic input data
of flow variables required for running the model experiments were processed from the
ECMWF Reanalysis (ERA-40) data sets downloaded from their ftp site.
The author remains indebted to Professor T.N. Krishnamurti of the Florida State
University, USA for readily supplying the FSU limited area model and all technical support
while the author was serving in India Meteorological Department. The author would like to
convey his grateful thanks to the Director General of Meteorology, India Meteorological
Department for making the model codes available for carrying on further research in SMRC.
I convey my sincere thanks to Mr. Md. Akram Hossain, Director, SMRC for his constant
encouragement. I wish to thank Dr. Dewan Abdul Quadir, Head, Synoptic Division, SMRC
for useful discussions from time to time. Thanks are due to Mr. Majajul Alam Sarker, Mr.
Mizanur Rahman and Ms. Romee Afroz of the Theoretical Division, SMRC for their
assistance in compilation of data and information, and preparation of the manuscript. I
gratefully acknowledge the ready support provided by Bangladesh Meteorological
Department in collection of rainfall data of Bangladesh synoptic stations used in the study. I
thank Dr. S.K. Roy Bhowmik and Dr. Y.V. Rama Rao of India Meteorological Department
for providing their full support and cooperation. It would be our endeavour to work closely
with colleagues in the sister institutions in SAARC Member States for further development of
NWP models in the SAARC region.
(K. Prasad)
Head, Theoretical Division
May, 2005 SAARC Meteorological Research Centre, Dhaka
5
CONTENTS
Page
FOREWORD
PREFACE
List of Tables
List of figures
Summary
1. Introduction 1
2. The basic framework of the FSU limited area model 2

DYNAMICAL FRAMEWORK 2
LATERAL BOUNDARY CONDITIONS 6
PHYSICAL PROCESSES 6
3. Monsoon forecasting – the principal rain bearing weather systems 9

3.1 The seasonal monsoon trough 9
3.2 Monsoon depressions 12
DEFINITION AND STRUCTURE 12
FORMATION AND MOVEMENT 13
ASSOCIATED RAINFALL PATTERN 13
WATER POTENTIAL OF MONSOON DEPRESSIONS 14
IMPACT 14
4. Data and methodology 15
5. Case studies 18
5.1 Analysis and forecasts of heavy rainfall events in Bangladesh in
outstanding excess rainfall and flood years 1987 and 1998 18
5.1.1 Monsoon trough over north Bangladesh during last week
of July to beginning of August 1987 22
5.1.2 Land depression, 26-30 August 1987 29
5.1.3 Monsoon trough over Bangladesh during July, August 1998 36
5.1.4 Discussion 39
5.2 Track forecasts of monsoon depressions 45
5.2.1 Monsoon depression, 20-27 August 1997 45
SYNOPTIC HISTORY 45
TRACK FORECAST 51
5.2.2 Monsoon low, 20-28 June 2002 and heavy floods in Gujarat 57
SYNOPTIC HISTORY 57
FORECAST OF HEAVY RAINFALL 65
5.3 Heavy rainfall and floods in the northeastern region during July 2002 71
5.3.1 Heavy rainfall in the wake of monsoon low of June 2002 71
5.3.2 Heavy rainfall and floods during last week of July 2002 73
4. Concluding remarks 79
References 81
6
List of Tables
Table 1: List of symbols
Table 2: Significant amounts of 24h rainfall (cm) associated with monsoon low,
22-27 June 2002 (Source: IMD web site)
Table 3: Significant amounts of 24h rainfall (cm) from 21 – 24 July 2002 associated with
monsoon trough close to the foothills of Himalayas (Source: IMD web site/BMD)
7
List of Figures
Fig. 1. Vertical structure of the 15-layer version of the limited area model; solid lines are located at the layer
interfaces and dashed lines at the layer means. Variable u , v, T , and q are staggered with respect to the
variables P , σ& , and z .
Fig. 2. Horizontal structure of the limited area model. Variables are staggered following Arakawa C-grid;
Variables q , ln p s , σ& , T , and P (not shown) are also located at the z points.
Fig. 3. Orography used in the limited area model
Fig. 4. Mean sea level pressure (long term mean) in the month of July (based on NCEP/NCAR Reanalysis)
Fig. 5. Map of Bangladesh showing location of synoptic stations
Fig. 6. Map of India showing meteorological subdivisions
Fig. 7. Monsoon rainfall in Bangladesh during July & August - percentage departures of country average
rainfall from long period average (1971-2000); warm and cold episodes refer to the El Nino and La Nina events
Fig. 8. All India monsoon rainfall (July-September) - percentage departures from normal; warm and cold
episodes refer to the El Nino and La Nina events (Source: India Meteorological Department)
Fig. 9. Daily monsoon rainfall (country average) in Bangladesh during the monsoon Season (June to
September) 1987
Fig. 10. Daily monsoon rainfall (country average) in Bangladesh during the monsoon season (June to
September) 1988
Fig. 11. Daily monsoon rainfall (country average) in Bangladesh during the monsoon Season (June to
September) 1998
Fig. 12. Mean sea level pressure analysis on 24 July 1987
Fig. 13. Flow at 850 hPa on 24 July 1987 with orography (x100 M) in the background
Fig. 14. Integrated horizontal moisture flux divergence (x10-5g cm-2s-1) on 24 July 1987
Fig. 15. Flow at 500 hPa on 24 July 1987 (shaded area depicts wind speed in knots)
Fig. 17. 24 hour rainfall forecast (mm) valid 00 UTC of 25 July 1987
Fig. 19. 24 hour precipitation (cm) as observed on 25 July 1987
Fig. 22. 24 hour precipitation (cm) as observed on 1 August 1987
Fig. 23. 24 hour rainfall forecast (mm) valid 00 UTC of 1 August 1987
Fig. 24. Mean sea level pressure analysis on 26 August 1987
8
Fig. 26. Flow at 850 hPa on 21 August 1987 (shaded area depicts wind speed in knots); and track of the
remnant of tropical depression from South China Sea
Fig. 32. Flow at 850 hPa on 26 August 1987 with orography (x100 M) in the background
Fig. 33. 24 hour predicted rainfall (mm) valid 00 UTC of 26 August 1987
Fig. 34. 24 hour rainfall (cm) as recorded on 26 August 1987
Fig. 37. Flow at 500 hPa on 13 August 1998 (shaded area depicts wind speed in knots)
Fig. 40. Flow at 850 hPa on 20 August 1997 with orography (x100 M) in the background
Fig. 43. Mean geopotential contour height at 700 hPa in July 1987 (Source: NCEP/NCAR Reanalysis)
Fig. 50. Relative vorticity (unit: 10-5 s-1) at 850 hPa on 20 August 1997
9
Fig. 56. Flow at 850 hPa 20 August 1997 (shaded area depicts wind speed in knots); and track of monsoon
depression from Bay of Bengal
Fig. 57. 24h forecast flow at 850 hPa valid on 00 UTC of 21 August 1997 (shaded area depicts wind speed in
knots); and track of monsoon depression from Bay of Bengal
Fig. 59. Flow at 850 hPa on 21 August 1997 (shaded area depicts wind speed in knots); and track of monsoon
Fig. 62. Flow at 850 hPa on 26 August 1997 (shaded area depicts wind speed in knots); and track of monsoon
Fig. 65. Observed 24h rainfall (cm) on 27 August 1997
Fig. 67. Mean sea level pressure analysis on 22 June 2002, 00 UTC
Fig. 68. Flow at 850 hPa on 22 June 2002 (shaded area depicts wind speed in knots)
Fig. 75. Relative vorticity (unit: 10-5 s-1) at 850 hPa on 24 June 2002
10
Fig. 79. Meteosat cloud photograph of the monsoon low on 27 June 2002 (Source: IMD web site)
Fig. 80. Meteosat cloud photograph of the monsoon low on 28 June 2002 (Source: IMD web site)
Fig. 81. 24h forecast flow at 850 hPa valid on 00 UTC of 26 June 2002 (shaded area depicts wind speed in
knots)
Fig. 82. 24 hour predicted rainfall (mm) valid 00 UTC of 26 June 2002
Fig. 83. 24h forecast flow at 850 hPa valid on 00 UTC of 27 June 2002 (shaded area depicts wind speed in
knots)
Fig. 85. Integrated horizontal moisture flux divergence (x10-5g cm-2s-1) on 25 June 2002
Fig. 89. Flow at 850 hPa on 28 June 2002 with orography (x100 M) in the background
September) 2002
Fig. 91. Satellite photograph of monsoon trough located close to Himalayan foothills obtained from Meteosat-5
on 2 July 2002 (Source: IMD web site)
Fig. 92. Floods in Brahmaputra river in Assam due to break monsoon conditions in the third and last week of
July 2002 (Photo: courtesy Sentinel, Guwahati, 23 July 2002)
Fig. 93. Mean sea level pressure analysis on 20 July 2002, 00 UTC
Fig. 94. Flow at 850 hPa on 20 July 2002 with orography (x100 M) in the background
Fig. 95. Satellite photograph of monsoon trough located close to Himalayan foothills obtained from Meteosat-5
on 22 July 2002 (Source: IMD web site)
Fig. 96. 24 hour predicted rainfall (mm) valid 00 UTC of 21 July 2002
Notes:
(i) The basic data used in construction of the synoptic maps of mean sea level pressure, flow
pattern and derived products etc. in the above figures have been processed from the ECMWF
ERA-40 Reanalysis data sets downloaded from their ftp site.
(ii) The map outlines and coastlines drawn in all the above figures are approximate and do
not depict any political boundaries. The locations of meteorological stations shown on the
maps are also approximate. These are drawn purely for scientific illustrations. The maps are
produced by the GrADS (COLA) software package used for preparation of graphics
representation of the model output gridded fields in this study.
11
Summary
University (Krishnamurti et al., 1990) has been implemented in the SAARC Meteorological
Research Centre (SMRC), Dhaka for NWP research. The FSU model is a high resolution
multi-level primitive equation model based on a semi-Lagrangian dynamics, cast in sigma
coordinate system. The model is considerably advanced in the dynamical, physical and
initialization aspects and is specially suited for prediction of meso scale events and heavy
rainfall. The SMRC version of the model is run on a horizontal grid resolution of 0.5o x 0.5o
lat./long. and 16 sigma levels (15 layers) in the vertical on a SUNFIRE V480 system. The
vertical extent of the model ranges from σ = 0.05 to σ = 1 at the Earth’s surface. The
model is integrated in a domain 30E to 120E and 15S to 45N. A smoothened orography
obtained from 30 minute terrain field is used in the model.
The above model has been used in the present work for several case studies of heavy
rainfall and movement of monsoon depressions during the summer monsoon season. The
cases chosen for experiments are in respect of some heavy rainfall episodes in 1987 and
1998, the two most prominent excess rainfall and flood years in Bangladesh, as also some
cases of monsoon depressions originating in the Bay of Bengal in regard to their exceptional
behaviour. The main focus of the study is to document the synoptic situations responsible for
causing heavy rainfall episodes and floods in Bangladesh. The efficacy of the limited area
model in forecasting heavy rainfall and movement of monsoon depressions is also
investigated. The analytical structure and track forecast of a monsoon depression of 20-27
August 1997, which originated in the Bay of Bengal and traveled all the way up to northwest
India and adjoining north Pakistan in a rather unusually long track and maintained its
intensity as deep depression until its last stage, is dealt with. Another case of a monsoon low
of 20-28 June 2002, which created catastrophic conditions in Maharashtra and Gujarat in
India due to continuous heavy rains for three to four days, has been studied with an objective
of interpreting the NWP model outputs of some derived fields, such as vorticity and
integrated moisture flux divergence, in foreshadowing the exceptionally heavy rainfall.
Further, two events of heavy rainfall and flooding in the northeastern states of India during
July 2002 are also investigated.
The basic data for preparing the weather maps for diagnostic analysis and for running the
forecast model are processed from the ECMWF (ERA-40) Reanalysis data sets downloaded
from their ftp site. The data downloaded are those of the five basic flow variables:
geopotential ( gZ ), temperature ( T ), u & v components of wind and relative humidity (RH)
on 15 isobaric levels at 6 hourly map times. These are first deGRIBed using an ECMWF
GRIB decoding software package, also downloaded from their ftp site, and customised to suit
our requirements. The gridded fields are then interpolated to the limited area model grid and
transformed from pressure to the 15 model σ layers via a pressure to sigma converter. The
lateral boundary conditions for running the model are calculated from the 6 hourly analysed
fields as perfect boundary conditions.
The most common synoptic situation responsible for flood producing rainstorms in
Bangladesh is the positioning of the axis of seasonal monsoon trough in a northerly latitude
over north Bangladesh and persistence of the above conditions under the influence of large
scale circulation anomalies. The years 1987 and 1998, in particular, witnessed a situation in
which the persistence of monsoon trough in its northerly position was prominently reflected
even in the mean monthly maps of flow pattern. The analysis brought out that the above
12
anomaly of the synoptic situation occurred on a super-synoptic time scale and that it was an
interplay of global scale circulation anomalies responsible for creating quasi-stationarity of
the monsoon trough in a northerly position. A land depression may sometimes form over
Bangladesh, which may have its origin farther east over South China Sea. This is a
favourable situation for occurrence of heavy rainfall in south Bangladesh and the coastal belt
off Arakan coast. The study showed that when a depression forms in its usual position over
north Bay of Bengal with the territory of Bangladesh lying to its north and east, the rainfall
over the country is only light to moderate, which is in conformity with the typical structure of
a monsoon depression.
The limited area forecast model was run for short range forecasts of precipitation during
the selected days of heavy rainfall occurrence in Bangladesh during the above episodes of
1987 and 1998, as also a few cases of heavy rainfall and flooding in the northeastern region
of India. The entire region comprising the northeastern states of India and Bangladesh is
dotted with significant orographic features which have a dominating influence on the
occurrence of heavy rainfall in this part of the subcontinent. The forecasts of rainfall
produced by the model did bring out this influence prominently and were reasonably well
placed in regard to simulating the observed areas of rainfall although the predicted
magnitudes remained underestimated.
The forecast model was also run in two cases of monsoon disturbances forming in the
Bay of Bengal and moving across central and northern parts of India during August 1997 and
June 2002. These disturbances had some exceptional features. The depression of 20-27
August 1997 had a long history in that it travelled all the way upto northwest India and north
Pakistan and maintained its intensity as a deep depression throughout its movement until it
dissipated. The model was able to simulate its movement to a good degree of accuracy even
in 48 hour forecast. The rainfall forecasts were also reasonably well placed. The monsoon
low of June 2002, also went through a long history with a slow movement and intensified
after entering land. Even though the disturbance was classified as only a low pressure area
on the surface, its associated upper air cyclonic circulation was intense and deep, as much as
the deep depression of August 1997. The low caused exceptionally heavy rainfall for
consecutive three days in the Gujarat-Maharashtra belt, disrupting rail and road traffic due to
heavy flooding. While the model simulations of the movement of this disturbance were
accurate enough, the rainfall amounts were much underestimated compared to observations.
Forecasts of the rainfall magnitudes that were recorded in this episode are by no means
possible to foresee by synoptic analysis either. This points towards the need for developing
model output interpretation techniques for ‘event’ forecasting in such cases, based on
additional derived products like vorticity, moisture flux divergence etc., which could provide
useful numerical guidance in such cases. The computation of integrated moisture flux
divergence in the June 2002 case did indicate an unusually strong field, which was a
supportive evidence for the occurrence of exceptionally heavy rainfall. Development of such
methodology is targeted as our future work programme.
13
Monsoon forecasting with a limited area
numerical weather prediction system
1. Introduction
Monsoon forecasting is a challenging problem in the whole of south Asia where

