Flood Forecasting and Warning in Bangladesh

Bangladesh is located downstream of three large river basins: the Ganges, Brahmaputra and Meghna river basins. The total catchment area of these basins is 1.72 million km², with almost 93% of the catchment area situated outside the territories of Bangladesh – in Bhutan, China, India and Nepal. The topography, location and discharge from each of these three basins shape the annual hydrological cycle of the country. ^[1]

Over the course of a year, Bangladesh experiences periods of extreme water availability – too much and too little water. Monsoon precipitation from June to September is the main source of water, and the country has less water available outside of this season, termed the “dry period.” Heavy rainfall during the monsoon period is the main cause of flooding; this occurs almost every year, with a devastating flood every 5–8 years (FFWC, 2004). Such flooding causes severe damage to agriculture and infrastructure and the loss of human lives.

Bangladesh has implemented flood control and drainage projects since the 1960s. However, structural measures alone cannot totally protect the people and infrastructure from floods. Complete flood control in a country like Bangladesh is neither possible nor feasible. With this understanding, Bangladesh started developing flood forecasting and warning systems (non-structural measures) for flood management (Bhuiyan, 2006). The objectives were to enable and persuade people, communities, agencies and organizations to be prepared for floods and take action to increase safety and reduce damage. The goal was to alert people on the eve of a flood event.

 

Developing flood forecasting services

Bangladesh Water Development Board (the Board) is responsible for flood management through structural and non-structural measures. It also provides hydrological services in Bangladesh. As part of non-structural measures, the Board has been providing flood forecasting and warning services through its Flood Forecasting and Warning Centre (FFWC), established in 1972. Since then, the development of flood forecasting and warning services has made stepwise progress, which can be divided into three stages.

Initial stage (1972–1988) Initially, 11-gauge points were used for real-time flood monitoring and forecast purposes. In this early phase, gauge-to-gauge statistical correlation and Muskingum–Cunge methods were used for predicting water levels. In 1981, WMO and the United Nations Development Programme provided technical assistance for computerization of the hydrological database. Computer programs were also developed to carry out operations that had previously been performed manually. During devastating floods in 1987 and 1988, flood forecasts of the major river systems proved to be fairly accurate.

Second stage (1989–1999) After the 1987 and 1988 floods, an initiative was launched to develop a flood forecasting system based on a numerical model. WMO engaged the Danish Hydraulic Institute (DHI) to create a flood forecasting model for Bangladesh. During 1989–1991, the national flood forecasting model was developed using a MIKE 11 modelling system. From 1991, additional deterministic flood forecasting efforts were pursued, resulting in forecast lead times being increased to 48 hours. The number of real-time forecasting stations was increased to 16. From 1995 to 1999, the flood forecasting model was further upgraded to improve its forecast accuracy, under the Bangladesh Flood Action Plan. A geographic information system (GIS) module was added to the flood forecasting model, and the number of stations used to support forecast modelling was increased from 16 to 30.

Bangladesh again experienced severe flooding in 1998, for which the flood forecasting and warning services yielded productive and successful results. An internal analysis of the 1998 flood concluded that flood forecasting and warning services should be extended to all flood-prone areas of the country. In addition, the need for dissemination of flood information to vulnerable communities became very evident.

Third stage (2000 to date) Many lessons were learned from the 1998 floods. Foremost was that the people of vulnerable communities require flood information with a greater lead time. Further, they wish to know when their homesteads are going to be inundated and for how long. This showed that people were demanding area-specific flood forecasts. Moreover, field-level flood and water-related disaster managers also expressed their eagerness to receive timely flood forecasting information. In this third stage, FFWC received support to improve the accuracy and extend the lead time of flood forecasts, expand the provision of flood forecasting services to all flood-prone areas of the country, improve flood information dissemination at the vulnerable community level and build a sustainable institution.

FFWC efforts focused on improving the forecast lead time. It started to use ensemble precipitation forecasts from the European Centre for Medium-Range Weather Forecasts to provide medium-range flood forecasts. Since 2004, FFWC has provided deterministic flood forecasts to 3 days and medium-range probabilistic forecasts to 10 days. FFWC also started to develop its basin model in 2012.

 

Development of the basin model

WRF model rainfall forecast Source: Bangladesh Meteorological Department, Bangladesh

The concept for the basin model was introduced under the Comprehensive Disaster Management Programme Phase-II to increase forecast lead time. As Bangladesh is located downstream of three big river basins, an integrated basin model was needed to effectively increase the forecast lead time for Bangladesh. Fundamental to this was using the advances that have been made in numerical weather modelling and ensemble forecasting.

FFWC uses the Weather Research Forecast (WRF) model for precipitation forecasting. A typical WRF over the region and the three basins is shown below. The basin model, which is currently used for flood forecasting purposes in Bangladesh, uses quantitative precipitation WRFs for establishing a deterministic flood forecast with a lead time extended from 3 to 5 days.

 

Flood forecasting and warning activities

Flood forecasting and warning activities run from April to October every year in Bangladesh. In this period, the field-level hydrological measurements division works closely with the flood forecasting centre to provide observed data. FFWC remains open 24 hours a day, 7 days a week during this period.

 

Data collection and transmission

Today, the hydrology division of the Board has an extensive network of 60 rain gauges and 90 hydrological stations where water level, discharge, sediment or water quality are measured. Network design reflects the need for field data based on requirements of the flood forecasting model. Daily operational requirements of the flood forecast model are for real-time water level and rainfall data. Water level gauge readers for the 90 stations send data to FFWC twice daily. Data are usually collected from 6 a.m. to 6 p.m. at 3 hourly intervals every day. Rainfall records are available for 24 hour periods for the 60 gauges all over the country.

Data are now transmitted from the field using a mobile SMS system (see below). Prior to this development, hydrological data had been orally transmitted using landline telephones. The Board piloted automatic collection of water level data using a radar level sensor as part of another project^[1].

 

Operation of the flood forecast model

The flood forecast basin model is based on the DHI MIKE 11 hydrodynamic modelling system. The computational core of the hydrological forecasting system is the DHI MIKE 11 software, which contains two modelling components: (i) a hydrodynamic model and (ii) a hydrological model (NAM; a rainfall-runoff model). The hydrodynamic module contains an implicit finite-difference computation of unsteady flows in the rivers based on St Venant equations. The flood forecasting model is customized with the Flood Watch database, which uses a GIS. The MIKE GIS module is also integrated with the digital elevation model (DEM) of Bangladesh to generate an inundation model.

Quality checked, processed data are used in the model to generate 5-day deterministic forecasts. The operational flood forecasting system is based on real-time data received from available stations in Bangladesh, relevant online data received from riparian countries (based on an existing data-sharing protocol), and quantitative precipitation forecasts from numerical weather prediction models provided by the Bangladesh Meteorological Department and the Indian Meteorological Department. FFWC also uses satellite-based observation data for flood forecasting purposes.