DOI: 10.

1007/s00267-004-0010-4

Consensus Seasonal Flood Forecasts and Warning

Response System (FFWRS): An Alternate for
Nonstructural Flood Management in Bangladesh
MD. RASHED CHOWDHURY* important component in a comprehensive and participatory
Bangladesh Water Development Board approach to seasonal flood management. A general con-
WAPDA Building (FFWC) sensus in producing seasonal forecasts can thus be
Dhaka 1000, Bangladesh achieved by enhancing the existing FFWRS. Therefore, the
primary objective of this paper is to revisit and modify the
ABSTRACT / Despite advances in short-range flood fore- framework of an ideal warning response system for issuance
casting and information dissemination systems in Bangla- of consensus seasonal flood forecasts in Bangladesh. The
desh, the present system is less than satisfactory. This is five-stage FFWRS—i) Flood forecasting, ii) Forecast inter-
because of short lead-time products, outdated dissemina- pretation and message formulation, iii) Warning preparation
tion networks, and lack of direct feedback from the end-user. and dissemination, iv) Responses, and v) Review and
One viable solution is to produce long-lead seasonal fore- analysis—has been modified. To apply the concept of con-
casts—the demand for which is significantly increasing in sensus forecast, a framework similar to that of the Southern
Bangladesh—and disseminate these products through the African Regional Climate Outlook Forum (SARCOF) has
appropriate channels. As observed in other regions, the been discussed. Finally, the need for a climate Outlook Fora
success of seasonal forecasts, in contrast to short-term has been emphasized for a comprehensive and participa-
forecast, depends on consensus among the participating tory approach to seasonal flood hazard management in
institutions. The Flood Forecasting and Warning Response Bangladesh.
System (henceforth, FFWRS) has been found to be an

One of the most important objectives of the Flood people/agencies by taking appropriate counteractions
Forecasting and Warning Response System (FFWRS) is before and during floods to save lives and property,
a better comprehension of variability of seasonal and reduce damage and disruption.
flooding and forecast products for community adjust- The Flood Forecasting and Warning Center (FFWC)
ments. Consequently, improved translation of this of the Bangladesh Water Development Board (BWDB)
information into products, and their coordination to is responsible for monitoring floods in unified and
various governments, nongovernments, and private multipurpose basis throughout the country. At present,
agencies is essential. Railway, water/sewerage authori- the FFWC produces short-range forecasts with (2–3
ties, industries, and the commercial enterprises need daysÕ lead time). Such forecasts entail several stages
advance forecasts for their flood preparedness. Fore- including the real-time rainfall and water level data
casts, either short- or long-range, remain underused if collection, meteorological forecasting, and boundary
these are not disseminated properly to the lowest estimation. The flood forecasting procedure involves
administrative tier (i.e., the village level of any flood- carrying simulation into forecast; therefore, boundary
vulnerable community). This objective can be achieved conditions (rainfall, water levels, and flow) must be
by implementing an integrated FFWRS. The issuance available to the forecast horizon. At present, regional
of an advance warning of the increasing flood risk daily rainfall (millimeters) data are the main input for
enables the FFWRS to help in the preparedness of boundary estimation. Typically, 72 hours (i.e., 24, 48,
and 72 hours) input from the day after the time of
forecast to the forecast horizon period is used (see
KEY WORDS: Flood; Seasonal forecast; Warning response; Outlook
Fora; Bangladesh
Chowdhury 2000 for a review).
As per mandate, the FFWC disseminates a daily
Published online May 27, 2005 flood bulletin through special messengers on an ur-
Present address: Pacific ENSO Applications Center, University of gent basis to the PresidentÕs Office, Prime MinisterÕs
Hawaii, 2525 Correa Road, HIG 350, Honolulu, HI 96822 Office, the Water Resources and Relief & Rehabilita-

Environmental Management Vol. 35, No. 6, pp. 716–725 ª 2005 Springer Science+Business Media, Inc.
 Flood Forecasts and Warning in Bangladesh 717