monsoons constitute a major weather system affecting a large population. The monsoon
forecasting problem has several aspects. While long range forecasts on monthly and seasonal
time scales and for larger spatial extent are required for long term economic planning, short
and medium range forecasts are essential for various weather sensitive activities such as
farming operations, flood forecasting etc. The synoptic methods have been the mainstay of
tropical weather forecasting. Of late, numerical weather prediction methods have acquired
greater skills and are playing an increasingly important role in the tropical predictions, though
the progress of dynamical modelling effort in the tropics has been rather slow in comparison
with the extratropics. This is because of some inherent problems associated with the
dynamics of the tropical motion systems. The problems involved are much more difficult
than in the middle latitudes. In the extratropics the primary energy source for the
atmospheric motions is the zonal available potential energy associated with the strong
temperature gradients, and there exists a satisfactory dynamical theory of these motions
outside the tropics. In the tropics, on the other hand, the storage of available potential energy
is very small due to the very small temperature gradients. Latent heat release in cumulus
convection is the primary energy source. Parameterization of cumulus convection in tropical
models is therefore very important and is a difficult problem. Added to this, there is the
problem of large perennial data gaps in the tropical regions, which are largely oceanic. The
tropical numerical weather prediction system is required to address these problems
adequately. Much progress has been made in recent years in the development of numerical
models for the low latitudes. The World Weather Watch, now supported by a variety of
surface based and space based observing platforms, has considerably enhanced the
observational data base for numerical modelling. The availability of faster computers has
enabled a large volume of tests on analysis, initialization, sensitivity to physical
parameterization, and statistical evaluation of numerical weather prediction, resulting in an
overall improvement in the skill of tropical dynamical models.
In the recent past global models have rapidly evolved towards a substantial increase in the
horizontal resolution and emerged as the quasi-static type mesoscale models. Studies show
that the global models are now capable of handling the tropical weather disturbances quite
well. Krishnamurti and Oosterhof (1989) have demonstrated an example of the successful
prediction of the formation and motion of typhoon Hope of August 1979, which made a
landfall near Hong Kong. Our own experience shows that the numerical guidance products
that are routinely available on the Global Telecommunication System (GTS) from the
advanced global NWP centres like ECMWF and NCEP Washington do capture the formation
and movement of the tropical depressions and cyclones to a satisfactory level of reliability.
Global models have the obvious advantage of dispensing with the need for lateral boundary
conditions, which are invariably required for running a limited area model. While global
models can provide good forecasts of large scale weather systems, the regional models have
their own place in the field of NWP modelling. The regional models, through the nesting
approach, can be run on very fine horizontal resolutions of even 5-10 km and with a non-
hydrostatic dynamics, which can handle the high impact and disastrous weather events like
heavy rainstorms, severe thunderstorms, tornadoes, and tropical cyclones in their core
14
structure. Several such promising limited area models have emerged in recent years. The
most recent example is that of the widely popular MM5 model developed by NCAR/Penn
State University.
University (Krishnamurti et al., 1990) has been implemented in the SAARC Meteorological
Research Centre (SMRC), Dhaka for NWP research. The FSU model is a high resolution
multi-level primitive equation model based on a semi-Lagrangian dynamics, cast in sigma
coordinate system. The model is considerably advanced in the dynamical, physical and
initialization aspects and is specially suited for prediction of meso scale events and heavy
rainfall. Various versions of this model have been used in tropical prediction experiments by
Krishnamurti et al. (1979, 1987), Krishnamurti and Ramanathan (1982), Krishnamurti (1985)
and Krishnamurti et al. (1990). Results of these experiments have amply demonstrated the
capability of the FSU model in forecasting the tropical weather systems effectively. A
version of this model implemented in India Meteorological Department, New Delhi was used
for tropical cyclone track prediction experiments in conjunction with a bogusing scheme
(Prasad et al., 1997).
The above model has been used in the present work for several case studies of heavy
rainfall and movement of monsoon depressions during the summer monsoon season. The
cases chosen for experiments are in respect of some heavy rainfall episodes in 1987 and
1998, the two most prominent excess rainfall and flood years in Bangladesh, as also some
cases of monsoon depressions originating in the Bay of Bengal in regard to their exceptional
behaviour. The main focus of the study is to document the synoptic situations responsible for
causing heavy rainfall episodes and floods in Bangladesh. The efficacy of the limited area
model in forecasting heavy rainfall and movement of monsoon depressions is also
investigated. We describe in section 2, the basic framework of the FSU model. Section 3
provides some general information on the principal rain bearing systems – the seasonal
monsoon trough and the monsoon depressions – of the southwest monsoon in the Indian
subcontinent. A brief description of the data and methodology is given in section 4. Results
of case studies are presented in section 5. Section 6 gives concluding remarks.
2. The basic framework of the FSU limited area model@
DYNAMICAL FRAMEWORK
The forecast model used in this study is a multi level primitive equation model cast in the
σ (sigma) coordinate system. The dimensionless σ is defined as:
p
σ= (2.1)
ps
where p is the pressure and p s is the pressure at the Earth’s surface.
_________________________________________________________
@
Material for this section is adapted from Krishnamurti et al. (1990)
The basic dynamical framework of the model is constructed from the following basic set
of equations:
15
Equations of motion
Du ∂u u ∂z ∂ ln p s
= − σ& + v( f + tan φ ) − g − RT + Fx (2.2)
Dt ∂σ a ∂x ∂x
Dv ∂v u ∂z ∂ ln p s
= − σ& − u ( f + tan φ ) − g − RT + Fy (2.3)
Dt ∂σ a ∂y ∂y
Thermodynamic energy equation
DT ∂T ⎡ σ& D ln p s ⎤ H + D T
= − σ& + κT ⎢σ + ⎥+ (2.4)
Dt ∂σ ⎣ Dt ⎦ CP
Mass continuity equation
D ln p s ∂ σ& r
= − − ∇ .V
∂σ
(2.5)
Dt
Moisture continuity equation
Dq ∂q
= − σ& + E − r + Dq
∂σ
(2.6)
Dt
Equations (2.2)-(2.6) constitute the prognostic equations of the model. To close the above
system of equations, the following diagnostic equations are introduced:
Hydrostatic equation
∂z RT C pθ ∂ π κ
= − = −
∂σ gσ g ∂σ (2.7)
Equation of state
RT RT
α = − =
p σ Ps (2.8)
The quantity π κ in equation (2.7) is obtained from the Poisson equation:
Poisson equation
κ (2.9)
T κ ⎡ p ⎤
=π = ⎢
θ ⎣ 1000 ⎥⎦
16
The list of symbols is provided in Table 1.
The operator D / Dt in equations (2.2)-(2.6), which represents the horizontal Lagrangian

operator, can be expressed in the spherical coordinate system as:
D( ) ∂( ) u ∂( ) v ∂( )
= + + (2.10)
Dt ∂t a cos φ ∂λ a ∂φ
which can be written in abbreviated form as:
D( ) ∂( ) u∂( ) v∂( )
= + + (2.11)
Dt ∂t ∂x ∂y
A one-step semi-Lagrangian advection (Mathur, 1983) and a semi-implicit time-
differencing scheme is used for integration of the model. A special feature of this scheme is
that it accounts for time changes of dependent variables and the advecting velocities over the
trajectory traced by the parcel in time step ∆t . The inclusion of this formulation of the
advective processes is expected to provide a reasonable prediction of the phase and amplitude
of rapidly developing disturbances. From a physical point of view semi-Lagrangian
advection is a direct approach. To make this scheme economical, it is coupled to the semi-
implicit time integration scheme.
The finite differencing grid utilizes the Arakawa-C grid (Mesinger and Arakawa, 1976)
for the horizontal discretization. This grid is very well suited for the representation of
divergence and for the formulation of Helmholtz’s equation, which arises in the use of the
semi-implicit time integration scheme (McGregor and Leslie, 1977). The grid is staggered in
the horizontal and the vertical. Variables u , v , q , and T are defined at the centre of σ
layers and the vertical velocity σ& and z are defined at the layer interfaces. Fourth-order
finite differences are used in the formulation of most of the numerical representation of the
model. The physical processes include shallow and deep moist convection, large-scale
condensation, air-sea interaction, diffusion processes, planetary boundary layer, a detailed
calculation of radiative processes including cloud feedback processes, and a variable solar
zenith angle. The surface fluxes are based on similarity theory and are coupled to the
computation of surface hydrology and surface energy balance. Time-dependent boundary
conditions and a dynamic normal mode initialization are the other features of the model.
Table 1
List of symbols
a Earth’s radius
17
α specific volume
Cp specific heat (for dry air) at constant pressure
Dq convergence of eddy flux of moisture
DT convergence of eddy flux of sensible heat
E moisture sources or sinks
f (=2 Ω sin φ ) Coriolis parameter
Fx convergence of eddy flux of x-momentum
Fy convergence of eddy flux of y-momentum
g acceleration due to gravity
H diabatic heating rate per unit mass
κ = R/Cp
λ longitude, positive eastward from Greenwich meridian
Ω angular velocity of Earth (s-1)
p atmospheric pressure
ps atmospheric pressure at Earth’s surface
φ latitude, positive northward from equator
κ
⎡ p ⎤
π = ⎢ ⎥
⎣1000 ⎦
q specific humidity
r rainfall
R specific gas constant (for dry air)
σ dimensionless vertical coordinate
σ& (= dσ / dt ) vertical velocity in sigma coordinate
t time
T temperature
θ potential temperature
u zonal component of wind vector
v meridional component of wind vector
r
V horizontal wind vector
x x -coordinate
y y -coordinate
z height
LATERAL BOUNDARY CONDITIONS
The limited area models used for forecasting in limited space domains essentially need
specification of lateral boundary conditions. Numerical treatment of lateral boundary is a
difficult but important aspect of limited area modelling as the flow in the interior of the
domain gets influenced by changes outside the domain. This influence is important even in
18
short range forecasting for a couple of days. In the present model we make use of the time
dependent boundary conditions. Here the prognostic variables on the boundary are allowed
to change with time. The future values of the prognostic variables are usually obtained from
a concurrently running global model forecast. In research mode, however, these can be
obtained from future observations (analysis) – the so called perfect boundary conditions.
These specified values are then merged with the forecast produced by the model. If at any
time t the model forecast is x f and the externally specified value is x s , where x is any
prognostic variable, then the following expression is used to give the merged forecast:
xc = (1- α ) x f + α x s
Where α is constant and equal to 1.0 at the boundary. The value of α decreases as we
move away from the boundary and becomes zero in the interior from the sixth boundary point
inward. The boundary updating is done at every time step of model integration. The
procedure is carried out for u , v, z , ln p s , and q on the σ surfaces.
PHYSICAL PROCESSES
The physical processes that are incorporated in the model are:
• Large scale condensation

