The state-of-the-art
A forecasting system has to exploit
available observations and knowledge and build on the experience
gained in forecasting in the past. In order to build a forecasting
system for India, it is pertinent to examine the state-of-the-art in
countries that have developed such systems, and to review the current
status in India. The following description is based on the concept
paper by Shetye and Radhakrishnan (2004).
The forecasting experience in Europe
The state of ocean forecasting that existed in Europe in the early
1990s has been described by Dugan (1993). By the early 1990s, several
countries of the European Union had an operational forecasting
system: United Kingdom, The Netherlands, Belgium, Germany, Denmark,
Norway, and France. A major player in the forecasting system of all
these countries was the European
Centre for Medium-Range Weather Forecasting (ECMWF).
The ocean
variables that figured in the forecasts were tide (water level),
storm surge, surface waves, water depth, and currents. The following
were under consideration for inclusion in forecasts: water quality,
pollution, and variables of interest to aquaculture and fisheries.
The models that were used in tidal and storm surge prediction were
typically two dimensional (covering the two horizontal dimensions)
and were found to be reasonably accurate for water level predictions
but not for water currents. For predicting wind waves, the model used
most often was the one developed by Wave Modelling Group and usually
known as WAM. The wind field used in the forecasts was provided by
ECMWF using their fine resolution atmospheric model. In his
evaluation of the forecasting systems, Dugan noted that the models
used in the systems were comparable to the best available then.
There also
existed a wide network of users of these forecasts (organizations
involved in navigation, water quality, and safety, the offshore
industry, fisheries and aquaculture industry, navies and coast
guards), and these users were serious about the advice they received
and provided real feedback to the developers and operators of the
forecasting systems. In the early 1990s, Europe had plans to extend
the forecasts to water quality, ecology, fisheries, aquaculture,
sedimentation, and pollution.
The forecasting experience in the
United States
The national
organization identified for coastal ocean forecasting in the United
States is the National Centre for Environmental Prediction (NCEP).
Rao (2003) provides an overview of the present status of NCEP’s
operational regional ocean forecasting systems for the US coastal
domains. NCEP has been involved in the development of nowcast and
forecast models for predicting the state of the US coastal oceans and
implementing the models to produce daily forecasts on an operational
basis.
As a first step
in this direction, an area off the US east coast was chosen as the
pilot domain to test the feasibility of producing coastal ocean
nowcast and forecast fields on a daily basis in real-time. The model
domain extended from approximately 26.5°N
to 48°N and from the
coast out to 50°W.
This domain off the east coast was initially chosen because of the
robust and identifiable signal provided by the Gulf Stream, which
covers a major portion of the domain, and the availability of better
atmospheric forcing compared to the US west coast.
The Princeton
Ocean Model (POM)
was used as a prototype to produce daily nowcast and forecast
oceanographic fields. The horizontal resolution in the ocean model
changed from 10 km near the coast to about 20 km offshore and the
model had 19 sigma levels in the vertical. The prescription of
open-ocean and landward boundary conditions was based on monthly
climatologies. The ocean model was forced by surface fluxes derived
from the Eta model, which is NCEP’s operational meso-scale meteorological
model. Forcing due to three semi-diurnal and three diurnal tidal
components were also included in the model. Assimilation of sea
surface temperature (SST) data from in-situ and satellite sensors and
the sea level anomalies from TOPEX/Poseidon altimeter and ERS-2 scatterometer was also included in the forecast system. The system
produced a nowcast and a 48-hour forecast every day.
The
oceanographic products produced by the POM-based system have been
extensively evaluated internally by the Marine
Prediction Center of NCEP and also by a group of selected marine
users during two marine demonstrations projects conducted during the
summer of 1999 and winter of 2000 (funded by the National
Oceanographic Partnership Program (NOPP)).
The nowcast-cum-forecast system for the east coast became fully
operational at NCEP in March 2002. Graphic products and grid-point
numerical fields are made available to the user community over the Internet.
Rao (2003) also
points out that the then operational east coast system was limited in
its geographical coverage and that improvements were needed in the
prescription of the lateral open-ocean boundary conditions. In order
to rectify these deficiencies, work is continuing on the development
of basin-scale models to provide the means for extending the
forecasting capability to other coastal areas of the US and to
generate more realistic open-ocean forcing for these regional
domains.
