The study revealed that Bangladesh climate -- despite weak quantitative correspondence between the strength of ENSO and the rainfall anomaly -- has particularly a strong relation when SOI (Southern Oscillation Index) extremes, indicate negative value of SOI to dry extremes and positive value of SOI to wet ones. In case of moderate anomaly in the SOI, the index-climate relation appeared to be contradictory and, in particular, Bangladesh is wet during the moderate El Ni-o years. Therefore, while the SOI-rainfall relation in the upstream Ganges-Brahmaputra-Meghna (GBM) basins in India shows strong casual connection indicating El Ni-o (both strong and moderate) to dry and La Ni-

a to wet, the same SOI-rainfall relationship offers limited applicability in the context of Bangladesh climate. So, understanding the broad-scale features of the Asian monsoon with especial emphasis to Bangladesh is very essential.

Asian monsoon is one of the important components of the coupled ocean-land-atmosphere system. Although in India considerable work has been done on variability of climate and particularly rainfall by analysis of the extensive data sets of the Indian Meteorological Department (IMD), the nature of climate variability over other parts of the sub-continental monsoon regime, such as Bangladesh, Sri Lanka, Nepal, and Pakistan, is not so well documented. Therefore, detailed investigation on the relationship between the Bangladeshi monsoon and monsoons of the other regions in Asia, in particular India, will enhance our understanding of the large-scale features of the monsoon in the greater Ganges-Brahmaputra-Meghna (GBM) basin system.

El Ni-o and La Ni-a years

Given that there are typical characteristics of El Ni-o and La Ni-a, how are specific 'ENSO events' defined? How large must the value of the index be, and for how long must it persist in order for an El Ni-o or La Ni-a to be identified as strong or moderate? Any definitive objective procedure for classifying intensity is yet to be explored. However, a common method used for this purpose is based on the Ni-o 3.4 SST index. In this method, an El Ni-o or La Ni-

a event is identified if the five-month running average of the Ni-o 3.4 index exceeds +0.4Êdeg. C (for El Ni-o) and 0.4 deg. C (for La Ni-

a) for at least six consecutive months. According to this multivariate ENSO index, seven major El Ni-o years are 1997-98, 1991-92, 1986-87, 1982-83, 1972-73, 1965-66 and 1957-58, and seven major La Ni-a years are 1998-99, 1988-89, 1975-76, 1973-74, 1970-71, 1964-65, 1954-55 and 1949-50. This ranking would even vary if based on an averaged Ni-o 3.4 index over different seasons. Further it was observed that the relative ranking of events would vary if the ranking were based on an index other than Ni-o 3.4. For example, the classification system in the Western Regional Climate Center's (WRCC) that is based on the average value of SOI for the months of June-November provides a different ranking of events. With this WRCC approach, the ENSO phase is determined by atmospheric quantities (SOI) (value of SOI = -1.0: Strong El Ni-o, SOI = -0.5:Moderate El Ni-o, SOI = + 0.5.: Moderate La Ni-a, and SOI = +1.0: Strong La Ni-a). Another classification that is based on cold (La Ni-a) and warm (El Ni-o) episodes is also available. This has been compiled to provide a season-by-season breakdown of cold and warm conditions in the tropical Pacific.

Based on this information we have classified five strong (or major) El Nino years -- 1951, 1958, 1972, 1982, and 1997, and five strong (or major) La Nina years -- 1964, 1973, 1975, 1988 and 1998.'

The years we have finally identified for moderate El Ni-o are 1963, 1965, 1969, 1974, and 1987, and for moderate La Ni-a, 1956, 1970, 1971, 1984 and 1999.