• Parameterization of shallow moist convection
• Parameterization of deep cumulus convection
• Surface fluxes via similarity theory
• Vertical diffusion
• Short-wave radiative processes
• Long-wave radiative processes
• Surface energy balance
• Orography
Except for the long-wave and short-wave radiation algorithms, each of the above physical
processes are computed at each time step of the model forecast. The radiative processes are
computed every six hours, being highly computationally expensive.
The full details of the dynamical and physical aspects of the model may be seen in
Krishnamurti et al. (1990).
σ 1 = 0.050 P1 , σ& 1 , z1
u1 , v1 , T1 , q1 ------------------------------------------------ σ~ 1
σ 2 = 0.125 P2 , σ& 2 , z 2
u 2 , v 2 , T2 , q 2 ------------------------------------------------ σ~ 2
19
σ 3 = 0.175 P3 , σ& 3 , z 3
u 3 , v3 , T3 , q3 ------------------------------------------------ σ~ 3
σ 4 = 0.225 P4 , σ& 4 , z 4
u 4 , v 4 , T4 , q 4 ------------------------------------------------ σ~4
σ 5 = 0.275 P5 , σ& 5 , z 5
u 5 , v5 , T5 , q5 ------------------------------------------------ σ~ 5
σ 6 = 0.350 P6 , σ& 6 , z 6
u 6 , v6 , T6 , q 6 ------------------------------------------------ σ~ 6
σ 7 = 0.450 P7 , σ& 7 , z 7
u 7 , v7 , T7 , q 7 ------------------------------------------------ σ~ 7
σ 8 = 0.550 P8 , σ& 8 , z 8
u8 , v8 , T8 , q8 ------------------------------------------------ σ~ 8
σ 9 = 0.650 P9 , σ& 9 , z 9
u 9 , v9 , T9 , q9 ------------------------------------------------ σ~ 9
σ 10 = 0.750 P10 , σ& 10 , z10

u10 , v10 , T10 , q10 ------------------------------------------------ σ~ 10
σ 11 = 0.800 P11 , σ& 11 , z11

u11 , v11 , T11 , q11 ------------------------------------------------ σ~ 11
σ 12 = 0.850 P12 , σ& 12 , z12
u12 , v12 , T12 , q12 ------------------------------------------------ σ~ 12
σ 13 = 0.900 P13 , σ& 13 , z13
u13 , v13 , T13 , q13 ------------------------------------------------ σ~ 13
σ 14 = 0.950 P14 , σ& 14 , z14

u14 , v14 , T14 , q14 ------------------------------------------------ σ~ 14
σ 15 = 0.990 P15 , σ& 15 , z15

u15 , v15 , T15 , q15 ------------------------------------------------ σ~ 15
σ 16 = 1.0 P16 , σ& 16 , z16
Fig. 1. Vertical structure of the 15-layer version of the limited area model; solid lines are located at
the layer interfaces and dashed lines at the layer means. Variables u , v, T , and q are staggered
with respect to the variables P , σ& , and z .
v v v v
u z u z z u z u
20
v -------- v --------------------------------- v ------ v
u z u z z u z u
u z u z z u z u
∆φ v ---- ---- v ----------------------------------v ------ v
u z u z z u z u
v v v v
∆λ
Fig. 2. Horizontal structure of the limited area model. Variables are staggered following
Arakawa C-grid. Variables q , ln p s , σ& , T , and P (not shown) are also located at the
z points.
The SMRC version of the model is run on a horizontal grid resolution of 0.5o x 0.5o
lat./long. and 16 sigma levels (15 layers) in the vertical on a SUNFIRE V480 system. The
vertical extent of the model ranges from σ = 0.05 to σ = 1 at the Earth’s surface. The sigma
values in the present version from top to bottom are: 0.05, 0.125, 0.175, 0.225, 0.275, 0.350,
0.450, 0.550, 0.650, 0.750, 0.800, 0.850, 0.900, 0.950, 0.990, 1.0. The model is integrated in
a domain 30E to 120E and 15S to 45N. The horizontal and vertical arrangement of the model
variables is shown in Figures 1 & 2 respectively. The orography used in the model is the one
obtained by smoothening a 30 minute terrain field. Figure 3 illustrates the orography for a
section of the domain.
21
Fig. 3. Orography used in the limited area model
3. Monsoon forecasting – the principal rain bearing weather systems

