On the predictability of Indian summer monsoon rainfall in general circulation model at different lead time

Abstract

The objective of this present study is to analyze the predictability of all India summer monsoon rainfall (AISMR) and its dependence on lead time using general circulation model (GCM) output. For the purpose, six GCMs for the hindcast run from 1982 to 2008 are used at three different initializations viz. April (lead 2), May (lead 1), and June (lead 0) for seasonal mean rainfall of June–July–August–September (JJAS). Among these models, four of them are the coupled ocean–atmosphere GCMs (CGCMs) and the remaining two are the atmospheric GCMs (AGCMs). The analysis is made on the basis of statistical measures of predictability including climatology, interannual variability, root mean square error, correlation, signal to noise ratio, potential model predictability and index of agreement. On the basis of these measures it is found that all the GCM having the minimum prediction skill is at lead 2 compare to lead 1 and lead 0. It is also noticed that higher predictability in the lead-1 forecasts is found in coupled models whereas, the predictability of atmospheric models exhibit high in lead 0. Rather than rainfall, teleconnection of rainfall with large scale features (such as sea surface temperature, zonal wind at 850 hPa) and monsoon dynamic index (Indian monsoon index (IMI)) are also examined in GCMs. The results depicted that there is not much variation in the teleconnection pattern in two leads (lead 0 and lead 1) whereas; the dynamic index being predicted closer to the observed value at lead 1 in the CGCMs. The GCMs are also examined during four typical monsoon (excess/deficit) years, among which 1983 and 1988 are excess and 1987 and 2002 are deficit. Results indicate that the coupled (atmospheric) models capture the extreme rainfall signal in lead 1 (lead 0). The probabilistic prediction skill of GCM predicted rainfall is also evaluated which supports our initial analysis and results.