tion Ministries, Disaster Management Bureau, Army averaged period) (Sarah and others 2001). Because of
Headquarters, Public Information Department, most its long lead time, global nature, interannual time
important government departments, news agencies scale, and inherent lag relationships, this is highly
such as radio and television, nongovernment organi- recommended for advance flood forecasting in Ban-
zations, international relief organizations (i.e., Red gladesh (Chowdhury 2003, Chowdhury and Ward 2004,
Crescent/Cross Society), foreign embassies, field wire- Douglas and others 2001). The demand for seasonal to
less stations, and other places as directed by the gov- interannual forecasting has already been established in
ernment. a report detailing the need analysis for forecasting
Other media used for these disseminations are the tools and means by Environment and GIS Support
Internet, e-mail, fax, telephone and wireless, radio, and (EGIS) (a project under the Water Resources Planning
television. Many other agencies, both public and pri- Ministry). This report further stated the need for long-
vate, fall under this dissemination network; however, in term seasonal flood forecasts for use in flood man-
this case, the receivers are to take a proactive role to agement preparedness actions in Bangladesh (River-
collect this bulletin from the FFWC either by phone, side and EGIS 2000). Additionally, there is mounting
fax, e-mail, or by personal contact. Other than some evidence that seasonal forecasts can be successfully
informal feedback through the subdistrict officers of made for many parts of the world including Bangla-
the government agencies, this system, at present, does desh with a lead times of up to several months (Ward
not receive any other feedback, either formal or and Folland 1991, Hastenrath 1995, Chowdhury and
informal, from the end usersÕ community. Therefore, Ward 2004).
the existing FFWRS is somewhat dysfunctional because The focus of this paper is to develop an operational
of almost nonexistent opportunities for inputs from framework for FFWRS in Bangladesh. Each of the
the end users. The end usersÕ feedback is a prerequisite stages of this five-stage FFWRS is explored in detail and
for improving the FFWRS; therefore, some formal augmented on the basis of a comprehensive and par-
feedback mechanism is essential to understanding the ticipatory approach to flood hazard management in
usersÕ need and to estimate the error in forecasts and Bangladesh (section 5). The first stage of the inte-
warnings. grated FFWRS (i.e., the seasonal forecast preparation)
As mentioned, the rise/fall of the water level with is not the subject of inquiry here; this is reported
respect to the danger level of the main rivers is the elsewhere (Chowdhury 2003, Chowdhury and Ward
basis of the existing warning message preparation in 2004). The other pertinent area of discussion is to
the FFWC. Because of the state of weather, it is difficult promote a dialogue between the international climate
to predict it after 6 days and almost impossible to science community and forecast users for developing a
predict it in 10–14 days because of nonlinearity within method for formulation of consensus of forecasts, and
the climate system and the growth of numerical fore- also aims to address the gaps in training and technical
casting model errors over time (Folland and Wood- capabilities.
cock, 1986, Murphy and Palmer 1986, Pierce and This paper is limited to the interpretation and dis-
others 1997, Washington and Downing 1999). In semination of the consensus seasonal flood forecasts.
addition to this major limitation of short lead time, the For this purpose, recent developments and experi-
present deterministic type of forecasts (i.e., ‘‘the river ences from other countries actively seeking to establish
is likely to rise/fall by 20 cm tomorrow’’) is not or improve seasonal forecasts and improved warning
understandable, and also not equally accessible to all response system are consulted. In addition, a model
users. For example, it is very difficult to interpret the similar to SARCOF (the Southern Africa Regional Cli-
meaning of water-level rise by 6.0 cm (What does it mate Outlook Forum) has been proposed here.
mean to him or her?) by someone whose village is lo-
cated far away from the major river such as the Ganges.
Consensus Forecast: An Example of SARCOF
Therefore, the present flood forecast messages are
insufficient for response and preparedness action by Several international initiatives are in progress to
vulnerable communities. Consequently, improved develop models with stronger orientation to users.
translation of this information to the village commu- There are areas in the world, mainly the tropics, for
nity, and also a better comprehension of seasonal cli- which forecasting is more reliable and consequently
matic variability and change, are needed. seasonal forecasts are already considered to be rela-
Seasonal forecasting is the outcome of a shift from tively reliable for practical use. In Brazil, seasonal
deterministic predictions to probabilistic forecasting forecasts are also sufficiently reliable to enable farmers
schemes (i.e., above/below a mean state over a time- to plan for the anticipated drought. Lessons can be
718 R. Chowdhury

