The importance of seasonal flood forecasts in Bangladesh
Bangladesh is essentially an agrarian country in a deltaic environment that relies on floods and rich soils to produce its annual grain requirement every year. However, early monsoon floods, late onset of floods, and other climate variations significantly impact food production and quality of life.
Recognition of the danger of unexpected flooding led to the development of short-range flood forecasts (24- to 72- hours in advance) more than a decade ago. While the role of short-range forecasts is important, they are insufficient for response and preparedness action by vulnerable communities, particularly in terms of dislocation of cropping practices.
These are forecasts of average seasonal conditions over a region that are made many months in advance due to slowly changing parts of the climate system. Analyzing ocean temperatures provides some ability to forecast average conditions for months in advance. If it is possible to create successful seasonal forecasts, then the generation of a probabilistic outlook of flooding well ahead of seasons (3-6 months before the occurrence of flooding) is also possible.
The beauty of seasonal forecasting techniques is that it enables advance forecasts at least 3-6 months in advance. In the case of short-range deterministic flood forecasts, we can generate information 3-7 days in advance and the probabilistic seasonal forecasts can produce information for about 3-6 months in advance with reasonable accuracy.
However, while the deterministic forecasts are mostly accurate, due to their shorter time scales (hours to days), the probabilistic forecasts have some uncertainties because of their longer time scales (3-6 months).
When reading a probabilistic flood forecast in three tercile format, eg, 40:40:20, it means the possibility (or probability) of higher-than-normal flooding is 40% (upper tercile), normal flooding is 40% (middle tercile), and lower-than-normal flooding is 20% (lower tercile).
Therefore, even if we forecast a 40% possibility for higher-than-normal flooding, there is still a 20% possibility for lower-than-normal flooding. Compared to deterministic forecasts, this is the kind of limitation probabilistic forecasts have.
However, while the deterministic method can produce forecasts only during the occurrence of an event -- eg, the “Flood Forecasting and Warning Centre” of the Bangladesh Water Development Board produces information on daily rise/fall in river water-level in a flooding season -- the probabilistic method can produce forecasts in a hot/dry spring on the possibility of a rainy summer and monsoon flooding.
The El Niño-Southern Oscillation (ENSO) climate cycle, which has two phases -- El Niño and La Niña -- has been widely used to develop these probabilistic seasonal forecasts. In Bangladesh, El Niño is associated with drought and La Niña with flooding.
La Niña refers to the appearance of colder-than-average sea surface temperatures (SSTs) in the central or eastern equatorial Pacific region, the opposite to conditions during El Niño. It is a cold event where the SSTs become anomalously colder compared to the long-term average for the central and eastern equatorial Pacific.
La Niña episodes also feature large-scale changes in the atmospheric winds across the tropical Pacific, including increased easterly (east-to-west) winds across the eastern Pacific in the lower atmosphere, and increased westerly (west-to-east) winds over the eastern tropical Pacific in the upper atmosphere.
These conditions reflect an enhanced strength of the equatorial Walker circulation. When the Walker circulation is strong, the upper-tropospheric winds in the Australasian region are easterly and, consequently, tropical disturbances are transported westward into the Bay of Bengal. Therefore, rainfall becomes very active in the region of the western Pacific, including Bangladesh.
The causal connection between La Niña and seasonal flooding in Bangladesh
There is evidence of teleconnections between La Niña’s strength and seasonal climate anomalies (eg, rainfall, flood, and cyclone) in Bangladesh, which normally faces a surplus of rainfall during La Niña years.
For example, all previous La Niña years (1964, 1973, 1988, and 1998) recorded excessive basin-wide rainfall. During any La Niña year, the trade wind strengthens, and as a result, rainfall increases significantly along the greater Ganges-Brahmaputra-Meghna (GBM) basins, causing flooding along the whole catchments.
This, in turn, severely floods Bangladesh, as it is the lowest riparian country in these basins. The stream-flow of the major rivers in Bangladesh is the result of monsoon rainfall in upstream India.
How helpful are the seasonal forecasts?
Although La Niña-based seasonal products are used widely and successfully for flood hazard management in one-quarter of the globe, the scientific research in Bangladesh related to seasonal products is just beginning.
The government and water experts in Bangladesh will decide how effectively they can use the products of seasonal forecasts. For example, the onset of this year’s La Niña was visible in March–April 2020. At that time, the seasonal SST forecast showed a cooling tendency (an indication of borderline La Niña) for seasons Jul-Aug-Sep, and Aug-Sep-Oct.
Currently, as the season advances, a trend of further cooling has been projected and a weak-to-moderate La Niña is likely to continue in Oct-Nov-Dec. This information, which was available by April 2020, could be utilized with reasonable accuracy for probabilistic flood forecasting during Jun-Jul-Aug of 2020 for Bangladesh.
To improve the forecast, the observed climate data for Bangladesh and the state of the science global datasets for other climate features could also be used. This information could help develop a strategy to address stakeholders’ needs through the flood response group.
Therefore, in addition to short-term deterministic forecasts, medium-to-long term seasonal forecasts are essential in developing a real-time response plan for hazards management. This would significantly enhance the agricultural decision support system in Bangladesh.
Dr Rashed Chowdhury is a Principal Research Scientist at Pacific ENSO Applications Climate Center Joint Institute for Marine and Atmospheric Research, the University of Hawaii at Manoa.