3.1 The seasonal monsoon trough
The summer monsoon rainfall in the Indian subcontinent is known to be characterised by
active and weak (or break) spells on a time scale of about two weeks. These active/break
spells are the result of fluctuations of the seasonal monsoon trough. Monsoon trough is an
extended trough of low pressure which runs across Gangetic plains of north India with its
western end anchored to the seasonal heat low over northwest India and Pakistan and eastern
end emerging into the head Bay of Bengal. In the mean the axis of this trough runs from
Ganga Nagar in Rajasthan to Kolkata via Allahabad. A map showing the distribution of
mean sea level pressure (long term mean) for the month of July is given in Fig. 4 for the
purpose of illustration. The axis of the monsoon trough slopes southward with height and its
position at 3 km above mean sea level runs along nearly 20o N latitude over the Indian
longitudes. Paradoxically, the zone along the monsoon trough axis on sea level has a minima
of annual and seasonal monsoon rainfall. The maxima of rainfall lies to the south of its axis.
22
Fig. 4. Mean sea level pressure (long term mean) in the month of July (based on
NCEP/NCAR Reanalysis)
The monsoon trough is subjected to meanderings north and south and this helps in
distributing rainfall over the whole of Gangetic plains. These northward and southward
movements of the monsoon trough axis are generally indicative of the large scale active and
weak monsoon conditions. The position of monsoon trough axis is an important factor in the
monsoon activity over the subcontinent. When the trough stays in its normal position or
south of its normal position, this is considered to be an active monsoon situation. The rainfall
is well distributed over the northern plains and central parts of India. When the trough dips
into the Bay, conditions become favourable for the formation of monsoon depressions/lows
in the Bay of Bengal, which eventually move westward across the mainland and produce
good rainfall activity. On the other hand, sometimes the axis of the monsoon trough shifts
northward to the foothills of Himalayas. The northward shift may take place either in its
entire stretch from west to east or only in one part. When the trough in its whole stretch
shifts northward, the rainfall over the central and the northern plains of the country is
drastically reduced and increases along the foothills of Himalayas. This situation has come to
be known as a ‘Break’ monsoon condition in Indian meteorological parlance, so called
because of large scale reduction of rainfall over major part of the country. Under certain
favourable conditions of the atmospheric flow, a heavy incursion of moisture from the
adjoining seas due to strong southwesterly/southerly flow takes place. Rainfall in the
foothills of Himalayas is substantially enhanced leading to floods in the Himalayan rivers.
An eastward movement of a wave disturbance in the middle latitude westerlies across
Himalayas often leads to excessive rainfall in some pockets producing heavy floods. The
23
break monsoon conditions usually last for a few days. Prolonged breaks sometimes occur
leading to dry conditions over large parts of India.
The northward movement of the monsoon trough axis is generally seen to occur in the
wake of a monsoon depression having travelled across the country. The active period of
monsoon depression is mostly followed by a weak phase when the monsoon flow weakens,
the low level easterlies over the Gangetic plains north of the monsoon trough axis are
replaced by westerlies with the shifting of the trough close to the foothills. The monsoon
trough often gets completely effaced in the mean sea level pressure and low level wind fields.
The trough again shifts southward to its normal position with the formation of a fresh
disturbance in the Bay of Bengal, often in quick succession.
It has been hypothesized that the fluctuations in the position of the monsoon trough and
hence the active/break spells in monsoon rainfall are identified with the fluctuations of the
regional tropical convergence zone (TCZ) or the local Hadley circulation (Yasunari, 1979,
1980, 1981; Sikka and Gadgil, 1980; Gadgil, 1988; Gadgil and Asha, 1992), which lies over
the heated continent of the Afro-Asian land mass. Sikka and Gadgil (1980), through analysis
of satellite imagery, showed that the warm equatorial Indian Ocean is also a favourable
location for the occurrence of TCZ. Thus there seem to be two favourable zones of the TCZ,
one over the continent and the other over the equatorial Indian Ocean. Active spells of
rainfall are associated with an intense tropical convergence zone of the continental origin.
Sikka and Gadgil (1980) found that active spells of the continental TCZ occur with weak
spells of the oceanic TCZ and vice versa. This observation was also supported by Gadgil and
Asha (1992), Yasunari (1979) and Lau and Chan (1986), who found an inverse relationship
between daily OLR over the continent and the equatorial Indian Ocean TCZs. The major
difference between excess/normal and deficient monsoon years is in the duration and
intensity of the breaks or dry spells, suggesting that the inter-annual variation between good
and poor monsoons is a manifestation of the variation of the duration and intensity of active
TCZ over the continent. The fluctuations of the TCZ over the monsoonal area is due to insitu
generation and demise of the continental TCZ as well as northward propagation of the
oceanic TCZ from the equatorial Indian Ocean (Sikka and Gadgil 1980; Yasunari, 1979,
1980, 1981). The maximum cloudiness (or the heat source) also fluctuates between the two
preferred positions, though averaged over the whole season, it appears to be located over
continental India. It turns out that the continental heat source dominates over the oceanic
heat source for a longer period of time within the season. During an active spell the
continental heat source dominates while during a break spell the oceanic one dominates
(Goswami, 1993). Thus the difference between an active monsoon and a weak monsoon is
determined by how much more time the system spends in one regime compared to the other.
The cumulative monsoon performance in any year thus depends upon the behaviour of intra-
seasonal oscillations. These oscillations, being super-synoptic in nature, appear even in the
monthly/seasonal means.
Thus the principal rain bearing weather systems in the context of day to day monsoon
rainfall in the Indian subcontinent are the monsoon depressions/lows forming in the Bay of
Bengal during the active monsoon conditions and the seasonal monsoon trough located in a
northerly position close to the foothills of Himalayas (the so called ‘break’ monsoon
conditions). Monsoon depressions have some characteristic features which are summarised
in the following paragraphs.
3.2 Monsoon depressions
24
DEFINITION AND STRUCTURE
As the TCZ moves to sufficiently high latitudes and the monsoon establishes over central
parts of the country and the Bay of Bengal, conditions become favourable for the formation
of low pressure areas and depressions in the north Bay of Bengal. These are known as
monsoon depressions. Monsoon depressions are the principal rain bearing systems in the
monsoon season. On a synoptic surface weather map these low pressure systems appear as
closed isobars. On an upper air map they appear as closed circulations with winds rotating
around the centre cyclonically (anticlockwise) in the northern hemisphere. In the upper air
the circulation generally extends up to the middle troposphere (about 6-7 km), occasionally
even higher (about 9 km) if the system is deep. The upper air circulation usually slopes
southward with height due to the presence of colder monsoon air to the south. The more
intense a system is, more will be the number of closed isobars, as the atmospheric pressure in
the central region will drop considerably. In India Meteorological Department a low pressure
system is classified as a low if the sustained winds in its circulation are less than 17 knots
(nautical miles per hour), and a depression if the wind speed lies between 17 and 27 knots.
The systems with wind speeds exceeding 27 knots and upto 33 knots are called deep
depressions. Above this threshold the systems fall in various categories of cyclonic storm.
Formation of cyclonic storms in the monsoon season is rare due to unfavourable
environmental conditions.
FORMATION AND MOVEMENT
The monsoon depressions/lows, after formation and maturing, cross the coast and move
generally in a west to northwesterly direction across central parts of the country. On an
average two to three depressions develop per month in the Bay of Bengal at the height of the
monsoon season. The frequency of formation is, however, highly variable from year to year
and depends on the mode of the large scale monsoon environment. There may be no
formations in the sluggish monsoon years while active monsoon years may witness more than
average formations. Some of these systems may not develop into depressions and move only
as monsoon lows. The monsoon depressions (for the purpose of this write up we shall refer to
all the three categories of monsoon low pressure systems as monsoon depressions), as they
move inland, produce heavy to very heavy rainfall in the areas which come under their
influence. This is one of the major sources of floods in the central and northern plains of
India and the river catchments originating in the central highlands. The location, intensity,
and the horizontal extent of area covered by the flood producing rainstorms associated with
these depressions is something that is determined by the behaviour of these depressions such
as their point of formation in the Bay of Bengal, crossing of coast, and the subsequent path of
movement and the speed of movement. Mostly, the monsoon depressions form in the head
Bay of Bengal north of latitude 18o north, cross Orissa-West Bengal coast, and move across
central India in a direction between west and northwest, guided by the upper easterly flow
over India during this season. They usually cover about 400 to 500 km per day on an average.
The average life period of a monsoon depression in general is about three to four days. They
usually weaken after crossing coast and tend to die out after reaching Madhya Pradesh and
merge with the seasonal monsoon trough. Some of these depressions, however, have longer
life span. Some of them continue to travel in a westerly direction, all the way up to the
western fringe of the country, even finally emerging into the Arabian Sea. As they near the
western India shores, they often pick up strength, fed by enhanced moisture supply from the
Arabian Sea. Some, which have a little northerly component travel up to the seat of seasonal
heat low over northwest India and merge with it. The tracks of monsoon depressions in the
mid monsoon months of July and August, being generally in a west to northwesterly
25
direction, have a tendency to be concentrated in a fairly narrow belt running across central
India, confined south of 25o north latitude. However, some of the disturbances may deviate
from the normal path and their movement may be slowed down or stalled in exceptional
circumstances, particularly if they happen to cross 25o north latitude.
During the late monsoon period of September and October, as the monsoon circulation
starts retreating southward, the depressions start forming in a comparatively lower latitude in
the Bay of Bengal, and their tracks are more spread out, moving in any direction from west to
north and later northeast in higher latitudes. Here, though October is outside the calendar
definition of southwest monsoon period, the actual monsoon rainfall sometimes extends well
into October, particularly in the eastern and peninsular India and therefore it is considered as
continuation of the southwest monsoon in these parts. The middle latitude westerly belt,
which is confined to higher latitudes during the height of monsoon period, starts retreating
southward and the monsoon depressions have a chance to come under the influence of wave
disturbances moving eastward in this belt. In such cases the depressions have a tendency to
change course and show an erratic behaviour. The northward tendency of movement and the
erratic behaviour of monsoon depressions is though a feature of the late monsoon period,
such a behaviour in the mid monsoon months of July and August is not ruled out in
exceptional cases, influenced by some unusual changes in the atmospheric flow pattern in the
neighbouring middle latitude belt, which, in turn, affect the steering flow over the field of
monsoon depression.
Though the usual seat of formation of monsoon depressions is the head Bay of Bengal,
some of them may have an origin farther east in the South China Sea and adjoining west
Pacific ocean, which has a significant typhoon activity during the southwest monsoon period.
Some of these typhoons or tropical depressions cross the Indochina peninsula, move
westward and emerge into the Bay of Bengal and neighbouring areas of Bangladesh. Though
they usually weaken during the course of their movement over land and generally lose
identity on a synoptic weather map, they travel westward as a low pressure wave and get
rejuvenated once they approach the Bay of Bengal, aided by a fresh supply of moisture. They
may redevelop into a depression.
ASSOCIATED RAINFALL PATTERN
The distribution of rainfall associated with a monsoon depression is rather unique in each
case, depending on its intensity, direction & speed of movement and the prevailing large
scale environmental conditions with which it interacts. Nevertheless there are some common
features. In the case of a normal westward moving depression in the mid monsoon months of
July and August, maximum rainfall is concentrated in a belt ahead of its current position to a
distance of about 500 km and to the left of the track, viz., in the southwest sector of the
depression field, in about 400 to 500 km wide belt (Rao, 1976). However, this relationship
does not always hold good, particularly in the late monsoon season, i.e., September and
October. A typicality of the late monsoon depressions is that the rainfall associated with
them is often concentrated in a small area and is much more intense than their mid monsoon
counterparts. The late monsoon depressions have a special significance in view of their high
flood producing potential due to a combination of concentrated heavy rainfall associated with
them and a high antecedent precipitation index.
WATER POTENTIAL OF MONSOON DEPRESSIONS
26
The monsoon depressions produce heavy to very heavy rainfall in the areas through
which they pass. The intensity of rainfall being heavy or very heavy is conventionally
measured in terms of the cumulative rainfall in 24 hours at an observing station. As per the
convention in Indian meteorology, the magnitude range of 7 to 12 cm in a 24 hour period is
classified as heavy rainfall. 13 cm and above is known as very heavy rainfall. This
definition precludes those cases where very intense short duration precipitation may occur
leading to inundation, but does not qualify to be called a heavy rainfall day because of not
satisfying the cumulative rainfall criterion. Rainfall amounts of 20 to 30 cm in 24 hour are
not uncommon with monsoon depressions. Exceptionally heavy rainfall of as much as 60 to
80 cm per day may also occur in some intense situations. The highest 24 hour point rainfall
on record at a plains station in India is 99 cm on 2nd July 1941 at Dharampur in south Gujarat
(Dhar et. al, 1984; Kulshrestha, 2000). This event was associated with a monsoon
depression. This has been found to be the most severe rainstorm on record in India. The
storm led to severe floods in the rivers of the region resulting in very heavy loss of life,
livestock and property especially in the districts of Surat, Baroda and Valsad in south Gujarat.
The storm also caused heavy landslides in the Western Ghats of the adjoining north Konkan
region leading to complete disruption of rail and road traffic between Mumbai and Pune for
nearly ten days (Kulshrestha, 2000).
The belt of heavy rainfall keeps on shifting as the disturbance moves and its adverse
impact in any particular area is thus contained. However, if a situation arises where the
depression’s movement is retarded or stalled due to some unusual developments in the
atmospheric flow pattern, heavy rainfall may persist in a particular area longer than usual,
which becomes a potential source of floods. Sometimes it may so happen that the
depressions form in the Bay of Bengal and pass through a particular area in quick succession
thus giving effect to persistence in heavy rainfall leading to floods.
IMPACT
The depressions of July and August, which cross the coasts of Orissa and West Bengal
and generally move in a westerly to northwesterly direction across central parts of the country
affect the river catchments of central highlands. In such cases the areas likely to be most
affected are Orissa, Gangetic West Bengal, Jharkhand, Madhya Pradesh, Vidarbha, parts of
Maharashtra, Gujarat, Saurashtra and Rajasthan depending on the life span of the particular
disturbance and the length of its track. The systems which form and move in a slightly higher
latitude may also affect the catchments of the northward flowing tributaries of Ganga and
Yamuna, such as Chambal, Sone etc. which have their source regions in Madhya Pradesh
hills. In some cases the depressions have a tendency to form in the Bay of Bengal in slightly
lower than normal latitudes off Andhra coast even during the mid monsoon months of July
and August. These systems cause heavy rainfall and flooding in Andhra Pradesh. The
depressions which have a longer life span and happen to move upto Gujarat and Maharashtra,
produce extremely heavy rainfall in these states due to strengthening of moisture inflow from
the Arabian Sea. The rivers having their catchments in West Madhya Pradesh and Western
Ghats go in spate. The depressions of late monsoon period, as mentioned earlier, have a
tendency to move in a more northerly direction and affect the Gangetic plains and the river
catchments in Himalayas. These northward moving depressions of late monsoon period often
come under the influence of a middle latitude wave disturbance extending southward and
moving eastward across northern parts of the country in the middle and upper troposphere.
This type of synoptic situation is a potentially hazardous condition for producing
exceptionally heavy rainfall and severe widespread floods in the rivers of Himalayan origin.
27
4. Data and methodology
The basic data sets required for preparing the synoptic maps and the initial and lateral
boundary conditions for running the limited area model in the present study are drawn from
the ECMWF Reanalysis (ERA-40) data downloaded from their ftp site. The ERA data sets in
GRIB coded form are available on a coarse resolution of 2.5o x 2.5o lat./long. grid on constant
pressure surfaces. The data downloaded are those of the five basic flow variables:
geopotential ( gZ ), temperature (T ), u & v components of wind and relative humidity (RH)
on 15 isobaric levels: 1000, 925, 850, 775, 700, 600, 500, 400, 300, 250, 200, 150, 100, 70,
50 hPa at 6 hourly map times 00, 06, 12 and 18 UTC. These are first deGRIBed using an
ECMWF GRIB decoding software package, also downloaded from their ftp site, and
customised to satisfy our requirements. The gridded fields are then interpolated to the limited
area model grid and transformed from pressure to the 16 model σ levels (as outlined in
section 2 above) via a pressure to sigma converter. The lateral boundary conditions in this
study are calculated from the 6 hourly analysed fields, the so called perfect boundary
conditions. The input fields for the model are the five basic flow variables Z , T , u , v , RH
and the fixed fields of orography (terrain), the sea surface temperature (SST) and the albedo.
The highest point in the orography field carries a value of about 5400 meters in the
Himalayan region. The SST are the monthly climatology and albedo are available for June
and December, which are appropriately selected in the model run. The outputs are the same
basic fields on sigma surfaces and the vertical motion and accumulated precipitation. The
sigma fields are post processed on to the standard isobaric levels and graphically displayed
through the GrADS software package.
The observed daily rainfall data of about 35 rainfall stations of Bangladesh (Fig. 5) were
obtained from the Bangladesh Meteorological Department (BMD). The data were subjected
to a visual inspection and message, before being used in the computations, to deal with the
erroneous and missing entries, and each station’s file was reconstructed. For the purpose of
this study we have constructed the country average rainfall on day to day basis. The daily
country average rainfall was used as the basic element for constructing the long period
averages of the monthly & seasonal rainfall. The above method of computing the cumulative
monthly and seasonal country average rainfall using the daily country average rainfall as the
basic element was considered superior as it obviates the need for artificial interpolations to
fill the missing rainfall entries when individual station data are taken as the working
elements. Such missing entries were substantially large in number in some cases and the
interpolation procedures using graphical methods and/or substitution by long period averages,
which are normally resorted to, have their own limitations. The case studies undertaken also
required the daily rainfall data of Indian synoptic stations, which were picked up from the
published Indian Daily Weather Reports.
28
28
BHUTAN
NEPAL
27
26 INDIA
DinajpurRangpur
25 Bogra Sylhet
Mymensing
Rajshahi Srimangal
Ishurdi
BANGLADESH
24 INDIA Dhaka
Faridpur
Comilla
Jessore
23 M. court
SatkhiraKhulna Barisal
Bhola
Sandwip Rangamati
Hatiya
Chittagong
Khepupara
22
Cox`s Bazar
21 Teknaf MYANMAR
BAY OF BENGAL
20
19
86 87 88 89 90 91 92 93 94
Fig. 5. Map of Bangladesh showing location of synoptic stations
29
Fig. 6. Map of India showing meteorological subdivisions
5. Case studies
5.1 Analysis and forecasts of heavy rainfall events in Bangladesh in outstanding excess
rainfall and flood years 1987 and 1998
30
Bangladesh, known as the country of rivers, is frequently visited by floods, often
devastating, during the summer monsoon period. During recent times the country
experienced severe floods in 1987, 1988 and 1998 (SMRC, 2003; Quadir et al., 2003). Based
on an analysis of the discharge data by Mirza et al. (2001), Quadir et al. (2003) reported that
the discharge levels in the Ganga, Brahmaputra and Meghna river systems attained their
highest ever peaks in these years at some locations. An analysis of the seasonal rainfall
departure from the long period average over Bangladesh during the monsoon period revealed
that the floods in 1987 and 1998 occurred due to excess rainfall within Bangladesh as also in
the upstream portion of the river basins. Fig. 7 shows the year to year time series of the
percentage departures of the country average monsoon rainfall of Bangladesh during the two
main monsoon months of July & August with reference to the long period average (1971-
2000), where the years 1987 and 1998 stand out clearly. The year 1988, on the other hand,
did not experience any excess rainfall within Bangladesh (Fig. 7) and the floods occurred
mainly due to the contributions to river discharge from upstream. 1988 was an excess
monsoon year in India (Fig. 8) and was characterised by highly excess rainfall in some of the
meteorological subdivisions of northeast India such as Arunachal Pradesh, Assam &
Meghalaya, and Sub-Himalayan West Bengal & Sikkim that constitute a part of the mighty
Brahmaputra and Teesta river basins (SMRC, 2003), which eventually merge with the river
systems of Bangladesh. A map of India showing the meteorological subdivisions is included
in Fig. 6. Interestingly, 1987, which was a significant El Nino year, was a major drought
year in India (Fig. 8). It turned out to be the most significant excess monsoon and a major
flood year in Bangladesh. While the rainfall in Bangladesh by itself was highly in excess
during 1987, the adjoining meteorological subdivisions of northeast India, viz. Arunachal
Pradesh, Assam & Meghalaya, Sub-Himalayan West Bengal & Sikkim and even Bihar
Plains, which all contribute to the river discharge entering Bangladesh, also recorded large
excess (SMRC, 2003), in the backdrop of the large scale drought in other parts of India. 1998
was overall a good and normal monsoon in India. On the other hand, the year 2003, which
had a bountiful rainfall in India, had one of the most deficient rainfall in July & August in
Bangladesh (Fig. 7). This prompted us to look into the anomalies of the atmospheric
circulation in the region vis-a-vis the large anomaly in the rainfall distribution in the two
neighbouring regions. The objective of this study is to identify the typical synoptic situations
leading to flood producing heavy rainfall in Bangladesh with particular reference to the
monsoons of 1987 and 1998. We have chosen the episodes of heavy rainstorms in the above
two prominent flood years for a diagnostic study. The focus of this study is twofold – (i) to
identify the typical synoptic situations causing heavy rainstorms in Bangladesh; and (ii) to
examine the efficacy of the forecast model in predicting these events in short range time
scales. With this end in view we constructed the weather maps on selected days of heavy
rainstorms by post processing the high resolution sigma fields to pressure levels and
computing some derived parameters for a diagnostic analysis. The forecast model was run up
to 48 hours for rainfall prediction.
31
60
Jul+Aug
50
Warm episode
40 Cold episode
30
20
srf_dep
10
0
-10
-20
-30
-40
-50
1948
1950
1952
1954
1956
1958
1960
1962
1964
1966
1968
1970
1972
1974
1976
1978
1980
1982
1984
1986
1988
1990
1992
1994
1996
1998
2000
2002
Year
Fig. 7. Monsoon rainfall in Bangladesh during July & August - percentage departures of country average
rainfall from long period average (1971-2000); warm and cold episodes refer to the El Nino and La Nina
events
30
All India Monsoon Rainfall (1901-2003)