learned for South Asia by examining forecast strategies provided by international and national experts
applied in above-mentioned areas. Globally, the re- through technical sessions at the Forum meetings.
search on seasonal products is increasing while the Although initially experimental, the SARCOF process
research related to seasonal forecasting in Bangladesh is continuing each year and has been adopted
is just beginning. Some of the recent findings show a worldwide as a model for seasonal forecasting. Re-
great potential for seasonal flood forecast in Bangla- gional Climate Outlook Fora are currently operating
desh (Hossain and others 2001, Chowdhury 1994, for other regions including south and Central
Douglas and others 2001, Chowdhury 2003, Chowdh- America, other parts of Africa, and Southeast Asia
ury and Ward, 2004). (Harrison 2000). Products from a number of RCOFs
The success of seasonal forecasts requires a closer are available online (http://www.cpc.ncep.noaa.gov/
cooperation between the forecast providers and the products/african_desk/rain_guidance/).
forecast users. One example of a project designed to It is envisioned that the model structure of SARCOF
bring together forecasters and users is entitled is an applicable tool to bring the forecasters and users
SARCOF (the Southern Africa Regional Climate together in Bangladesh. Moreover, the SARCOF model
Outlook Forum), a pilot scheme in which many of handles climate information (wet/dry), and flood is
the issues of forecast creation and provisions for essentially a component of the total climate system.
estimation of benefits are being examined. The Here in Bangladesh, the focus is on the combined ef-
SARCOF process first operated before the 1997– fect of rainfall and runoff. Additionally, the socioeco-
1998 wet seasons in southern Africa. The process nomic setting in Bangladesh, which is presumably
consists of an annual program of three meetings, agrarian, makes Bangladesh a potential candidate
with the first meeting held in September before the country that can benefit from the experience of the
onset of the wet season. The principal participants SARCOF model.
to the Forum are delegations from the National
Meteorological and Hydrological Services (NMHS)
Conceptual Perspectives of the FFWRS
of the Southern Africa Development Community
(SADC) region, which have the ultimate responsi- The basic concepts here are similar to those of
bility to disseminate the forecast product to their other existing FFWRS (see Parker, 2003 and refer-
user communities. Within a workshop format, tech- ences therein), and are being applied to real-world
nical and discussion sessions are held on global and problems in Bangladesh. The dissimilarity lies in the
regional climate dynamics, forecast methodologies, fact that long-range seasonal products are being ex-
and seasonal forecast presentations. Based on the pressed in a broader aspect of consensus forecast. In
predictions provided by each of the forecast groups contrast to short-term forecasts, the seasonal forecasts
for different homogeneous zones of the region, a raise issues of consensus. (Figure 1) portrays the
consensus forecast is achieved through discussions existing FFWRS as a five-stage process for reducing
among the participating climate community. For the impacts of floods, which begins with detection of
each zone, rainfall forecasts for the coming season the climatological conditions leading to flooding. An
are expressed as probabilities of occurrence of important feature of FFWRS is that it spans the
rainfall in three tercile classes: average, below aver- technical disciplines of meteorology, hydrology and
age, and above average rainfall. This probabilistic hydraulic engineering, and the human behavioral,
product is then distributed to users. In December, a sociological, and organizational disciplines. Some-
midseason correction is held. Again, through a times (as in A in Figure 1) these disciplines are not
process of consensus, an assessment of the early successfully brought together and integrated as they
season forecast is made. This assists in the update of need to be (as in B in Figure 1), and this is one of the
Forum forecasts for the remaining main wet season fundamental causes of ineffectiveness or partial
period. Finally, a postseason meeting is held in April effectiveness of the FFWRS. FFWRS needs to be de-
or May in which extensive validation of the Forum signed by multidisciplinary teams, and the design
product is conducted and the process is debated process requires cross-agency integration and under-
within the context of user feedback. This allows the standing. Experience indicates that effective FFWRSs
aims and methods of the process to be related di- are rarely designed and implemented without a
rectly to the requirements of a range of users. An thorough learning process (stage 5 in Figure 1) in
important component of the process is the capacity which the system is progressively refined on the basis
building exercises in which training on climate sci- of monitoring of performance and lessons learned
ence and seasonal forecasting methodologies is from operational events (see Parker 2003 for details).
 Flood Forecasts and Warning in Bangladesh 719

Figure 1. Stages and characteristics of FFWRS.

