Teleconnections in recent time and prediction of Indian summer monsoon rainfall

• Main Authors: Sahai, AK, Pattanaik, DR, Satyan, V, Grimm, AM
• Format: Article in Journal/Newspaper
• Language: unknown
• Published: Springer 2003
• Subjects: Meteorology and Climatology; Indian; Pacific; North Atlantic
• Online Access: http://moeseprints.incois.gov.in/1487/; http://download.springer.com/static/pdf/228/art%253A10.1007%252Fs00703-002-0595-1.pdf?auth66=1427347803_5c8101bb196823b5b27d3a986734b822&ext=.pdf

The prediction of Indian Summer Monsoon Rainfall (ISMR) is vital for Indian economic policy and a challenge for meteorologists. It needs various predictors among which El Nino-Southern Oscillation (ENSO) is the most important. It has been established by various researchers that ENSO and ISMR relationship is weakening in recent years. It has been also argued that changes in ENSO-ISMR relationship may be due to decadal fluctuations, or it may be the indicative of longer-term trends related to anthropogenic-induced climate changes. In the present communication, an attempt is made to discuss the variability and predictability of ISMR in recent years. It is found that three different indices associated with different regions in the tropics and extra-tropics at different levels of the atmosphere-Asian land mass index represented by geopotential height at upper troposphere (AI), Caribbean-North Atlantic index represented by geopotential height at middle troposphere (A2) and tropical Pacific index at surface level (A3) - have different mechanisms to interact mutually and separately with ISMR in different periods. In recent years ISMR shows weak association with A1 and A3 while strong association with A2. Thus, if these three indices could be combined objectively, they can give rise to the predictability of ISMR. This objective combination is achieved here using Artificial Neural Network (ANN) and a model is developed to predict ISMR. This model has predicted reasonably well during the whole period of consideration (1958-2000) with a correlation coefficient of 0.92 in last 11 years (1990-2000) whereas most of the models fail to predict the variability in recent time.