Elements of a coastal forecasting
system
The experience
of forecasters in the United States and Europe enables us to
delineate the following as the essential elements of a coastal
forecasting system.
Basin-scale
models that can provide the boundary conditions needed by the
coastal models: a hierarchy of models is needed because there is an
interaction between the coastal ocean and the open ocean, and the
coastal models, which cover a smaller domain at a higher resolution,
need to be provided with the conditions at their open-ocean end.
These open-ocean boundary conditions are provided by models that are
run over a larger domain (basin-scale models) but at a coarser
resolution.
- Models
of the coastal ocean and atmosphere with data assimilation capability: The coastal ocean model must incorporate the knowledge
concerning behaviour of the coastal ocean. The atmospheric model
must be able to provide forecasts of fields of winds, pressure, etc.
Both these models must have the ability to assimilate data so that
the simulated fields stay close to the observed.
- An observing system that provides crucial information to check
the performance of the model(s). The modelling studies, in turn,
often help decide observational strategy using these platforms.
- A set-up
for dissemination of forecasts.
- A system to
objectively evaluate performance of the prediction system and
provide feedback on performance of the forecasting system.
The relation
between these elements can be depicted in a schematic.
Illustration
1: Schematic showing the relation between the different elements of
a coastal forecasting system
The state-of-the-art in India
Given the above
elements of a coastal forecasting system, we now describe the present
state-of-the-art in India for each of these elements. We focus on the
physical variables because we understand them much better than we
understand other variables such as those related to the ecology of
the coastal region. We examine, in turn, the models used in India,
the observing systems available today, the dissemination of the
forecasts being made now, and the user-feedback process currently in
place.
Basin-scale
models
During the last
decade and a half, considerable progress has been made in
understanding the currents in the waters around India ( McCreary
et al., 1993; Shetye and Gouveia, 1998; Shankar, 1998; Schott and
McCreary, 2001; and Shankar et al., 2002). This knowledge was the
result of a series of observations, mostly relying on ship-based
cruises, and modelling studies with a suite of models. Though the
models ranged from very simple dynamical models to complex Ocean
General Circulation Models (OGCMs), they had one feature in common:
all these studies were simulations on the scale of the basin,
typically modelling the entire Indian Ocean north of about 30°S.
These studies
show that the current along the east or west coast of India (called
East India Coastal Current (EICC) and West India Coastal Current
(WICC) respectively) can be forced by events anywhere in the north
Indian Ocean, stretching from the Asian landmass in the north to
about 10°S. For
example, if we want to explain the variability seen in a time series
of currents measured off Goa, we cannot correlate the current merely
to the local wind around Goa: the current off Goa, the WICC, is
forced not just by the local winds, but is also affected by winds
elsewhere in the basin. At the seasonal time scale (typically
monthly-mean wind forcing and the resulting monthly-mean currents),
the WICC off Goa is forced more by the winds blowing along the east
coast of India than by the local winds. This phenomenon is called
remote forcing: the cause (winds) of the observed behaviour (of, say,
currents) is not local, but remote from the region where the
(current) observation is made. This basin-wide integration is made
possible by waves (that are distinct from the surface waves we see on
the sea surface) that propagate long distances, transmitting the
disturbance generated at a location to remote areas. The studies over
the past decade show that this combination of local and remote
forcing merges the Arabian Sea, the Bay of Bengal, and the equatorial
Indian Ocean (down to about 10°S)
into a single dynamical entity, the north Indian Ocean, which must be
modelled as a whole to simulate the circulation even in its parts.
Hence, any
modelling system that is used for the forecasting system has to cover
at least this domain. Such a system can be built with just one model
that covers the entire domain of interest, but the typical solution
in such cases is to run two models, one with a somewhat coarse
resolution, but at the scale of the basin (covering, say, at least
the entire Indian Ocean), and the other at a much higher resolution,
but over a limited domain covering the region of interest (say, the
Indian EEZ). The former is a basin-scale model and the latter could
be called a coastal-ocean or regional model. The coastal ocean is
affected by the conditions in the open ocean. Hence, the coastal
model needs the conditions at its open-ocean end, and these
conditions are provided by the basin-scale model. Computational
limitations usually preclude running the high-resolution model over
the entire, larger domain.