There has been a tendency in many regions of the 2003 for rainfall and flooding patterns in Bangla-
world (including Bangladesh) to focus attention and desh).
investment on stages 1 and 2 of the process, i.e., Bangladesh experiences flash floods, river floods,
detection and forecasting. Until recently, compara- and tidal floods along with serious river erosion that
tively little attention has been paid to stages 3, 4, and 5, causes thousands of families to become homeless and
and comparatively little investment has been put into landless each year. More than 90% of the flow drains
these stages. The consequence has been that where a out through three major rivers—the Brahmaputra,
capability to detect and forecast floods has developed, Ganges, and Meghna, and its branches spilling over in
an equivalent capability has not developed to formu- and around the country—and flows to the sea through
late appropriate warning messages, to disseminate Lower Meghna estuary, the confluence of the three
them in a timely manner, and to respond to such major rivers (BWDB and others 1989). Occasionally,
warning appropriately. local rainfall aggravates the total flooding situation in
the country, but in generally flood peaks are essentially
determined by the confluence of three major rivers.
Description of Study Area (Bangladesh)
Bangladesh occupies an area of about 145,000 km2 Five Stages of FFWRS: Bangladesh
between latitudes 20–30¢N and 26–45¢N, and longi-
Perspectives
tudes 88–0¢E and 92–45¢E (Figure 2). It is bounded
on the west, north, and east by India, on the Southeast Decision-makers worldwide face a difficult challenge
corner by Myanmar, and on the south by the Bay of in developing an effective response to the threat of
Bengal. water-induced disasters. In Bangladesh, this task is
Bangladesh has a tropical monsoon-type climate more complex, not only because of the multitude of
with a hot and rainy summer and a pronounced dry such disasters, but also because of the countryÕs vul-
season in the cooler months. January is the coolest nerable geography. Major parts of Bangladesh occupy
month of the year, with temperatures averaging near the deltaic plain of the Ganges–Brahmaputra–Meghna
17–19C, and July is the warmest month, with tem- river system. Bangladesh is criss-crossed by a labyrinth
peratures ranging 28–29C. The climate is one of the of rivers. Because the country is located at the lowest
wettest in the world; most places receive more than reach of the fluvial system, and in the area affected by
1,525 mm (60 in) of rain a year, and areas near the coastal surges, flood disasters of different types are
hills receive 5,080 mm (200 in). Most rain falls during frequent, including river, urban, and tidal or storm
the monsoon (June-September) and little during the surge. It has been observed that the technical compo-
dry season (November-February) (see Chowdhury, nents of the FFWRS are relatively better developed in
720 R. Chowdhury

Figure 2. Bangladesh and its physiographic settings.

Bangladesh, but the nontechnical (soft) components

need more attention that can only be achieved by clear
understandings and cooperation between various
agencies. An operational framework for an ideally
suitable five-stage FFWRS for Bangladesh, based on
seasonal outlook, is described in the following sections.

Stage 1: Forecast Preparation

On the basis of previous and current meteorologi-
cal, oceanic, and hydrological conditions, this stage
predicts the occurrence and magnitude of a flood in
advance.
Systematic efforts to bring together the scientific
outputs and usersÕ needs can increase the utility of
seasonal forecasts (Stern and Easterling 1999).
Therefore, preparation and presentation of seasonal
forecast information is the first stage in addressing
how applications can be maximized (Figure 3, Stage
1). As mentioned before, the South African Regional
Climate Outlook Fora (SARCOF) (initiated by the
NOAA Office of Global Programs (OGP)) is one of
the mechanisms through which the Climate Infor-
mation and Prediction Services (CLIPS), as repre-
sentatives of national and regional meteorological
services, and the usersÕ community can be brought
Figure 3. Flood Forecasting and Warning Response Sys-
together in the formulation of a consensus forecast tem—Bangladesh case.
for the region each year (Harrison 2000). Such a
formal mechanism (the Fora) then promotes a dia-
logue between the international climate science Similar to the process of SARCOF, the probabilistic
community and forecast users, and then develops a flood forecasting approach in Bangladesh may involve
method for consensus forecasts. three program meetings, with the first meeting being
 Flood Forecasts and Warning in Bangladesh 721