Warm Episode
Cold Episode
20
10
% departure
-10
-20
-30
1901
1903
1905
1907
1909
1911
1913
1915
1917
1919
1921
1923
1925
1927
1929
1931
1933
1935
1937
1939
1941
1943
1945
1947
1949
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. 8. All India monsoon rainfall (July-September) - percentage departures from normal; warm and cold episodes
refer to the El Nino and La Nina events (source: India Meteorological Department)
32
We identified the dates of heavy rainstorms by computing the daily country average
rainfall of Bangladesh. Figures 9-11 show the daily country average rainfall of Bangladesh
during the period 1 June to 30 September for the years 1987, 1988 and 1998. The monsoons
of 1987 and 1998 were characterised by several episodes of heavy rainstorms resulting in
highly excess rainfall over Bangladesh, as may be seen in Figures 9 and 11 respectively. The
day to day rainfall distribution in both the years is marked by occurrence of several peaks in
the bar diagram. A high peak in the country average is a reflection of widespread heavy
rainfall with many stations reporting high values of 24 hour precipitation, at times exceeding
20 to 30 cm in a continuous spell of several days. The most significant peaks in 1987
occurred in the last week of July, continuing into the beginning of August, and in the last
week of August, the latter recording a country average rainfall exceeding 90 mm. The one in
the last week of July exceeded 80 mm. The highest peaks in 1998 were comparatively
moderate as compared to 1987. The most significant peaks in 1998 occurred in the second
week of July and the second week of August. During 1988 on the other hand, the month of
June appeared to be quite active but the rainfall in July and August was considerably subdued
(Fig. 10).
The synoptic situations in respect of the above four rain spells are discussed in the
following paragraphs.
Daily monsoon rainfall (country average) - 1987
100
90
80
70
60
Rainfall (mm)
50
40
30
20
10
0
1 11 21 31 41 51 61 71 81 91 101 111 121
Day (1 June - 30 September)
September) 1987
33
100
90
80
70
60
Rainfall (mm)
50
40
30
20
10
0
1 11 21 31 41 51 61 71 81 91 101 111 121
September) 1988
100
90
80
70
60
Rainfall (mm)
50
40
30
20
10
0
1 11 21 31 41 51 61 71 81 91 101 111 121
September) 1998
34
5.1.1 Monsoon trough over north Bangladesh during last week of July to beginning of
August 1987
During the third week of July 1987, a low pressure area appeared over Bangladesh on 17
July (India Met. Dept., 1987). The low shifted slightly southward to north Bay of Bengal on
18 July. It then moved inland and was located over north Orissa and adjoining Bihar Plateau
and Gangetic West Bengal on 19th. The low further moved northwest and lay over Bihar,
adjoining northeast Madhya Pradesh and southeast Uttar Pradesh on 20th. It persisted over
Bihar on the subsequent three days. Consequent upon movement of the low pressure area
from north Bay to Bihar the axis of seasonal monsoon trough got aligned in a northerly
position passing across Bangladesh during the last week of July. This is demonstrated in Fig.
12, which shows the mean sea level pressure analysis (mslp) on 24 July 1987. A sharp
monsoon trough is seen running eastward across northern parts of Bangladesh and thence
eastward across northeastern states of India. A north south trough runs along the east coast of
India. Fig. 13 shows the streamline flow at 850 hPa with orography in the background. A
prominent reflection of the sea level trough along the east coast is seen at the 850 hPa level.
The flow to the east of this trough has a strong northward meridional component causing a
large moisture incursion in the Assam and neighbouring areas from Bay of Bengal and
striking the hill features in the eastern Himalayas, almost at right angle, causing strong
orographic ascent of moist air. There exists an area of strong accumulation of moisture as
seen in the pattern of integrated horizontal moisture flux divergence in Assam and
neighbourhood (Fig. 14). The integrated horizontal moisture flux divergence was computed
with the following expression (Prasad, 1992):
300 hPa
∫ ∇ ⋅ qV
Moistflux = 1/g
1000 hPa
where,
g = acceleration due to gravity
q = specific humidity
V = horizontal wind vector
Figures 15 & 16 contain the streamline flow at 500 hPa and 300 hPa respectively. The
prominent feature on these two levels is the southward displacement of the subtropical ridge,
which in its normal position runs along about 28-30 deg. latitude in the Indian longitudes. At
500 hPa a centre of anticyclonic circulation over central India lies near Nagpur, embedded in
the subtropical ridge. The centre of anticyclonic circulation at 300 hPa lies over Gangetic
West Bengal close to Kolkata. The 500 hPa map contains another important feature in the
form of a middle latitude westerly wave apparently moving across western Himalayas with a
cyclonic circulation located close to Himachal Pradesh in north India. The totality of
synoptic situation with the westerlies prevailing in the lower troposphere (Fig. 13), shifting of
the middle and upper tropospheric subtropical ridge southward and movement of a middle
latitude westerly wave across extreme northern India resembled that of a typical ‘break’
monsoon condition in the Indian monsoon system. The synoptic setting was thus very
favourable for heavy precipitation in the northeastern region of India and neighbourhood.
The heavy rain spell in Bangladesh during the above period was thus found to be associated
with the monsoon trough located in a northerly position, close to the foothills of eastern
Himalayas and the upper air flow pattern typical of a ‘break’ monsoon situation.
35
Fig. 13. Flow at 850 hPa on 24 July 1987 with orography (x100 M) in the
background
36
Fig. 14. Integrated horizontal moisture flux divergence (x10-5g cm-2s-1) on 24 July
1987
37
The Forecast model was run based on the initial conditions of 24 July with lateral
boundary conditions calculated from analyses at 6 hourly intervals. The 24 hour and 48 hour
predicted rainfall valid for 25 July and cumulative for 25 & 26 July 1987 are presented in
Figures 17 & 18 respectively. The model produced heavy rainfall in the areas covering
Assam & adjacent states and Bangladesh, consistent with the synoptic patterns and the areas
of heavy moisture convergence. Fig. 19 & 20 show the observed rainfall on 25 July and 26
July as verification. Widespread heavy rainfall along the coastal areas of Bangladesh,
influenced by orography, as well as northern and central parts was the prominent feature.
The monsoon trough stayed north for a considerably long period of about 10 days through
the beginning of August. As a corroboration of this statement, the mean sea level pressure
analysis on 31 July is presented in Fig. 21, which shows the monsoon trough persisting in the
same location across north Bangladesh. The observed rainfall distribution on 1 August (the
day of peak in the country average rainfall in 1987) is shown in Fig. 22. Sylhet, which lies at
the foothills of Khasi-Jaintia hills near Cherrapunji, recorded 30 cm of rainfall on this day.
Many stations in northern and central parts of Bangladesh reported heavy rainfall exceeding
10 cm. The 24 hour predicted rainfall valid for 1 August 1987, based on the initial conditions
of 31 July 1987 is shown in Fig. 23. The model simulated the observed precipitation
reasonably well.
38
39
40
Fig. 22. 24 hour precipitation (cm) as observed on 1 August 1987
41
Fig. 23. 24 hour rainfall forecast (mm) valid 00 UTC of 1 August 1987
5.1.2 Land depression, 26-30 August 1987
The second significant rain spell occurred in the last week of August 1987 from 26 to 28
August. A land depression appeared over Bangladesh on 26 August (India Met. Dept., 1987).
Fig. 24 shows the mean sea level pressure analysis on 26 August, 00 UTC. It may be
pertinent to state here that the maps presented in these analyses are constructed from the post
processed fields after interpolation from a coarse resolution of 2.5x2.5 deg. lat./long. to
0.5x0.5 deg. lat./long. In this process the geometry and shape of the isobaric fields is likely
to get distorted. This is reflected in somewhat elongated shape of the depression and its
centre a little displaced from the synoptic centre. Nevertheless the synoptic patterns have
been well reproduced from the reanalysis data sets via the post processing procedures
adopted in this study. In an attempt to trace the origin of this land depression, we constructed
the analysed daily synoptic maps for the preceding few days – upto 20th of August.
The origin of this disturbance could be traced to South China Sea. A tropical depression,
probably a typhoon, existed in the South China Sea on 20th August. The mslp analysis on
20th August 1987 is presented in Fig. 25. It is possible that this system had even an earlier
history farther east. The disturbance moved westnorthwestward and, travelling across
Indochina peninsula, arrived over Bangladesh territory on 25 August. The track of the
disturbance from 20 to 30 August is also plotted in the mslp analysis map in Fig. 25. Though
the system lost strength gradually after crossing coast, in the course of its travel over land, it
42
remained identifiable throughout its movement across Indochina peninsula, even as a weak
circulation in the lower troposphere. This is evident in the daily streamline charts at 850 hPa
from 21 to 26 August 1987 presented in Figures 26-31. As the disturbance arrived over
Bangladesh, it got a fresh lease of life due to proximity to the Bay of Bengal as a source of
moisture supply. After formation, the land depression travelled further westward and was
located over Bihar Plateau and adjoining Gangetic West Bengal on 27th; Bihar Plateau and
adjoining northwest Madhya Pradesh on 28th; and northwest Madhya Pradesh and adjoining
south Uttar Pradesh on 29th (India Met. Dept., 1987). It weakened into a well marked low
pressure area on 30 August and finally dissipated over the plains of west Uttar Pradesh.
43
Fig. 26. Flow at 850 hPa on 21 August 1987 (shaded area depicts wind speed in
knots); and track of the remnant of tropical depression from South China Sea
44
45
46
The above system caused heavy rainfall in the southern and the hilly southeastern region
of Bangladesh. The streamline pattern at 850 hPa with orography in the background (Fig. 32)
shows strong westerly flow striking the Arakan range of hills. The 24 hour predicted rainfall
based on the initial conditions of 25 August and valid for 00 UTC of 26 August is shown in
Fig. 33. A plot of 24 hour rainfall as recorded on 26 August 1987 is shown in Fig. 34. The
southern parts of Bangladesh, south of the depression, received very heavy rainfall in this
case, while the northern and central parts got only moderate rainfall. The role of orography
in producing heavy rainfall along the coastal areas is clearly brought out.
47
Fig. 32. Flow at 850 hPa on 26 August 1987 with orography (x100 M) in the
background
48
5.1.3 Monsoon trough over Bangladesh during July, August 1998
The monsoon season of 1998 was also characterised by quite a few active rain spells
during July and August. The peaks in the daily distribution of country average rainfall
occurred on 5 July (59 mm), 13 July (59 mm) and 14 August (73 mm). The analysis of
synoptic situations corresponding to these heavy rainstorms showed that these events were
also associated with the monsoon trough running across Bangladesh. The mean sea level
pressure analyses on 12 July and 13 August 1998 are shown in Figures 35 and 36
respectively. The analysis pattern is similar to the one presented earlier in respect of the July
1987 event, with a trough aligned along the east coast of India and the resulting southerly
moist flow from the Bay of Bengal penetrating into Bangladesh. The flow pattern at 850 hPa
(not shown) also exhibited similar features as in July 1987. In this particular case though the
axis of seasonal monsoon trough was in a northerly position, it could not be described as a
break monsoon situation. On the contrary, the upper air flow pattern was typical of a strong
monsoon condition. The 500 hPa flow was characterised by a prominent east west trough
(Fig. 37) along about 24 deg. north with an embedded circulation over central India. This is a
reflection of the typical monsoon trough sloping southward with height in a strong monsoon
situation. The flow at 300 hPa (Fig. 38) was deep easterly over the north Bay of Bengal and
neighbourhood, with the axis of the subtropical ridge line located farther north, again typical
of a strong monsoon condition. Thus the heavy rain spells in these instances during 1998
were associated with a strong monsoon situation over India with the surface monsoon trough
in a favourable northerly latitude. This is in sharp contrast with the synoptic situation in July
49
1987, when the surface monsoon trough located in a similar position over north Bangladesh
was coupled to a break monsoon type upper air flow as described in the previous section.
50
knots)
51
knots)
5.1.4 Discussion
It turns out from the analysis of synoptic situations in respect of the selected heavy
rainfall events in the two outstanding excess monsoon rainfall years 1987 and 1998 that the
typical synoptic setting for heavy rainfall in Bangladesh is the positioning of the axis of
seasonal monsoon trough in a northerly latitude passing across the country’s territory. The
shifting of the monsoon trough northward usually takes place in the wake of a monsoon
depression or a low pressure area from the Bay of Bengal moving across central and northern
parts of India. In one of the cases analysed in the present study, the surface monsoon trough
located over northern Bangladesh, close to the foothills of Himalayas, was accompanied by
the upper air flow pattern typical of a break monsoon situation in India. A middle
tropospheric cyclonic circulation was seen over western Himalayas across extreme north
India and the upper tropospheric subtropical ridge line (Tibetan High) was displaced
southward with the centre of anticyclonic circulation at 300 hPa positioned over Bangladesh
and neighbourhood. This was the case in July 1987. It may be recalled that 1987 was a
major drought year in India. In another case, i.e. in August 1998, the surface monsoon trough
located in the same position as in 1987 over north Bangladesh, the upper air flow pattern was
typical of a strong monsoon situation, viz., the monsoon trough extending upto the middle
tropospheric levels, sloping southward with height, and the subtropical ridge in the upper
troposphere located in its normal position in a northerly latitude. 1998 was a good monsoon
year in India as well as Bangladesh.
52
In yet another case a tropical depression, probably a typhoon from South China Sea,
moved westward across Indochina peninsula and arrived over Bangladesh as a weak
disturbance. It developed into a land depression after reaching Bangladesh possibly aided by
increased moisture supply from the Bay of Bengal. The depression produced heavy rainfall
in south Bangladesh, particularly in the coastal hill areas. Thus it appears that a prolonged
residence of the axis of the monsoon trough in a northerly latitude across north Bangladesh is
the primary factor leading to persistent heavy rainfall in the country.
It would be necessary to corroborate the inference in the preceding paragraph by