The most popular
basin-scale model is the Modular Ocean Model (MOM),
which has a history stretching back to the 1960s. It has also been
used extensively for simulating the circulation in the Indian Ocean.
This rigorous testing over two decades, as exemplified in several
research papers, makes MOM a leading candidate for basin-scale
simulations.
Models
of the coastal ocean
Models used for
coastal forecasting include POM and the Regional Ocean Modelling System (ROMS).
POM has been used for forecasting on the US east coast (Rao, 2003);
ROMS has been used for modelling the California
Current System on the US west coast. Both models are currently
being used in India. Another model for coastal studies is MIKE
21, a commercial software that has been used for Environmental
Impact Assessment (EIA) studies in the Indian coastal waters. The
documented research using these models in the Indian seas, however,
is meagre in comparison with that using basin-scale models like MOM.
Also, it is only now that some groups in India are beginning to use
models like ROMS in association with basin-scale models, with the
open-ocean boundary-conditions from the larger-scale models being
used as a forcing for the high-resolution, regional model.
A model,
however, is only as good as are the forcing functions. In the Indian
coastal regime, the two most important forcing functions are the
tides and the winds; in some areas like the northern Bay of Bengal,
river runoff will also be important.
For tides, a
reasonably good database is available for the Indian seas (see, for
example, Unnikrishnan et al., 1999).
The traditional
source of gridded wind fields for the Indian research community has
been the National Centre for Medium-Range Weather Forecasting
(NCMRWF), New Delhi; other
sources include ECMWF and NCEP. Before these products can be
assembled into an ocean forecasting system, however, it is essential
to test them extensively: ocean models need to be forced using these
wind products and the model outputs compared with available
observations. Such validation, necessary before a forecasting system
can be assembled, is lacking in the Indian seas. There is also ample
evidence that the quality of wind forcing in the region around India
improves if the atmospheric models are run at higher resolution (see,
for example, Roy Bhowmik, 2003), calling for a combination of
meso-scale atmospheric models and Atmospheric General Circulation
Models (AGCMs) (as required for the ocean) to generate the forcing
fields needed for an ocean forecasting system.
There is,
however, a dearth of data on the discharge of rivers into the seas
around India. Discharge gauges are not installed on all rivers. Also,
data on river discharge are often difficult to obtain. Both
limitations make numerical modelling of river discharge, the models
being forced by precipitation, an important, if auxiliary, element of
an ocean forecasting system. In regions like the northern Bay of
Bengal, into which flow the largest rivers of India, it is unlikely
that a reasonable forecasting system can be assembled without this
element. Hence, hydrological
modelling forms a part of this programme.
A key component
of modern oceanic and atmospheric forecasting systems is data
assimilation. Models used for forecasting assimilate available
observations in real-time to improve the quality of forecasts. Though
some work on ocean data assimilation has been done in India, it is in
a nascent state and cannot yet be included as an element of a
forecasting system. Data assimilation capability is yet to be tested
for the ocean models identified above. Hence, this is one area where
considerable basic work has to be done before it can be integrated
into a forecasting system for the Indian seas.
Observing
system
Since the successful completion of the
International Indian Ocean Expedition (IIOE) in the 1960s, India has
made handsome investment in ocean research. Besides developing
institutions dedicated to ocean research and teaching, the country
has also invested in infrastructure for sustained observations of the
ocean. The data from the observing system has two applications in the
context of a forecasting system. First, the data are required to
validate the model simulations: how good are the predictions? Thus,
the observing system is required to provide the benchmarks for
evaluating the forecasting system itself. This requirement calls for
a carefully planned, sustained observational programme. Second, if
the forecasting system includes data-assimilation capability, then
the data have to be available in real-time or near-real-time in order
to be assimilated into the models. At present, very few of the
observing systems in India have this capability.
The components of the present ocean
observing system that either have this capability or could be
upgraded to this state are the following.