held in early April before the onset of the monsoon cultural adjustments by early irrigation) are essential
season (July–August–September). Within a workshop elements to attract the attention and consideration of
format, technical and discussion sessions can be held vulnerable communities to flood warning (Figure 3,
on global and regional climate dynamics, forecast Stage 2). The temporal generality that characterizes
methodologies, and seasonal forecast presentations. seasonal forecast information may not be specific en-
Such a coordinated effort can be achieved through ough for some stakeholders. Water managers may be
data exchange and close cooperation among the par- able to use the forecasts effectively without knowing
ticipating local institutions [e.g., Space Research and the timing and distribution of rain throughout the
Remote Sensing Organization (SPARRSO), Bangla- season. Farmers, on the other hand, may require more
desh Meteorological Department (BMD), FFWC of the accurate forecasts on seasonal scales, for example, the
BWDB, regional institutions [e.g., Indian Metrological seasonal distribution and the onset of floods. This
Department (IMD), India; International Center for stage, therefore, outlines how this information may be
Integrated Mountain Development (ICIMOD), Nepal] used most effectively to minimize risk, with emphasis
and those of existing international forecast centers on flood or drought in the agricultural sector.
[e.g., International Research Institute for Climate In Bangladesh, there is a demand for information
Prediction (IRI), USA; European Centre for Medium- on the height, time, and duration of forecasted floods
Range Weather Forecasts (ECMWF)]. It is then possi- in specific areas. At present, it is very difficult to gen-
ble to achieve a consensus forecast (for example, the erate this kind of information with long lead time be-
probability of strong El Niño is above average in this cause of the nonlinearity within the climate system.
year) through discussions among these participating However, seasonal forecasting provides a basis for ear-
climate communities, and expressed as probabilities of lier warning on a probabilistic forecasting scheme (i.e.,
flood occurrence in three tercile classes: average, below above/below a mean state over a time-averaged peri-
average, and above average. od). This earlier warning is important to individual
This probabilistic estimate could then be distributed decision-making in many sectors, including the agri-
to the main users only (i.e., government agencies, re- cultural sector, which uses this information to set up
search institutions, etc.). Here in this connection, the harvesting time of two major crops, aus and jute.
participating local institutions can constitute an Normally, the harvesting time of Aus and Jute is in late
‘‘Outlook Fora’’ under the administrative control of May and early June; any variation in flooding onset/
the Ministry of Water Resources, Bangladesh; the recession causes serious impact on the cropping sea-
FFWC, being an agency of the same ministry, may take sons in Bangladesh. If the onset of flooding period is
the lead role to organize meetings/seminars, and so earlier than normal, the harvesting of aus and jute is
on. delayed, which is damaging to crops. On the other
Through a process of consensus with the same hand, if the onset of shallow flooding (i.e., depth is 50
participants of the first meeting, a midseason correc- cm below the danger level, which occurs during the
tion meeting can follow in early July to update the months of April to May and submerges low-lying lands
existing forecast for the monsoon season (July–August– only) is late, then the Broadcast Aman crops cannot
September). This product is then delivered to every thrive well in May/June because of scarcity of water.
possible section of society, including the village com- Similarly, if the flood recession is late and the duration
munity. of flood peak is above normal, then it overtops the
highland Transplanted Aman crops, causing severe
Stage 2: Interpretation and Stakeholders damage to the vegetation and flowering stage of
This stage identifies the probable impacts of flood Transplanted Aman crops (Rasid and Paul, 1987).
upon vulnerable communities, to warn people about
the onset of flood and its probable impacts. Stage 3: Process of Dissemination
A broad-based development approach with a clear This stage communicates and distributes the warn-
understanding of characteristics of the target audience ing messages to disaster management agencies and
at risk leads to increase confidence in the system. Clear vulnerable communities.
and area-specific information indicating what is likely If the time lag between the development and dis-
to happen (for example, a strong El Niño (like 1997) is semination of seasonal forecasts could be decreased,
likely to develop this year), what it means to the people the seasonal forecast would be more usable. By pro-
(for example, this advance El Niño information means viding information when needed, decisions such as the
that the probability for a dry summer is above average), type and timing of crops to be planted could be altered
and what can they do (for example, they can try agri- to reflect the uncertainty in the forecasted climate.
722 R. Chowdhury

