Teleconnections and long range prediction of summer monsoon rainfall variability over Bangladesh

Abstract

In Bangladesh, the agricultural economy with large growing population is closely linked with the performance of its monsoon systems, namely summer monsoon (SM), active during June to September, which receives 70.7% of the annual rainfall of the country. In the view of critical influence of summer monsoon rainfall (SMR) on agricultural activities, industrial production and other water based enterprises in Bangladesh, prediction of the SMR is important for the policy making and planning of the country in mitigating efforts. Therefore, long-range forecasting (LRF) of SMR is a high priority in Bangladesh as there is no dynamical and statistical model to give the LRF for Bangladesh. In this study, interannual and decadal variability of SMR using long term historical data for 48-year (1961-2008) has been used. An attempt is made to study the rainfall variability over Bangladesh on administrative divisional scale in detail. The teleconnections of SMR with various global parameters have been investigated and an empirical regression model has been developed for prediction of SMR. A high resolution (20 km) climate model named MRI-AGCM (Atmospheric General Circulation Model) output data has been used to predict the seasonal SMR during past 28 years (1979-2006) and results have also been verified with the regression model. The first 25 years (1979- 2003) data of AGCM has been used for calibration and remaining three years data (2004- 2006) has been used for validation. SMR scenario/projection has also been generated by this model for near future (2015-2034) and future (2075-2099) for Bangladesh. The interannual variability of SMR shows random fluctuations, the decadal variability shows alternate epochs of above and below normal rainfall. The epochs tend to last for a decade or two. However, no long-term trends have been detected. The long-term average of SMR is 1701 mm, with a standard deviation of 199 mm. Its coefficient of variation is 11.6%. In Bangladesh, the interannual variability of SMR is very high; the interannual variability of SMR occasionally leads to large-scale deficient/droughts and excess/floods over the different parts of Bangladesh. Teleconnections between SMR and global parameters have been investigated, and it has been found that above normal SMR is associated with warm Sea Surface Temperature (SST) in the month of February over southwest Indian Ocean and it is positively and significantly correlated with correlation coefficient (CC) 0.44 (significant at 1% level). The correlations are weak over rest of the Indian Ocean and Bay of Bengal. The correlations are insignificant throughout the Indian Ocean and Bay of Bengal for the others month. It is observed that above-normal SMR is associated with warm Surface Air Temperature (SAT) in the month of April over Somalia and is positively and significantly correlated with CC=0.59 (significant at 1% level). The correlations are weak over rest of the part of the globe. Above-normal SMR is associated with high Mean Sea Level Pressure (MSLP) in the month of April over central Pacific Ocean is positively and significantly correlated with CC=0.53 (significant at 1% level). The statistical model has been developed using three predictors namely SST, SAT and MSLP anomalies and it represents various forcing on SMR. The model showed reasonably good result during the training period 1979-2002 and performed well during the independent verification period 2003-2012 (10-year). AGCM generated SMR scenario is calibrated with observed (rain-gauge) data during the period 1979-2003. The bias correction method of the World Climate Research Programme (WCRP) is utilized for validation of AGCM generated SMR during the period 2004-2006. Better performance of AGCM through validation encourages utilizing it in SMR forecast for Bangladesh. The change of near future SMR was forecasted for Bangladesh by -27.6 to 24.7% for the period 2015-2034. Similarly, the change of future SMR was forecasted for Bangladesh by -29.4 to 29.4% during the period 2075-2099. On an average near future and future SMR may change by -0.5% and 0.4% during the period 2015-2034 and 2075-2099, respectively. AGCM calibrated and regression model mean SMR are compared with observed seasonal mean SMR during the period 1979-2006, it is seen that both AGCM and regression model SMR are close to observed rainfall for the period 1979-2006.