exemplifying the rainfall distribution in the contrasting synoptic situation, when the monsoon
trough is positioned in a southerly latitude with its eastern end shifting to the head Bay of
Bengal. This is usually considered an active monsoon condition for the Indian monsoon
rainfall and is generally a precursor to the formation of a monsoon depression in the Bay of
Bengal. The case in example is the monsoon depression of August 1997 in Bay of Bengal.
The depression formed in the northwest Bay on 20 August and moved inland. The mslp
analysis on 20 August 1997 and track of the depression are presented in Fig. 39. Fig. 40
shows the streamline flow at 850 hPa with orography in the background. The flow over the
territory of Bangladesh is southeasterly and parallel to the hill features in the coastal areas.
Fig. 41 shows the observed rainfall distribution over Bangladesh and neighbourhood as
reported on 21 August. It would be seen that the rainfall over Bangladesh, which lies entirely
to the north of the monsoon trough in this case, was only light to moderate. This distribution
is significantly different from the one that occurred with the synoptic situations described
earlier, viz., the monsoon trough passing across north Bangladesh. The observed rainfall
distribution here is in conformity with the well known feature of a monsoon depression,
where heavy rainfall is mostly confined to the south of the westward moving track due to the
maximum convergence there. The northern sector has strong easterly flow in the lower
troposphere and the rainfall is minimal, more often even dry weather. The forecast model
was run on the initial conditions of 20 August. The 24h forecast precipitation from the
model, shown in Fig. 42, matches with the observed rainfall distribution to a fair degree as far
as spatial distribution is concerned. The predicted rainfall distribution over Bangladesh is
light to moderate as in the observed distribution. The small patch of rainfall seen to the right
of the Chittangong-Cox’s Bazar coast is due to orographic influence.
53
Fig. 40. Flow at 850 hPa on 20 August 1997 with orography (x100 M) in the
background
54
55
As a natural corollary to the above findings in regard to the distribution of rainfall over
Bangladesh being controlled by the position of the monsoon trough, one would be led to
think that the exceptionally high seasonal scale anomaly of the monsoon rainfall in
Bangladesh during 1987 and 1998 would have been the result of super-synoptic seasonal
scale anomalies in the position of the monsoon trough. With this end in view we constructed
the monthly mean and anomaly maps of some flow variables using the NCEP/NCAR
Reanalysis data sets. We present here the monthly mean maps of geopotential contour
heights at 700 hPa for July 1987 and 1998 respectively in Figures 43 and 44. 700 hPa is the
most preferred level in describing the monsoon trough on a seasonal time scale. It may be
seen that in both the cases there exists a well marked low in the contour field in the eastern
Himalaya’s foothills north of Bangladesh. A noteworthy contrast in the two maps, however,
is that a well defined trough runs from the contour low across central parts of India to the
north Arabian Sea in 1998. This is indicative of strong monsoon conditions and 1998 was a
good monsoon year both in Bangladesh and India, as mentioned earlier. On the other hand,
the contour pattern in 1987 has no such east west trough, a north-south trough from the centre
of the contour low to head Bay of Bengal is prominent, and the flow over central and
northwestern parts of India is northerly. 1987 was an exceptional year in that it was a major
drought year in India and an exceptionally high rainfall year in Bangladesh. In further
support of our hypothesis about the position of monsoon trough controlling the monsoon
rainfall in Bangladesh, we present in Fig. 45 the mean monthly contour field for July 2003,
which was an exceptionally poor monsoon year in Bangladesh with one of the lowest rainfall
deficiencies in the two main monsoon months of July and August (Fig. 7). By contrast, 2003
was a bountiful monsoon in India. It is interesting to see that the contour field has a
markedly symmetrical east west oriented monsoon trough running across the breadth of the
north peninsular India south of 25 degree north latitude, which is in sharp contrast to the
patterns of 1987 and 1998. A well defined low exists over Orissa and neighbourhood.
Bangladesh lies to the north of this trough and in a southerly flow regime east of the low. It
is easy to see that these anomalies in the flow pattern during the three exceptional years of
monsoon rainfall in Bangladesh would have been the main contributory factor in shaping the
realised rainfall distribution on the seasonal time scale.
56
Fig. 43. Mean geopotential contour height at 700 hPa in July 1987
(Source: NCEP/NCAR Reanalysis)
57
5.2 Track forecasts of monsoon depressions
5.2.1 Monsoon depression, 20-27 August 1997
SYNOPTIC HISTORY
In this section we describe the results of model experiment in forecasting the track of a
monsoon depression, which originated in the Bay of Bengal and travelled all the way up to
northwest India and adjoining north Pakistan in a rather unusually long track and maintained
its intensity as deep depression until its last stage. The depression formed in the northwest
Bay of Bengal on 20 August 1997, referred to earlier in section 5.1.4 (Fig. 39) from
intensification of a low pressure area which had appeared over the northwest Bay and
adjoining Orissa-West Bengal coast on 17th. It is interesting to note that the formation of this
low pressure area was preceded by appearance of a cyclonic circulation in lower troposphere
over Bangladesh and neighbourhood from 15 August onward. The appearance of cyclonic
circulation itself was preceded by formation of a north south trough in low level monsoon
westerlies, seen from 12 to 14 August. The flow at 850 hPa on 12 August is shown in Fig.
46. The cyclonic circulation shifted southward slightly on 16th and finally slipped into the
head Bay and developed into a surface low on 17th. Formation of a monsoon low in head Bay
of Bengal in this manner, viz., shifting of a cyclonic circulation from Bangladesh southward
and its subsequent development into a monsoon depression is not uncommon. It may be
recalled that under a similar situation described in section 5.1.1 a low shifted southward from
Bangladesh into the head Bay of Bengal on 18 July 1987 and subsequently moved
northwestward as a monsoon low. This phenomenon appears to be a manifestation of the
usual southward swing of the monsoon trough into the head Bay prior to the formation of a
monsoon depression. The streamline flow analyses at 850 hPa from 15 to 17 August are
58
shown in Figures 47-49. The low persisted over northwest Bay on 18 and 19 August and
concentrated into a depression, probably deep, on 20 August (Indian Daily Weather Report).
The mean sea level pressure analysis on 20 Augut 1997 was presented earlier in section 5.1.4
(Fig. 39). The deep depression crossed north Orissa coast on the afternoon of 20 August. It
moved initially in a westnorthwesterly direction across Chhattisgarh and Madhya Pradesh.
After reaching close to Nagpur, it’s direction of movement became northwesterly. It headed
towards Rajasthan and thence crossed over to north Pakistan, where it weakened and merged
with the seasonal trough on 27th evening.
A special feature of this depression was that it gained intensity after crossing coast and
maintained it during the course of its movement across central parts of India. This is evident
from the low level vorticity field from 20 to 24 August presented in Figures 50-54. Another
significant feature noticed was that between 24 and 26 August, as the depression moved
across west Madhya Pradesh and Rajasthan, its associated cyclonic circulation extended to
300 hPa level, which is rather unusual with a disturbance located over land. It appeared from
the flow geometry that the depression was undergoing a transformation into a westerly wave
trough during the above period. The streamline flow at 300 hPa on 26 August shown in Fig.
55 reveals this fact.
knots)
59
knots)
knots)
60
knots)
61
62
63
knots)
TRACK FORECAST
The forecast model was run to produce track forecasts up to 48 hours based on the initial
conditions starting from 20 August. The depression moved westnorthwest on first two days
after crossing the coast. The initial flow at 850 hPa on 20 August and the 24 hour & 48 hour
forecasts are shown in Figures 56-58 and those based on the initial conditions of 21 August in
Figures 59-61. The predicted positions of the depression match very well with the observed
positions as may be seen in the superimposed track of the depression. The movement of the
depression from 22 August onward was comparatively slow as it exhibited a tendency to
change direction towards northwest. The forecast based on the initial conditions of 21
August (Figures 59-61) did indicate this slow tendency in the 48 hour forecast. The model
forecasts based on the initial conditions of 26 August (Fig. 62), when the depression was
located over Rajasthan, indicated a slow northwesterly movement in agreement with
observation. The depression produced heavy to very rainfall in north Rajasthan and Punjab
as recorded on 27 August (Fig. 65). This was well captured by the model in 24 hour forecast,
shown in Fig. 66.
64
Fig. 57. 24h forecast flow at 850 hPa valid on 00 UTC of 21 August 1997 shaded
area depicts wind speed in knots); and track of monsoon depression from Bay of
Bengal
65
Fig. 58. 48h forecast flow at 850 hPa valid on 00 UTC of 22 August 1997 (shaded
Bengal
66
Bengal
Fig. 61. 48h forecast flow at 850 hPa valid on 00 UTC of 23 August 1997
(shaded area depicts wind speed in knots); and track of monsoon depression from
Bay of Bengal
67
Bengal
68
Bengal
Fig. 65. Observed 24h rainfall (cm) on 27 August 1997
69
5.2.2 Monsoon low, 20-28 June 2002 and heavy floods in Gujarat
SYNOPTIC HISTORY
This section deals with a monsoon low which formed in the Bay of Bengal in the third
week of June 2002 during the onset phase of the monsoon over India, crossed West Bengal
coast, travelled westward across central parts of the country all the way upto Gujarat and
created catastrophic conditions in Maharashtra and Gujarat due to continuous heavy rains for
three to four days. This disturbance was responsible for causing many deaths due to
inundation besides seriously disrupting rail and road communications in these states for a
number of days. An outstanding feature of this system was that it produced a high impact
weather event although it remained as only a low pressure area for most part of its life history
except on 25 July when two isobars in the mean sea level pressure analysis were seen and it
could possibly be classified as a depression. Ironically, the year 2002, which started with
devastating floods in Gujarat and Maharashtra in the beginning of the monsoon season due to
the above low pressure area, ended up as one of the worst drought years in the country.
The disturbance originated as a low pressure area, with circulation extending upto mid-
tropospheric levels, in the head Bay of Bengal on 20 June 2002. It crossed West Bengal
coast as a well marked low pressure area on 22 June (Fig. 67) and moved slowly in a
westnorthwesterly direction. The wind analyses at 850 hPa on each day from 22 June to 28
June are shown in Figures 68-74. The low moved due west across Madhya Pradesh to
Gujarat. After reaching Gujarat it changed direction and recurved northeastward, moving
70
over to south Rajasthan and neighbourhood on 28th. It weakened and continued moving
northeastward as a weak cyclonic circulation later.
Fig. 68. Flow at 850 hPa on 22 June 2002 (shaded area depicts wind speed in
knots)
71
knots)
72
knots)
knots)
73
knots)
74
The movement of this disturbance was rather slow as it took six days to travel from the
east coast to west coast. As it moved across Madhya Pradesh, the vortex showed
intensification as revealed by the low level vorticity fields, which are shown in Figures 75-77.
The wind speeds to the south of the low increased to 40-50 knots on 25 June (Fig. 71). The
vortex remained at the intensified stage on 26th and weakened on 27th (Fig. 78). As the low
moved westward, it deposited heavy precipitation along its track. In the initial phase when it
was crossing coast, it gave heavy rainfall at some stations in Gangetic West Bengal.
Subsequently the system caused heavy rainfall in Madhya Pradesh. Nagpur town reported
heavy inundation. Reports also came in of breaches in some bunds in Madhya Pradesh. As
the system further moved westward, it entered a zone where there exists a good chance of its
strengthening due to fresh surge of moisture being pumped into the circulation leading to
enhanced precipitation. Gujarat and northern parts of Maharashtra came in the grip of the
disturbance and very heavy rainfall was reported by many stations in these two states
consecutively for three to four days. The result was a complete disruption of the rail and road
communications due to damage to some railway bridges. Some significant heavy rainfall
reports from stations which fell in the path of the disturbance from east coast to west coast
are produced in Table 2. Gujarat region experienced continuous very heavy rainfall for three
consecutive days from 26 to 28 June 2002. Exceptionally heavy rainfall was reported from
many stations on 27 June with 24 hour rainfall amounts exceeding 50 to 60 cm. The cloud
photographs obtained from Meteosat operating in the Indian region, on two days when
maximum precipitation was recorded in Gujarat, are shown in Figures 79-80. These
photographs show dense, intense and organized cloud features covering Gujarat, Saurashtra
and Kutch, which received the maximum impact of heavy rainfall on these days.
75
76
Fig. 79. Meteosat cloud photograph of the monsoon low on 27 June 2002
(Source: IMD web site)
77
Fig. 80. Meteosat cloud photograph of the monsoon low on 28 June 2002
(Source: IMD web site)
FORECAST OF HEAVY RAINFALL
Occurrence of heavy rainfall of the magnitudes that were recorded in this episode was by
itself an unusual event. The rainfall occurred in association with a disturbance that was
classified only as a well marked low pressure area – only on 25th two closed isobars were
seen, as mentioned earlier. Though the surface pattern of this disturbance did not satisfy the
criterion of the system being classified as a depression, its associated upper air circulation
was intense and extended upto 500 hPa. The central vorticity magnitudes on 25 and 26 June
(Figures 76 & 77 respectively) were of the order of 10-15x10-5 s-1, which were as large as in
the case of deep depression of August 1997 discussed earlier (Figures 52-54). We ran the
model for rainfall prediction based on the initial conditions of each day during the period of
existence of the above low pressure area. The 24 hour forecasts of flow pattern at 850 hPa
and 24 hour rainfall based on the initial conditions of 25 & 26 June 2002 are presented in
Figures 81-84. The predicted positions of the vortex centres as on 26 June and 27 June match
very well with the verifying analyses (refer back to Figures 72 & 73 respectively). As far as
rainfall predictions are concerned, while the areas of heavy rainfall were fairly well indicated
by the model, the predicted maximum rainfall amounts were only of the order of 10-15 cm,
and therefore much underestimated. This is obviously a limitation of the model and points
78
towards a need for developing model interpretation techniques for ‘event’ forecasts. As a
part of this interpretation exercise to explain the exceptionally heavy rainfall, we looked at
the field of integrated horizontal moisture flux divergence (see section 5.1.1) on the
corresponding days. Figures 85-87 depict this field on 25, 26 and 27 June respectively. It is
seen that there exists an extensive area of strong moisture flux convergence on these days,
particularly on 25 and 26 June, covering the areas of Vidarbha, Madhya Maharashtra, and
Gujarat region, where heavy rainfall occurred. Thus though the intensity of the disturbance
on surface level was not equivalent to a depression, the flow and moisture fields were
unusually strong, which, coupled with slow movement of the system, produced conditions
favourable for exceptionally heavy rainfall.
Fig. 81. 24h forecast flow at 850 hPa valid on 00 UTC of 26 June 2002 (shaded
area depicts wind speed in knots)
79
Fig. 83. 24h forecast flow at 850 hPa valid on 00 UTC of 27 June 2002 (shaded
area depicts wind speed in knots)
80
Fig. 85. Integrated horizontal moisture flux divergence (x10-5g cm-2s-1) on 25 June
2002
81
Fig. 86. Integrated horizontal moisture flux divergence (x10-5g cm-2s-1) on 26
June 2002
Fig. 87. Integrated horizontal moisture flux divergence (x10-5g cm-2s-1) on 27