Timing is also imperative to aid agencies, which re- Once the information reaches the local authority, it
quire about 4 months to deliver food to an area (e.g., is their responsibility to contact the FRG, which can be
after a famine early warning system confirms need for formed by the representatives from the village com-
aid (Farmer 1997)). munity (see Chowdhury 2000 and references therein).
The timely process of dissemination has great po- In Bangladesh, for example, the most likely represen-
tential for adaptation of the forecasts (Figure 3, Stage tatives are those who are relatively educated and have
3) in Bangladesh. Presently, short lead time determin- better accessibility to the rural masses, for example,
istic forecasts of the FFWC are disseminated through a primary school teacher, chief of religious institutions
top–down approach (as mentioned earlier) and there is (i.e., imams of mosques), village leaders, and so on. At
no direct feedback from the usersÕ community. The this time, the local authority together with the FRG can
FFWC disseminates the daily flood bulletin by several generate community-based additional special informa-
means including their official Web site (http:// tion. For example, any local message advising the
www.ffwc.net). The main problem of this dissemination farmers to delay harvesting for aus crop for a period of
process is that the product does not reach the target 3 weeks from normal in any El Niño year would be
group in a timely manner, partly because of short lead relevant and useful for the community. The FRG will
time of the forecasts and, added to this, the presence of then propagate these messages to the villagers by
an outdated network. Therefore, forecasts remain megaphone (public announcement), drum beating,
inaccessible to the target users for which they were in- poster, drama, documentary, local training, and indi-
tended. In contrast, seasonal forecasts provide the vidual contacts. Additional training programs may be
opportunity to convey information with long lead time essential for the FRG/villagers to clearly understand
(1 to 3 months in advance). This long lead time allows the meaning of probabilistic forecasts and the under-
seasonal forecasts to be of wider and timely use, even if lying risk.
the dissemination network is outdated (as in Bangla-
desh). Therefore, the products of seasonal forecasts Stage 4: Response
have greater potential for dissemination in Bangladesh. This stage generates actions by concerned agencies
This is the stage where two full scales of dissemina- and threatened communities for protection against
tions takes place; the first is from the ‘‘Outlook Fora’’ flood hazards in response to warning.
to the level of local authorities/subdistricts, and the The source of a forecast can elicit various responses;
second is from the local authorities to the village level. therefore, a response-oriented approach could report
For example, the ‘‘Outlook Fora,’’ at first, disseminates response options for different sectors. Such options
area-specific national flood warnings (for example, should not be prescriptions; responsibility for action
information such as the probabilities of strong El Niño must be retained by the end users. The information
and the likelihood of a dry summer) to the local should be communicated in a way that allows it to be-
authorities via messengers, bulletin, letter, fax, phone, come the property of the user. Risk is an inherent part
e-mail, and seminars. The local authority then dis- of decision making; seasonal forecasting must be por-
seminates these area-specific flood warnings (for trayed in this context. Although the information may
example, the requirements for additional irrigation) to be of limited utility until forecast skill and timing can
the flood response group (FRG)/villagers via mega- be improved, it is important that users be given the
phone, drum beating, poster, drama, documentary, opportunity to understand the information. This allows
training, and individual contacts. individuals to decide for themselves whether to take
Even choosing the existing dissemination channels the risk of using the forecast information to improve
of the FFWC (i.e., messenger, fax, telephone, e-mail, their management plans.
and personal contact) and deciding on the time and At present, the FFWC neither organizes any feed-
frequency of issue ahead of time, the FFWC can sig- back process nor does it have any role in organizing
nificantly improve the dissemination of the forecasts to any coordination meeting among the institutions.
the local level because of long lead time of the forecast Thus, the feedback part remains largely ignored. In
products. Additionally, the national television and local Bangladesh, an effective feedback mechanism can be
radios can play a significant role in the dissemination implemented by organizing government/nongovern-
process of the seasonal products. Televised weather ment agencies working at the field and the FRG (also
forecasts carrying warning code icons of agreed ‘‘sea- see Chowdhury 2000 and references therein). Because
sonal flood outlook’’ and displaying the digital eleva- there are various government and nongovernment
tion flood map would be an essential instrument for agencies working in a variety of fields, the efforts re-
this purpose. quired are largely towards constant monitoring and
 Flood Forecasts and Warning in Bangladesh 723

