CFAB - Climate Applications in Bangladesh

Flood Forecasting
in a deltaic region such as Bangladesh, is a difficult problem. A large amount of data is required in order to initialize the hyrdological models.	Three factors guided the development of the CFAB flood forecasting schemes: • Hydrological data required to run models is difficult to obtain and much of it is sensitive and retained by different nations. • State-of-the-art hydrological models require a substantial investment in technology and man power. • Reliable and timely forecast are needed now.
for example, rainfall data over the entire catchment regions of the Ganges and Brahmaputra is required in addition to river flow data, soil and crop information and the amount of water that is retained within each catchment by agriculture and water resource use. Figure: Daily Ganges and Brahmaputra discharge at the boundaries of Bangladesh. The discharge is the primary predictand of CFAB.
Based on these factors, the generation of CFAB forecasts has got underway.
• The basic variables forecast are the discharges of the Brahmaputra and the Ganges into Bangladesh. • CFAB takes advantage of the short-term empirical forecast developed by the Flood Forecast and Warning Center (FWMC). • The CFABdepends on data from satellites from NOAA National Center for Environmental Protection (NOAA/NCEP) and model output from the European Center for Medium Range Weather Forecasts (ECMWF).	• Critical data sets are the discharge data at the boundaries of Bangladesh and the sea-level variability in the Bay of Bengal. • Forecasts are made on a range of times scales. • Models are as simple as possible and easily adaptable to the technology currently available in Bangladesh. • Models areadaptable for future improvement as science evolves. • Documentation will accompany the transfer of the forecast models to Bangladesh.

General Scheme	Forecasts for three main time scales:	• Short range (1-6 days): This enables the current Bangladesh forecasts to be extended to 8 days.
		• Medium range (20-25 days): Forecasts at these time scales enable significant agricultural adjustment and disaster mitigation programs to be enacted.
		• Long range (1-6 months): These forecasts provide sufficient lead-time for long term planning to be enacted.

Actual Forecast Modelling	Site	Description	Time Period	Scientists
	CU	short range EC model + empirical Probabilistic	5-19 days	Tom Hopson
	GT	medium range Observations + empirical Approaching probabilistic	~25 days	Carlos Hoyos
	ECMWF	long range NWP Probabilistic	2-3 months	Peter Webster, Bob Grossman

Short range (1-6 days):
The basic aim is to generate discharge forecasts at the boundaries of Bangladesh which will be coupled to the FFWC empirical flood forecasts of the intra-Bangladesh region. This combined scheme provides:
up to 8 days forecasts early decisions for flood mitigation early decisions for disaster management
These forecasts will benefit the agricultural sector. The farming community will be able to act or react to minimize loss and optimize gain, especially at planting and harvest times.
We have three forecasting techniques (A, B and C) devised for the short range forecasts.Method A is more complicated and has the capacity to enable risk management. Method B is a subset of Method A i.e. less complex but is readily transportable for near-time use in Bangladesh.
Method A :- This method uses the full forecasting products generated by The method uses the full forecasting products generated by NOAA/NCEP and ECMWF. Precipitation forecasts are integrated over each basin catchment. The ECMWF model uses ensemble techniques. Therefore, a probabilistic forecast of integrated rainfall is calculated. A total of 51 forecasts are integrated every 12 hours.	Method B :- This method uses a subsection of the ECMWF forecasts and a linear regression technique to generate the discharge forecast. The ensemble mean of the ECWMF forecasts is used to train a linear regression equation. Once the coefficients of the regression are calculated, the ensemble mean is extended for the 6-day period.	Method C :- This method is a more complex version of Method B.It uses an increasing number of linear regressions relating to ensemble members.
Sample Output - Method A
Advantages and disadvantages
possible to calculate the probability of various discharge intensities requires considerable computing power considerable technical training	relatively accurate forecasts only one time series is necessary inability to calculate probabilities	probabilistic forecasts are possible increase in computation time increase in amount of data transfer
Recommendation and Requirements
1) use Method B initially to gain experience, and to test the utility of the forecasts.	2) test and compare the discharge forecasts produced by these methods with those from the existing FFWC empirical models.	3) adopt either Method C or Method A

Medium range (20-25 days):
The forecast with maximum applicability to agriculture planning and water management is a forecast that is on the time scale of three to four weeks. This is referred to as the intraseasonal time scale. During a monsoon season, the precipitation goes through a series of low-frequency periods of wet and dry periods that are far greater in amplitude than the interannual variability of the monsoon. These medium range (or intraseasonal) forecasts can also provide sufficient forewarning so that resources can be marshaled for the mitigation of a disaster .
With field experiments and research investigations we have increased our appreciation and understanding of the intraseasonal variability of the monsoon. The CFAB research group has developed a new empirical method that provides relatively accurate forecasts of the week-to-week variability of the monsoon for several weeks in advance.
Samples of Output
		Figure :- 20-day forecast of rainfall over theGangetic Plain for the summer of 2002 using the new medium-range empirical forecast scheme. Black curve is the observed and red curve is the forecast made 20 days ahead.
Figure :- Forecast of the Brahmaputra discharge into Bangladesh for the 1997 summer. Blue curve is the observed and black curve is the forecast made for pentad periods.
Advantages and disadvantages
can be run on a small computer requires data for 8 predictors is readily adaptable for different predictands.

Long range (1-6 months):

Forecast on the time scales of 1-6 months provide an overview of expectations for climate. Currently, they are the least developed of the forecasting tools. These forecasts depend on the long-term forecasts prepared by ECMWF. The CFAB research group has tailored these forecasts to determination the water discharge into Bangladesh.

Scientific Basis

The scientific basis for long range climate forecasting rests in the assumption that the slowly varying surface features of the planet influence variations in the climate. Climate models attempt to capture the variability of the coupled ocean-atmosphere system such as El Nino, etc.

Sample of Output

Advantages and disadvantages

is relatively simple
only requires input from the ECMWF model
simple manipulation and statistical analysis

To accompany the forecasts, a risk analysis tool has been developed that combines the probablistic forcast with information that the user community provides. This tool is referred to as the "user metric"

The work described above is the result of the work of a number of people. In particular, the intermediate forecast scheme originates with Carlos Hoyos. Thomas Hopson performed most of the analysis of the short- and long-term forecasts with significant help from Hai-Ru Chang, Kamran Sahami and Jun Jian. Data was provided by ECMWF. Funding for these activities comes form USAID/OFDA. Science support was given by the Climate Dynamics Division of the US National Science Foundation.