June 2002
82
TABLE 2
Significant amounts of 24h rainfall (cm) associated with monsoon low,

22-27 June 2002 (Source: IMD web site)
23.6.2002
Gangetic Bankura 20, Kharidwar 20, Krishnanagar 18, Kangasabatidam 16,
West Bengal : Tusuma 16, Diamond Harbour 12, Durgapur 10, Simulia 8, Magra 7
24.6.2002
Gangetic: Midnapore 11, Diamond Harbour 10, Mohanpur 9, Rangagora 7
West Bengal
Jharkhand: Jamshedpur 8, Ranchi 7
Chhattisgarh: Raipur 16, Pendra 12
25.6.2002
Vidarbha: Chandrapur 19, Yeotmal 17, Nagpur 12, Akola 12, Gondia 9
Marathwada: Bhusaval 9, Jalna 8, Aurangabad 7
26.6.2002
Marathwada: Manjalegaon 17, Asti 17, Patoda 12, Beed 12, Gervai 11, Kaij 11,
Ambejogai 9, Auragabad 7, Buldana 7
Gujarat Ahwa 25, Karjan 17, Songadh 13, Mandvi 13, Madhuban 12, Navsari 11,
Region: Baroda 11, Valod 9, Dharampur 9, Padri 8, Bardoli 8, Palsana 8, Amod 8,
Silvasa 7, Hansot 7, Surat 7
27.6.2002
Gujarat Pardi 61, Valsad 54, Bansda 33, Umbergaon 30, Navsari 26, Chikhali 22,
Region: Dharampur 21, Karjan 19, Ahwa 19, Surat 15, Mahuva 14, Dhandhuka 12,
Valod 10, Broach 10, Kalol 9, Songadh 9, Vagra 9
Saurashtra, Palitana 17, Wankaner 15, Lilia 15, Amreli 14, Sinor 14, Wadhvan 13,
Kutch &Diu: Dharangadhra 9, Bhavnagar 9, Mahuva 9, Vallabhipur 9, Junagadh 7, Jasdan 7
Konkan & Bhira 32, Thane Belapur 12, Bhandup 9, Mumbai(SCZ) 8, Chiplun 8,
Goa: Dharavi 7
28.6.2002
Gujarat Pardi 28, Halol 21, Vagra 20, Dharampur 18, Karjan 17, Ahwa 15,
Region: Chikhali 15, Bansda 15, Madhuban 14, Umbergaon 13, Daman 12, Songadh 12,
Baroda 11,
Balasinor 11, Savli 10, Thasra 10, Navsari 10, Jambusar 9, Valsad 9,
Kalol 9, Anand 9, Padra 8, Hansot 8, Kheda 8, Ankleshwar 7, Amod 7, Mandvi 7,
Jambughoda 7, Ranpur 7, Nadiad 7, Dabhoi 7
Saurashtra, Bhavnagar 42, Bihor 31, Gocha 28, Jafrabad 26, Paliatana 23,
Kutch & Diu: Vallabhpur 16,Wadhvan 14, Mahuva 12, Morvi 12, Jamnagar 10, Chuda 8,
Bachau 8, Lilia 7
83
5.3 Heavy rainfall and floods in the northeastern region during July 2002
5.3.1 Heavy rainfall in the wake of monsoon low of June 2002
As the monsoon low described in the preceding sub-section 5.2.2 moved over to south
Rajasthan on 28 June 2002 and subsequently northeastward as a weak cyclonic circulation,
the monsoon trough shifted northward in its wake. The mean sea level pressure analysis on
28 June is shown in Fig. 88. Its eastern end passed across the northeastern states of India and
adjacent north Bangladesh. A north-south oriented trough formed in the mslp pattern running
off the east coast of India. The low level monsoon flow over the eastern region became
southwesterly striking the hill features in the area. The synoptic setting was similar to the
one described in section 5.1.1 earlier. A northerly position of the trough and the
accompanying southwesterly lower tropospheric flow striking the orographic features gave
rise to heavy rainfall in the eastern Himalayas, viz. Sub-Himalayan West Bengal, Sikkim,
submontane areas of Bihar, Nepal Himalayas, and the Brahmaputra valley further east. With
the shifting of the monsoon trough northward, the rainfall in Bangladesh also increased from
29 June 2002 onward and peaked on 2 July when a country average of 71 mm was recorded
(Fig. 90). The Meteosat cloud photograph of 2 July morning presented in Fig. 91 shows
dense clouds covering the eastern parts of India and Bangladesh. Damage and dislocation of
life due to landslides was reported from Sub-Himalayan West Bengal. Submersion of several
areas in upper Assam valley and breaches in a national highway were also reported due to
heavy floods in the Brahmaputra river.
84
Fig. 89. Flow at 850 hPa on 28 June 2002 with orography (x100 M) in the
background
80
70
60
50
Rainfall (mm)
40
30
20
10
0
1 5 9 13 17 21 25 29 33 37 41 45 49 53 57 61 65 69 73 77 81 85 89 93 97 101 105 109 113 117 121
Day (1 June-30 Sept.)
September) 2002
85
Fig. 91. Satellite photograph of monsoon trough located close to Himalayan foothills
obtained from Meteosat-5 on 2 July 2002 (Source: IMD web site)
5.3.2 Heavy rainfall and floods during last week of July 2002
Another example of a well developed break monsoon situation leading to high intensity
floods in the entire stretch of Brahmaputra river in Assam during the last week of July 2002
is presented in this section. Fig. 92 shows an interesting photograph of a flood scene in
Brahmaputra at Guwahati on 23 July 2002. The river was flowing more than half a meter
above danger level with the river top very close to the adjacent road level. The synoptic
situation which led to this flood wave was as usual a low pressure area from Bay of Bengal
crossing Andhra Pradesh, moving across central parts of the country, followed by shifting of
the monsoon trough close to the foothills of Himalayas in the third week of July. The mean
sea level pressure analysis on 20 July 2002 is shown in Fig. 93. The streamline flow map on
20 July with orography in the background is presented in Fig. 94. Southerly to southwesterly
flow from the Bay of Bengal to the east of a cyclonic circulation over eastern parts of India is
seen striking the north Bengal-Sikkim Himalayas in a perpendicular direction. Incessant
heavy rains followed in the eastern Himalayas covering the entire belt from Bihar to
Arunachal Pradesh in association with this synoptic situation. The catchments of rivers rising
in Nepal Himalayas also received intense rainfall during this spell leading to heavy discharge
and floods in north Bihar. Bangladesh also experienced widespread heavy rainfall during this
period, particularly the eastern parts along the coast and the hills to the east. Some of the
significant amounts of rainfall from 21 to 24 July are produced in Table 3. Fig. 95 contains
86
the satellite picture of 22 July showing extensive heavy clouding over the entire eastern
region including Bangladesh.
Fig. 92. Floods in Brahmaputra river in Assam due to break monsoon conditions in
the third and last week of July 2002 (Photo: courtesy Sentinel, Guwahati, 23 July 2002)
87
Fig. 93. Mean sea level pressure analysis on 20 July 2002, 00 UTC
Fig. 94. Flow at 850 hPa on 20 July 2002 with orography (x100 M) in the
background
88
Fig. 95. Satellite photograph of monsoon trough located close to Himalayan foothills
obtained from Meteosat-5 on 22 July 2002 (Source: IMD web site)
The forecast model was run during the period 20-24 July 2002. The 24 hour forecast
rainfall produced by the model based on the initial conditions of 20 July is shown in Fig. 96.
It is seen that the heavy rainfall belt along the eastern Himalayas from Nepal to Arunachal
Pradesh is very well simulated by the model. Bangladesh also experienced a spell of heavy
rainfall during this period with a peak of 55 mm on 22 July (Fig. 90). Fig. 97 shows the 24
hour rainfall forecast from the initial conditions of 21 July 2002. The simulation has captured
the heavy rainfall in Bangladesh as well as the neighbouring areas of India.
89
90
TABLE 3
Significant amounts of 24h rainfall (cm) from 21 – 24 July 2002 associated with monsoon trough
close to the foothills of Himalayas (Source: IMD web site/BMD)
21 July
Arunachal Pradesh: Passighat 20, Bhalukpong 9
Assam & Meghalaya: Chouldowghat 13, Aie N.H. X-ing 8, Goalpara 7
Sub-Himalayan: Chepan 25, Barobisha 22, Hasimara 18, Alipuduar 18, Coochbihar 17
West Bengal & Mathabhanga 11
Sikkim
Bangladesh: Chittagong 16, Rangamati 10, Kutubdia 8
22 July
Assam & Meghalaya: Chouldowghat 13, Nalbari 10, Beki Mathanguri 9, Matunga 9
Sub-Himalayan: Sevoke 25, Gajoldoba 21, Chanpasari 17, Baghdogra 16, Khanitar 15,
West Bengal & Darjeeling 14, Jalpaiguri 12, Singlabazar 9, Domohani 9, Tadong 7
Sikkim
Bihar: Sonbarsa 20, Galgalia 14, Chanpatia 13, Birpur 12, Samastipur 12,
Taibpur 10, Coalgaon 9, Khagaria 8, Jainagar 8, Triveni 7,
Jhanjharpur 7, Hayaghat 7, Chatia 7
Bangladesh: Sitakunda 14, Teknaf 13, Rangamati 13, Chittagong 12, Madaripur 9
Cox’s Bazar 8, Hatiya 8, Bogra 8, Barisal 8, Dhaka 7, Patuakhali 7
23 July
Arunachal Pradesh: Passighat 7
Assam & Meghalaya: Matunga 12, Shillong 11, Gosaigaon 10, Bihubar 8, Beki Rd. Bridge 7,
Dhubri 7, Goalpara 7
Sub-Himalayan: Coochbihar 9, Mathabhanga 7, Baghdogra 7
West Bengal &
Sikkim
Bihar: Birpur 16, Taibpur 10, Araria 10, Galgalia 10, Jainagar 7, Chargharia 7
Bangladesh: Kutubdia 15, Chittagong 9, Comilla 9, Srimongal 9, Rangamati 8,
Dinajpur 8, Chandpur 7, Rangpur 7, Teknaf 7
24 July
Arunachal Pradesh: Roing 8, Tezu 7, Bhalukpong 7, Passighat 7
Assam & Meghalaya: Choudowlghat 13, Beki Mathanguri 12, AIE N.H. X-ing 7
Sub-Himalayan: Chepan 17, Alipurduar 15, Cooch Bihar 14, Barobisha 14, Gajoldoba 13,
West Bengal & Hasimara 13, Jalpaiguri 12, Diana 11, Nagrakata11, NH-31 11,
Sikkim Domohani 10, Champasari 8, Sevoke 8
Bangladesh: Cox’s Bazar 19, Rangamati 15, Sylhet 11, Chittagong 7
25 July
Sub-Himalayan: Mathabhanga 12, Baghdogra 10, Cooch Bihar 10, Domohani 9, Sevoke 9,
West Bengal & Hasimara 8, Jalpiaguri 8, Alipurduar 9, Champasari 7
Sikkim
Bihar: Galgalia 14, Birpur 13, Taibpur 9
91
4. Concluding remarks
A diagnostic analysis of some flood producing heavy rainstorm events during the
southwest monsoon period and application of a limited area NWP model based on Florida
State University for short range forecasting of monsoon disturbances and associated rainfall
was undertaken in this study. The focus of the study was to identify the synoptic situations
responsible for causing persistent heavy rainfall and consequent floods in Bangladesh during
the southwest monsoon seasons of the two outstanding excess rainfall years 1987 and 1998.
The synoptic weather maps for diagnostic analyses and the initial and lateral boundary data
fields for running the forecast model were constructed from the ECMWF ERA-40 GRIB data
sets downloaded from their ftp site. The weather maps prepared from the above data sets
faithfully reproduced the flow patterns, which matched the descriptions of weather systems in
the published literature such as Indian Daily Weather Reports. The analysis of synoptic
situations in respect of selected heavy rainfall events in the above two years showed that the
typical synoptic setting for heavy rainfall in Bangladesh is the positioning of the axis of
seasonal monsoon trough in a northerly latitude passing across northern parts of Bangladesh.
The surface monsoon trough located over northern Bangladesh may either be associated with
a typical break monsoon situation in India as happened in 1987 or it may be associated with a
strong monsoon situation as in 1998. 1987, which recorded a significant warm episode of
ENSO, was a major drought year in India. It was one of the highest excess rainfall year in
Bangladesh. 1998 was a good monsoon year in India as well as Bangladesh. The monsoon
trough may often go into a quasi-stationary mode, persisting over the area for a long duration.
This situation may cause persistent heavy rainfall within Bangladesh and the catchment areas
upstream, leading to floods. The years 1987 and 1998, in particular, witnessed a situation in
which the persistence of monsoon trough in its northerly position was prominently reflected
even in the mean monthly maps of flow pattern. This obviously suggested that the
anomalous flow in the two years under consideration occurred on a super-synoptic time scale
and that it was an interplay of global scale circulation anomalies responsible for creating
quasi-stationarity of the monsoon trough in a northerly position. These aspects need to be
further investigated from the point of view of long range forecasting of monsoon rainfall.
A land depression may sometimes form over Bangladesh, which may have its origin
farther east over South China Sea. This is a favourable situation for occurrence of heavy
rainfall in south Bangladesh and the coastal belt off Arakan coast. The study showed that
when a monsoon depression forms in its usual position over north Bay of Bengal with the
territory of Bangladesh lying to its north and east, the rainfall over the country is only light to
moderate, which is in conformity with the typical structure of a monsoon depression.
The limited area forecast model was run for short range forecasts of precipitation during
the selected days of heavy rainfall occurrence in Bangladesh during the above episodes of
1987 and 1998, as also a few cases of heavy rainfall and flooding in the northeastern region
of India. The entire region comprising the northeastern states of India and Bangladesh is
dotted with significant orographic features which have a dominating influence on the
occurrence of heavy rainfall in this part of the subcontinent. The forecasts of rainfall
produced by the model did bring out this influence prominently and were reasonably well
placed in regard to simulating the observed areas of rainfall although the predicted
magnitudes remained underestimated.
The forecast model was also run in two cases of monsoon disturbances forming in the
Bay of Bengal and moving across central and northern parts of India during August 1997 and
92
June 2002. These disturbances had some exceptional features. The depression of 20-27
August 1997 had a long history in that it travelled all the way upto northwest India and north
Pakistan and maintained its intensity as a deep depression throughout its movement until it
dissipated. The model was able to simulate its movement to a good degree of accuracy even
in 48 hour forecast. The rainfall forecasts were also reasonably well placed. The monsoon
low of June 2002, also went through a long history with a slow movement and intensified
after entering land. Even though the disturbance was classified as only a low pressure area
on the surface, its associated upper air cyclonic circulation was intense and deep, as much as
the deep depression of August 1997. The low caused exceptionally heavy rainfall for
consecutive three days in the Gujarat-Maharashtra belt, disrupting rail and road traffic due to
heavy flooding. While the model simulations of the movement of this disturbance were
accurate enough, the rainfall amounts were much underestimated compared to observations.
Forecasts of the rainfall magnitudes that were recorded in this episode are by no means
possible to foresee by synoptic analysis either. This points towards the need for developing
model output interpretation techniques for ‘event’ forecasting in such cases, based on
additional derived products like vorticity, moisture flux divergence etc., which could provide
useful numerical guidance in such cases. The computation of integrated moisture flux
divergence in the June 2002 case did indicate an unusually strong field, which was a
supportive evidence for the occurrence of exceptionally heavy rainfall. Further, while
movement of a monsoon depression is fairly well predicted by the model, its initial
development is hardly captured. These are the limitations of the model which need to be
addressed. These issues are targeted as our future work programme.
93
References
Dhar, O.N., Kulkarni, A.K. and Mandal, B.N., 1984, The most severe rainstorm of India – A
brief appraisal, Hydrological Sciences, Journal des Sciences Hydrologiques, 29(2), 219-229.
Gadgil, S., 1988, Recent advances in monsoon research with particular reference to the
Indian monsoon, Aust. Met. Mag., 36, 193-204.
Gadgil, S. and G. Asha, 1992, Interseasonal variation of the summer monsoon I:

Observational aspects, Meteorological Society of Japan, 70, 517-527.
Goswami, B.N., 1993: Difficulties in simulation of time mean circulation: some thoughts,
Simulation and Prediction of Monsoons, Recent Results, TOGA/WGNE Monsoon Numerical
Experimentation Group, New Delhi, India, 12-14 January 1993, WCRP-80, WMO/TD-No.
546, 27-32.
India Met. Dept., 1987, Indian Daily Weather Reports, July & August,1987.
Krishnamurti, T.N., H.L. Pan, C.B. Chang, J. Ploshay, D. Walker, and A.W. Oodally, 1979,
Numerical Weather Prediction for GATE, Q.J.R. Meteorol. Soc., 105, 979-1001.
Krishnamurti, T.N. and Y. Ramanathan, 1982, Sensitivity of monsoon onset to differential

heating, J. Atmos. Sci., 39, 1290-1306.
Krishnamurti, T.N., 1985, Nymerical weather prediction in low latitudes, Adv. Geophys. 28,
283-333.
Krishnamurti, T.N., S. Low-Nam, A. Kumar, J. Sheng, M. Sugi, 1987, “Numerical Weather

Prediction of Monsoons, Monsoon Meteorology”, pp. 501-544, Oxford Univ. Press, London.
Krishnamurti, T.N. and Oosterhof, D., 1989, Prediction of the life cycle of a supertyphoon
with a high resolution global model, Bull. Am. Meteorol. Soc., 70, 1218-1230.
Krishnamurti, T.N., Arun Kumar, K.S. Yap, Ashu P. Dastoor, Noel Davidson, and Jian
Sheng, 1990, Performance of a high-resolution mesoscale tropical prediction model,
Advances in Geophysics, 32, 133-286.
Kulshrestha, S.M., 2000, “Flood Management in India”, Joint COLA/CARE Technical Report
No. 3, July 2000.
Lau, K.M. and P.H. Chan, 1986, Aspects of the 40-50 day oscillation during northern
summer as inferred from OLR, Mon. Wea. Rev., 114, 1354-1367.
Mathur, M.B., 1983, A quasi-Lagrangian regional model designed for operational weather
prediction, Mon. Weather Rev., 111, 2087-2098.
McGregor, J.L. and L.M. Leslie, 1977, On the selection of grids for semi-implicit schemes,
Mon. Weather Rev., 105, 236-238.
94
Mesinger, F. and A. Arakawa, 1976, “Numerical Methods Used in Atmospheric Models”,
GARP Publ. Ser. No. 14, WMO/ICSU J. Organ. Comm., World Meteorol. Organ., Geneva.
Mirza, M.M.Q., R.A. Warrich, N.J. Erichsen and G.J. Kenny, 2001, “Are floods getting
worse in the Ganges, Brahmaputra and Meghna basin?”, Environmental Hazards, 3, 38-48.
Prasad, K., 1992, A diagnostic study of flood producing rainstorm of September 1988 over
northwest India with the aid of a fine mesh numerical analysis system, Mausam, 43, 1, 71-76.
Prasad, K., Y.V. Rama Rao and Sanjib Sen, 1997, "Tropical cyclone track prediction by a
high resolution limited area model using synthetic observations", Mausam, 48, 3, 351-366.
Rao, Y.P., 1976, “Southwest Monsoon”, India Met. Dept. Monograph, Synoptic Meteorology,
No. 1, 1976.
Sikka D. R., and Gadgil S. 1980: On the maximum cloud zone and the ITCZ over Indian
longitudes during the southwest monsoon, Mon. Wea. Rev., 108, 1840-1853.
SMRC, 2003, SAARC Meteorological Research Centre (SMRC) Newsletter, Vol. 10, No. 1,
January-June 2003.
SMRC, 2003, SAARC Meteorological Research Centre (SMRC) Newsletter, Vol. 10, No. 1,
January-June 2003.
Quadir, D.A., Md. Amirul Hussain, Md. Akram Hossain, Nazlee Ferdousi, Md. Majajul Alam
Sarker and Md. Mizanur Rahman, 2003, Climate Change and its impacts on Bangladesh
floods over the past decades, Proceedings of SAARC Seminar on Climate Variability in the
South Asian Region and its Impacts, 10-12 December 2002, SAARC Meteorological
Research Centre, Dhaka, Bangladesh.
Yasunari, T., 1979, Cloudiness fluctuations associated with the northern hemisphere summer
monsoon, J. Meteor. Soc. Japan, 57, 227-262.
Yasunari, T., 1980, Quasi stationary appearance of 30-40 day period in the cloudiness
fluctuations during summer monsoon over India, J. Meteor. Soc. Japan, 58, 225-229.
Yasunari, T., 1981, Structure of an Indian summer monsoon system with around 40-day
period, J. Meteor. Soc. Japan, 59, 336-354.
95

Whether Calculation

Uploaded by

Document Information

Original Description:

Copyright

Available Formats

Share this document

Share or Embed Document

Sharing Options

Did you find this document useful?

Is this content inappropriate?

Copyright:

Available Formats

Whether Calculation

Uploaded by

Copyright:

Available Formats

MONSOON FORECASTING WITH A LIMITED AREA

NUMERICAL WEATHER PREDICTION SYSTEM

SAARC METEOROLOGICAL RESEARCH CENTRE

(A joint study by SAARC Meteorological Research Centre, Dhaka;

SAARC METEOROLOGICAL RESEARCH CENTRE

S.K. Roy Bhowmik

Majajul Alam Sarker

(Md. Akram Hossain)

2. The basic framework of the FSU limited area model 2

3. Monsoon forecasting – the principal rain bearing weather systems 9

Table 1: List of symbols

Fig. 3. Orography used in the limited area model

Fig. 5. Map of Bangladesh showing location of synoptic stations

Fig. 6. Map of India showing meteorological subdivisions

Fig. 12. Mean sea level pressure analysis on 24 July 1987

Fig. 19. 24 hour precipitation (cm) as observed on 25 July 1987

Fig. 20. 24 hour precipitation (cm) as observed on 26 July 1987

Fig. 21. Mean sea level pressure analysis on 31 July 1987

Fig. 22. 24 hour precipitation (cm) as observed on 1 August 1987

Fig. 24. Mean sea level pressure analysis on 26 August 1987

Fig. 25. Mean sea level pressure analysis on 20 August 1987

Fig. 34. 24 hour rainfall (cm) as recorded on 26 August 1987

Fig. 35. Mean sea level pressure analysis on 12 July 1998

Fig. 36. Mean sea level pressure analysis on 13 August 1998

Fig. 39. Mean sea level pressure analysis on 20 August 1997

Fig. 41. 24 hour rainfall (cm) as recorded on 21 August 1997

Fig. 65. Observed 24h rainfall (cm) on 27 August 1997

Monsoon forecasting is a challenging problem in the whole of south Asia where

2. The basic framework of the FSU limited area model@

where p is the pressure and p s is the pressure at the Earth’s surface.

Thermodynamic energy equation

Mass continuity equation

The quantity π κ in equation (2.7) is obtained from the Poisson equation:

The operator D / Dt in equations (2.2)-(2.6), which represents the horizontal Lagrangian

which can be written in abbreviated form as:

LATERAL BOUNDARY CONDITIONS

The physical processes that are incorporated in the model are:

• Large scale condensation

σ 10 = 0.750 P10 , σ& 10 , z10

σ 11 = 0.800 P11 , σ& 11 , z11

σ 14 = 0.950 P14 , σ& 14 , z14

σ 15 = 0.990 P15 , σ& 15 , z15

σ 16 = 1.0 P16 , σ& 16 , z16

∆φ v ---- ---- v ----------------------------------v ------ v

3. Monsoon forecasting – the principal rain bearing weather systems

3.2 Monsoon depressions

FORMATION AND MOVEMENT

ASSOCIATED RAINFALL PATTERN

WATER POTENTIAL OF MONSOON DEPRESSIONS

Fig. 5. Map of Bangladesh showing location of synoptic stations

All India Monsoon Rainfall (1901-2003)

Daily monsoon rainfall (country average) - 1987

Daily monsoon rainfall (country average) - 1998

Fig. 20. 24 hour precipitation (cm) as observed on 26 July 1987

Fig. 22. 24 hour precipitation (cm) as observed on 1 August 1987

5.1.2 Land depression, 26-30 August 1987

Fig. 24. Mean sea level pressure analysis on 26 August 1987

5.1.3 Monsoon trough over Bangladesh during July, August 1998

Fig. 35. Mean sea level pressure analysis on 12 July 1998

It would be necessary to corroborate the inference in the preceding paragraph by

5.2 Track forecasts of monsoon depressions

5.2.1 Monsoon depression, 20-27 August 1997

Fig. 65. Observed 24h rainfall (cm) on 27 August 1997