ENSO-rainfall: Bangladesh

It is evident that the seasonal-average rainfall in Bangladesh (ABR) shows a negative and general decreasing tendency during the strong El Ni-o events. Major El Ni-o years like 1951, 1958, 1972 recorded significant rainfall deficits: 38%, 48%, and 10% in these years. The same time-series showed that during the moderate El Ni-o years (1963, 1965, 1969, 1974, and 1987), there was a tendency of excessive rainfall -- other than 1963, the fall increased by +18% in 1965, +16% in 1969, +7% in 1974, and +34% in 1987. Similarly, except 1975, all the major (strong) La Ni-a years (1964, 1973, 1988 and 1998) recorded excessive rainfall (1975 appears to be the year of onset of El Ni-o). Recorded increases were +4% in 1964, +8% in 1973, +30% in 1988, and +10% in 1998. Like the strong La Ni-a years, the moderate La Ni-a years (1956, 1970, 1971, 1984, and 1999) recorded excessive rainfall. The increases were 1.7% in 1956, 11% in 1970, 5% in 1971, 18% in 1984, and 12% in 1999.

Evaluating from the composite picture of events-response relationship in these years, some interesting findings relating to SOI and seasonal rainfall deviation appeared. Strong El Ni-o years (1951, 1958, 1972, 1982, and 1997) provided negative values of SOI and the average deficits of seasonal rainfall were --19% and --16% for the Ganges and Brahmaputra basins. Meghna basin did not show any significant variation. The deficit was found to be significant for most of the successive monsoon months (i.e., June, July, August, and September). Findings revealed that a negative value of SOI for the months of JJAS (-1.9) was strongly correlated to rainfall deficit and drought in Bangladesh. During these El Ni-o years, negative values of SOIs started appearing from the previous season FMAM (-0.4), and the May minus March SOI was negative (1.5). While considering the moderate El Ni-o years, the negative values of SOIs appeared as usual, but values were not as high as before (JJAS: -- 1.2, FMAM: -- 0.2, and May minus Mar: -- 0.8). This resulted in a reverse rainfall picture in Bangladesh, which, instead of deficits, tended to show an increase of overall seasonal rainfall: + 13.6% for the Ganges and +20% for the Brahmaputra. All the successive months displayed similar trends too. In case of strong La Ni-

a years, the values of SOIs changed from negative to positive (FMAM: -- 0.4, May minus March: +1.0), and the high positive value of SOI in JJAS (+1.9) caused an increase of average seasonal and monthly rainfall in all the three basins ( 1). Other than the Meghna basin, the excessive fall remained active during the moderate cool years too.

The most important findings here are that the extreme negative and positive values of SOIs are associated to rainfall variability, i.e., rainfall decreases with the extreme negative SOIs in JJAS, FMAM, and May minus March (El Ni-

o phenomenon); it increases with the extreme positive SOIs in JJAS, FMAM, and May minus March (La Ni-a phenomenon). For other moderate ENSO years, the strength of SOI averages for two successive seasons (lag 2, lag 1) and the May minus March can provide useful predictions of ENSO-monsoon relationship in Bangladesh.

ENSO-rainfall: Up-stream basin-wide

Basin wide -- extending from upstream India to downstream Bangladesh monsoon rainfall patterns control flood peaks of the Ganges, Brahmaputra, and Meghna Rivers in Bangladesh. By and large, the seasonal rainfall in the upstream India is a major factor that contributes significantly to the stream-flows in the downstream Bangladesh. Therefore, ENSO-rainfall relation in the upstream India is essential for drawing any comprehensive climate perspectives in Bangladesh.

Basin-wide composite picture of events-response relationship shows some interesting association between SOI and seasonal rainfall deviation in upstream basins. Here also, all major El Ni-o years recorded moderate-to-weak rainfall deficit in all the three basins; the seasonal deficits were: 4.9% in Ganges, -- 8.8% in Brahmaputra, and 0.7% in Meghna. Similarly, all the major La Ni-a years recorded excessive rainfall: +5.2% in the Ganges, +6.6% in the Brahmaputra, and +3.5% in the Meghna basins. Rainfall during the moderate El Ni-o years displays significant deficit in the upstream Ganges (-- 18.4%) and shows moderately high in the upstream Meghna (+10%). These give a clear variability picture of SOI-rainfall relation along the greater GB basins (Meghna displays slightly different results) -- the relationship is linear in the upstream India indicating rainfall decreases during the El Ni-o years and increases during the La Ni-a years (both in strong and moderate case). In case of Bangladesh the moderate El Ni-o years are wet and experience excessive local rainfall.

