Increasing potential predictability of Indian summer monsoon wet and dry spells

An understanding of the limit on potential predictability is crucial for developing appropriate tools for extended range prediction of active/break spells of Indian summer monsoon (ISM). The global low frequency changes in climate modulate the annual cycle of the ISM and can influence the intrinsic predictability limit of the ISM intraseasonal oscillations (ISOs). Using 104 year (1901-2004) long daily rainfall data, the change in potential predictability of active and break spells are estimated by an empirical method. Using an ISO index based on 10-90 day filtered precipitation, the active and break phases were identified. The peak of each event was identified and signal and error estimates were made, starting from these peaks up to 30 days lead time. Corresponding to each lead day the signal (amplitude of the ISO) is found as the variance of 50 days starting from that particular lead day (covering approximately one complete ISO event) and averaged over all events. ‘Error' is defined as the variance among the different active / break events corresponding to each lead day. They showed that the monsoon breaks are intrinsically more predictable (20-25 days) than the active conditions (10-15 days) (Goswami and Xavier, 2003). In the present study the same analysis was repeated in 15 year sliding windows, about 45-50 active or break peaks were obtained in each 15 year period. We found that the potential predictability of both active and break spells have undergone a rapid increase during the recent three decades. The potential predictability of active spells (figure 1b) has shown an increase from one week to two weeks while that for break spells (figure 1a)increased from two weeks to three weeks. This result is interesting and intriguing in the backdrop of recent finding that the potential predictability of monsoon weather has decreased substantially over the same period compared to earlier decades due to increased potential instability of the atmosphere. The possible role of internal dynamics and external forcing in producing this change has been explored. The magnitude of initial errors, i.e. variance among peak active/break conditions shows a steady decrease over the years, indicating a lesser event to event variability in the magnitude of ISO peak phases in recent years. For understanding the changes in evolution characteristics of the two ISO phases, the mean evolution time from active peak to a break peak and vice versa were computed in each of the 15 year window. The mean period of evolution from an active phase to break phase was about 15 days, shows a slight decrease to 13 days after mid seventies. While the mean period of evolution from a break phase to an active phase shows a steady value of about 15 days until mid seventies, and increases to 21 days in recent time. The increase in number of slower evolving events also supports the higher potential predictability. Also, the increase in predictability of breaks after mid eighties in conjunction with the SST increase in recent decades implies that the increased air-sea coupling might have crossed a threshold of influence such that the boundary induced variability is well above the chaotic variability due to nonlinear internal dynamics. The ISO predictability may be closely linked to the error energy cascading from the synoptic scales and the interaction between these scales. An increase in potential predictability of ISOs implies an ineffectual influence of synoptic scale errors on the ISO scale. The synoptic-ISO energy exchanges were also examined to bring out the influence of synoptic scale errors on ISO predictability. The interdecadal changes in the nonlinear kinetic energy transfer between these scales were computed in the frequency domain. It was found that until 1980, the ISO scale was drawing energy from the synoptic scales, after eighties the exchange is a downscale transfer of kinetic energy was observed. Since the error energy spectrum is directly related to the total kinetic energy spectrum, these results support our argument of inefficient error cascading from synoptic to ISO scale. The changes in energy exchange between the synoptic and ISO scale and the different ISO modes as evidenced by energetics computations in frequency domain also support the increased potential predictability of ISO. Our finding provides optimism for improved and useful extended range prediction of monsoon active and break spells.Ref: Goswami BN, Xavier PK. 2003. Potential predictability and extended range prediction of Indian summer monsoon breaks. Geophys. Res. Lett. 30(18): 1966, doi:10.1029/2003GL017810.