Moored buoys. The National Data Buoy Programme (NDBP)
of the National Institute of Ocean Technology (NIOT),
an institute funded by the Ministry of Earth Sciences (MoES),
maintains about 20 moored
buoys at present in the waters around India; there are plans to
increase this number to 40. These buoys measure oceanographic and
surface meteorological variables, including wind speed and
direction, (ocean) current speed and direction, sea surface
temperature (SST), air temperature, barometric pressure, etc. Some
of the buoys also record thermal structure in the upper ocean and
transmit the data in real-time using communication satellite
systems.
- Sub-surface
moorings. In addition to the moored buoys
deployed by NIOT, NIO is deploying sub-surface moorings with current
meters and ADCPs (Acoustic Doppler Current Profilers) under this programme on
enabling a forecasting system for the Indian seas. Unlike the NIOT
buoys, these buoys are anchored at the bottom and current meters and
an ADCP (looking up) are deployed on the mooring line. These
moorings, being deployed for long-term current measurements in the
Indian EEZ, are part of the observational
component of this programme on building the science underlying a
forecasting system.
- Drifting buoys. The drifting
buoys are used primarily to measure SST, but they often include
other sensors. The Indian drifting
buoy programme, which launches about __ buoys every year, is
operated by the National Institute of Oceanography (NIO),
Goa, with funding from the Indian National Centre for Ocean
Information Services (INCOIS).
- ARGO floats. A recent advance in
near real-time in-situ observation of the thermohaline structure in
the upper 2000 m of the ocean is the system of profiling ARGO floats. The international community has a global array of 3000 ARGO
floats. India has contributed 160 floats to this global programme.
The Indian
ARGO programme is coordinated by INCOIS.
- Expendable
bathythermographs (XBT). XBTs,
launched from ships-of-opportunity, measure the temperature profile
along its route. Three major lines are being operated by India at
present: Mumbai to Mauritius, Chennai to Port Blair, and Port Blair
to Kolkata; an extensive XBT survey was also launched in the
southeastern Arabian Sea during the ARMEX (Arabian Sea Monsoon Experiment) programme. The present Indian XBT
programme (operated by NIO and funded by INCOIS) does not have the
capability for real-time transmission. It can, however, be upgraded
to transmit data in near real-time or even real-time.
- Tide gauges. The Survey
of India, Dehra Dun, operates a network of about 25 tide gauges
in major ports of India. Not all have real-time transmission
capability, but can be upgraded to have this ability.
- Meteorology. The India Meteorology Department (IMD)
operates a network of meteorological observatories, coastal radars,
etc. as a part of its weather forecasting system. Much of these data
are available on the Global Transmission System (GTS)
used worldwide for meteorological forecasting.
- Satellite
observations. The Indian
Space Research Organization (ISRO)
at present operates one satellite, Oceansat
I, dedicated to ocean observations. There are plans to have a
second satellite, Oceansat II. The most important sensor for coastal
forecasting is the ocean colour monitor (OCM),
which can play an important role in coastal fishery and ecology
forecasts. In addition, there are other satellites, operated by
other countries, whose data are available in real-time or
near-real-time. For an ocean forecasting system, the most important
amongst these satellites provide SST (for example, AVHRR and TMI),
sea surface elevation (Jason),
surface winds (QuikSCAT),
and rainfall (TRMM).
The
observational platforms listed above provide a reasonable starting
point for assembling a credible observing system for coastal ocean
forecasting.
Dissemination
At present, INCOIS maintains a website
for dissemination of its forecasts for potential
fishery zones, tidal
currents in the Gulf of Khambhat and Bombay High region, swells in the Indian Ocean, and ocean
surface wind wave field forecasts. The Indian
tsunami warning system is also housed at INCOIS. In addition,
INCOIS has worked on other means such as electronic bulletin boards
to disseminate the forecasts. The INCOIS website is the only Indian
site at present that keeps ocean-related forecasts.
Feedback
There
have been some attempts at providing to INCOIS feedback on its forecasts of potential fishery zones. However, as of now, our
forecasting effort is limited and so are the mechanisms for
generating well-thought-out feedback to the forecasts.
Summary
The elements of the coastal forecasting system that existed in Europe
in the 1990s are similar to the state-of-the-art in India today. The
US state-of-the-art at the turn of the century provides a viable goal
for the Indian forecasting effort: it provides a perspective on where
we might want to be in the future (Shetye and Radhakrishnan, 2004).
Back |