ENSO- seasonal flooding

The SOI-rainfall relation in the greater GBM basin systems shows strong casual connection to SOI extremes indicating negative value to dry and positive value to wet. Therefore, when SOI is negative (i.e. strong EL Ni-o years) the whole basin experiences less rainfall. The deficiency of rainfall causes Bangladesh rivers to be drying because of low-flow and, as a result, the country faces severe drought. On the other hand, when SOI is positive (both in strong and moderate La Ni-a years) there is significant increase of rainfall along the greater GB basins causing flooding along the whole catchments. This, in turn, severely floods Bangladesh, as it is the lowest riparian country in these basins. However, in case of moderate SOI-rainfall relationship (moderate EL Ni-o years), the basin-wide rainfall picture in downstream Bangladesh is relatively different from upstream India. With marginal deficit of rainfall (4%) in Meghna basin, Bangladesh experiences high rainfall along the Ganges (+13.6%) and Brahmaputra (+20.4%) basins -- these are all significantly higher than in the La Ni-a years ( 1). Although the upstream rainfall is not very dominant (18.4% for Ganges, 0.6% for Brahmaputra, and +10% for Meghna) ( 2), the exceptionally high and prolonged local rainfall contributes to flooding in Bangladesh ( 7: +60% deviation from the normal). On contrary, the excessive rainfall in the upstream (+18.5% for Ganges and +2.7% for Brahmaputra) and downstream (+19.8% for Ganges and +8.2% for Brahmaputra) of greater GB basins during the moderate La Ni-a years causes flooding inside Bangladesh ( 7: +24% deviation from the normal).

Why Bangladesh is wet during moderate El Ni-o years?

When SOI is negative, Walker circulation is weakened, then the easterly wind is also weakened or completely reversed. To clearly understand the mechanism, we have constructed two composites: (i) strong El Ni-o minus La Ni-a years, and ii) moderate El Ni-o minus strong La Ni-a years. The underlying hypothesis here is that the movement of tropical disturbances formed in the western Pacific governs the strength of the Bangladesh monsoon. It is stated in the climate literature that when the Walker circulation is weak, the Hadley circulation is strong; that is, the zonal wind weakens and the meridional wind strengthens causing absence of any significant convergence over Bangladesh. As a consequence, the tropical disturbances are transported northward or northeastward depriving Bangladesh from precipitation and causing either deficit rain or drought. On contrary, when SOI is negative (moderate El Ni-o years), the Hadley circulation is not as strong as during major El Ni-o years and, therefore, allows the tropical disturbances to cross into the Bay of Bengal and Bangladesh causing heavy rainfall and flooding.

ENSO-2002/03 and

Bangladesh climate

Following the dissipation of the 2002-03 El Ni-o in April, sea surface temperatures across the central and eastern tropical Pacific continued to decrease and are currently below average. There is now a significant possibility that a La Ni-a may develop. Based on the behaviour of past La Ni-a event onsets, and recent surface and subsurface observations and model forecasts, there has been an estimated likelihood of 55% that La Ni-a would develop by June (If a La Ni-

a does not develop, it is most likely that the tropical Pacific will remain in near-neutral conditions. If a La Ni-a does develop, associated climate effects could be experienced in June or July). This is the forecast provided by the IRI see (http://iri.

columbia.edu/climate/ENSO/currentinfo/update.html). Based on these findings, the likelihood of Bangladesh climate to be wet is above average during the monsoon of 2003. This is a probabilistic forecast that is based on monitoring of the ocean and knowledge of how the atmosphere has responded in the past to similar SSTs in Bangladesh, with a variety of lag times.

Dr. Md. Rashed Chowdhury is an UCAR visiting scientist working at the International Research Institute for Climate Prediction, The Earth Institute of Columbia University (on leave from the Flood Forecasting and Warning Center, Bangladesh)