INTRINSIC PROBLEMS IN THE SEASONAL PREDICTION OF THE INDIAN-SUMMER MONSOON RAINFALL

The latest non-parametric statistical tool Singular Spectrum Analysis (SSA) has been shown to extract deterministic oscillations present in a nonlinear dynamical system. It has been hypothesized that the tropical ocean-atmosphere system consists of both deterministic and stochastic parts in the interannual time scales. In the present study SSA has been employed to extract the deterministic and random parts present in the Indian summer monsoon (ISM) and its predictors time series data sets. The dominant eigenmode pair of the ISM does not emerge as a pure and deterministic oscillation. However, about 34% variance is deterministically predictable in the inter-annual range. The second pair is significantly related to the first pair of Darwin pressure tendency and both emerge as deterministic parts. This relationship partially answers the questions raised by Webster and Yang (1992). The low frequency component of ENSO emerges as a deterministic oscillation in all the variables, except in Bombay pressure tendency. The presence of decadal-scale oscillations may possibly be responsible for the instability in the relationship between the ISM and its predictors. Some plausible explanations for the percent variance explained by the predictors in the existing empirical models have also been discussed. It has been proposed that empirical models can be constructed only with the deterministic parts which may help improve the predictive skill of existing models.