periodic coordination of meetings and seminars. At to 5 (advanced). The usersÕ community (FRG, as ex-
present, the FFWC, under the overall guidance of the plained in stage 4) will evaluate locally, on the basis of
‘‘Outlook Fora,’’ can organize similar coordination CDM, and then communicate their findings to the
meetings to enhance the utility of flood forecasts in ‘‘Outlook Fora’’ via local authority for further action in
Bangladesh. improving the forecasts and warning dissemination. It
At the local level, the FRG will be responsible to is, however, a matter of fact that the ‘‘Outlook Fora’’
evaluate a) the accuracy and promptness of forecasts needs to pay significant attention towards development
(for example, the FRG will evaluate whether the sum- of a standard CDM for evaluation of seasonal flood
mer was really dry as per prediction), b) public forecasting and dissemination in Bangladesh.
awareness and response to warning messages (for Finally, an extensive validation of the Forum prod-
example, the FRG will evaluate whether the people uct is conducted and the process is debated within the
were well informed in advance about the dry episodes), context of user feedback at a postseason meeting to be
and finally c) close interaction with community resi- held each year in October. Such consideration will
dents through the use of questionnaire, interviews, and allow the process aims and methods to be related
contacts (Figure 3, stage 4). On evaluation, the FRG directly to the requirements of a diverse group of users.
will communicate their findings to the local authority
and from there it will gradually go up to the ‘‘Outlook
Conclusions
Fora’’ for further action in improving the forecasts and
warning dissemination. Sound structural measures to effectively mitigate
damages would take a long time and huge investments
Stage 5: Review and Analysis for implementation. In the meantime, nonstructural
This stage continuously monitors the performance measures can be particularly effective in mitigating the
of various components of the FFWRS for possible damaging effects of floods in Bangladesh. The flood
improvement. forecasting and warning are primary means to achieve
At present, other than some performance evalua- such measures in a highly cost-effective manner. For
tion of short-range forecast products of the FFWC practical purposes, this information is useful for natu-
(Chowdhury 2000), the performance of the flood ral hazard assessment and mitigation activities within
warning dissemination and warning response elements Bangladesh because the current short-term forecasting
in Bangladesh is virtually nonexistent. Despite ad- is not adequate to plan for extreme events in large river
vances in flood forecasting, FFWRS often underper- basins and to facilitate basin-wide planning and man-
forms because warning dissemination and response are agement of water resources in Bangladesh. It is desir-
unsatisfactory. It is envisioned that once ‘‘criteria’’ for able to have a forecasting mechanism involving lead
a uniform evaluation methodology are developed, time of several months in advance for desirable far-
problems may be highlighted, knowledge can be reaching economic ramifications.
pooled, and best practice can be transferred to the In order to enhance the existing FFWRS, Bangla-
forecasters and users groups. desh currently faces two types of problems: i) seasonal
The joint consultative committee of the ‘‘Outlook flood forecasts, the research of which has just begun;
Fora’’ with other government agencies (i.e., Ministry of and ii) warning dissemination and response, the re-
Water Resources, Disasters Management Bureau, and search of which has yet to start. For the forecast part,
Planning Commission) can monitor this performance Bangladesh has to rely on the latest ideas and tech-
procedure by undertaking regular sample surveys from nology from developed countries. Regular exchange of
the FRG, local authorities, and other participating data and close cooperation between the local institu-
agencies (Figure 3, stage 5). The ‘‘Criteria-develop- tions (e.g., SPARRSO, BMD, Flood Forecasting and
ment Matrix’’ (CDM) may be applied to evaluate the Warning Center of the BWDB) and those of existing
performances. The Criteria-development Matrix com- regional and international forecast centers would be
prises social, organizational, and institutional criteria essential. For the dissemination part, the model struc-
by which a FFWRS may be assessed using a five-stage ture of SARCOF as described and discussed in this
development scale. The concept of CDM has been paper in the context of Bangladesh could be an
successfully applied in the European Union to evaluate applicable tool that can bring the forecasters and users
the short-to-medium-range FFWRS system (see Parker together. The advantage of this model is that it in-
and Fordham 1996, Parker 1999, Quarentelli 1997). creases the skill of forecast for a better comprehension
The level of development of each criterion is gauged of seasonal flooding variability and change through
on a five-point categorical scale from 1 (rudimentary) providing significant opportunities to synthesize the
724 R. Chowdhury

