We have conducted three sets of experiments for the real-time prediction during 2011 and 2012 and the same sets of experiments were also conducted in the hindcast mode from 2001 to 2010.

The first set of experiment is with CFSv2 with its atmospheric component (GFSv2) at T126 horizontal resolution in coupled mode. Model integrations were started from 16 May and continued up to 28 September at every 5 day interval (16 May, 21 May, 26 May, ..., 23 Sep, 28 Sep). An ensemble of 11 members integration was performed for each start date and for the next 25 days period. This experiment is termed as CFST126.

In the second set of experiment we have identified the daily SST bias in the coupled model forecast in the control run of set one and corrected the bias with respect to the observed daily SST data (OISST). Bias correction is done by removing the daily mean bias (i.e. model daily climatology ̶ observed daily climatology) from the daily forecasted SST. This corrected daily SST is provided as the boundary forcing and GFSv2 at T126 horizontal resolution was run forced with this SST for the same period for next 25 days and same initial conditions in the ensemble mode as in set one. This experiment is termed as GFSbc.

In addition, we have conducted one more set of experiment to study the impact of resolution on the extended range forecasts over Indian monsoon region. By keeping all physics and ocean components same, resolution of atmospheric component of CFS (ie:GFS) has been increased to T382 (~30 km).