regional climate dynamics ahead of time by virtue of Chowdhury, A. M. 1994. Bangladesh floods, cyclones and
meetings among the local, regional, and international ENSO. Paper presented at the International Conference
on Monsoon Variability and Prediction. International
institutions.
Center for Theoretical Physics (ICTP). 9–13 May 1994.
It is very important that there be feedback between Italy.
forecasters and users so that forecasters focus on pro- Chowdhury, M. R. 2000. An assessment of flood forecasting
viding what users need. Forecasters must explain what in Bangladesh: the experience of the 1998 flood. Natural
might be possible, and users must take on the Hazards 22:139–163.
responsibility of providing definitions of adverse Chowdhury, M. R. 2003. The El Niño–Southern Oscillation
weather, and appropriate values for costs of treatments (ENSO) and seasonal flooding: Bangladesh. Theoretical and
and losses. The process of feedback can be strength- Applied Climatology 76:105–124.
ened by the participation of the FRG, who will peri- Chowdhury, M. R., and N. Ward. 2004. Hydro-meteorological
odically contact the local authorities with information variability in the Greater Ganges-Brahmaputra-Meghna
(GBM) basins. International Journal of Climatology 24:1495–
about the error of forecasts. This information will then 1508.
be transmitted to the ‘‘Outlook Fora’’ for corrections.
Douglas, W. W, S. A. Wasimi, and S. Islam. 2001. The El Niño–
It is in this sense that this is the best way to ensure Southern Oscillation and long-range forecasting of flows in
peopleÕs participation in the total decision-making the Ganges. International Journal of Climatology 21:77–87.
system. Farmer, G. 1997. What does the famine early warning com-
In conclusion, it is not only prudent but also munity need from the ENSO research community? Inter-
essential that the social aspects of FFWRS be thor- net Journal of African Studies (2) (Available at http://
oughly assessed. In some cases, it is necessary to rec- www.dir.ucar.edu/esig/ijas/ijasno2.farmer.html).
onceptualize FFWRS, with the involvement of effective Folland, C. K. and A. Woodcock. 1986. Experimental monthly
long-range forecasts for the United Kingtom. Part I: De-
multidisciplinary teams. There is no doubt that infor-
scription of the forecasting system. Metrological Magazine
mation and communication technology has opened up 115: 301–318.
new avenues for accessible and rapid flood warning Harrison, M. S. J. 2000. Seasonal to interannual climate pre-
dissemination in a manner that can be designed to diction and application—the CLIPS perspective. Preprints
address the needs of the at-risk community more than of AMS Sixth International Conference on Southern
ever before. The Bangladesh Government should, Hemisphere Meteorology and Oceanography. 3–7 April
therefore, strive to enhance the FFWRS for flood 2000. Santiago, Chile.
forecasting in Bangladesh. Hastenrath, S. 1995. Recent advances in tropical climate
prediction. Journal of Climate 8:1519–1532.
Hossain, E., S. S. Alam, K. H. Imam, and M. M. Hoque. 2001.
Bangladesh country case study: impacts and response to the
1997–98. Pages 44–50 in El Nino events, M. H. Glantz
Acknowledgments (eds.), Once burned twice shy. United Nations University
Grateful acknowledgment is made to the anony- Press.
mous referees for thoughtful comments and insights Murphy, J. M., and T. N. Palmero 1986. Experimental
regarding an earlier version of this paper. ReviewersÕ monthly long-range forecasts for the United Kingdom.
comments were particularly helpful to revise the paper Part II: A real-time long-range forecast by an ensemble of
numerical integrations. Metrological Magazine 115: 337–
in the present form. Special thanks are due to Ms. 349.
Nicole Colasacco of the Pacific ENSO Applications
Parker, D. 1999. Criteria for evaluation the condition of a
Center (PEAC, University of Hawaii) for her valuable tropical cyclone warning system. Disasters 23:193–216.
comments. Thanks are also due to the officials of the Parker, D. 2003. Designing flood forecasting, warning and
Flood Forecasting and Warning Center (FFWC) of the response systems from a societal perspective. Paper pre-
Bangladesh Water Development Board (BWDB) for sented at the International Conference on Alpine Meteo-
providing the author various research supports. rology and Meso-Alpine Programme. 21 May 2003. Brig,
Switzerland (Available at http://www.map.ethz.ch/
Icam2003/presentation/10.1/Brig_document.pdf).
Parker, D., and M. Fordham. 1996. An evaluation of flood
forecasting, warning and response systems in the European
Literature Cited Union. Water Resources Management 10:279–302.
BWDB (Bangladesh Water Development Board), French Pierce, C., M. Dukes, and G. Parker. 1997. Forecasting of the
Engineering Consortium, and Bets Consulting Company. British Isles weather. Pages 299–325 in M. Hulme, E. Bar-
1989. A joint report on pre-feasibility study for flood con- row (eds.), Climates of the British Isles: present, past and
trol in Bangladesh. Government of Bangladesh 2:2–6. future. Routledge, London and New York.
 Flood Forecasts and Warning in Bangladesh 725

Quarentelli, E. L. 1997. Ten criteria for evaluating the man- 2001. Seasonal forecasting for climate hazards: prospects
agement of community disasters. Disasters 21:39–54. and responses. Natural Hazards 23:171–196.
Rasid, H., and B. K. Paul. 1987. Flood problems in Bangla- Stern P., and W. Easterling (eds.). 1999. Making climate
desh: is there any indigenous solution? Environmental Forecasts Matter. National Academy Press, Washington
Management 11:155–173. D.C. 192.
Riverside Technology, Inc. USA, and the Environment Washington, R., and T. E. Downing. 1999. Seasonal forecast-
and GIS Support Project. 2000. Information for flood ing of African rainfall: prediction, responses and household
management in Bangladesh 1–2. CARE-Bangladesh 63– food security. The Geographical Journal 165:255–274.
64. Ward, M. N., and C. K. Folland. 1991. Prediction of seasonal
Sarah, M., R. Washington, T. Downing, R. Martin, G. Ziervo- rainfall in North Nordeste of Brazil using eigenvectors of
gel, A. Preston, M. Todd, R. Butterfield, and J. Briden. SST. International Journal of Climatology 11:711–744.