1. Click to edit Master title style Timeseries of dynamical and precipitation indices Black: observations; RGB: ensemble members for each start date case running left to right. Dynamical indices other than DMI show poor predictability. AIRWebster-Yang DMI Goswami MHI Wang-Fan circulation

2. Click to edit Master title style Mean state representation #1 (precip,SST) ObsModel Obs Indian precipitation pattern reasonable. Clear westward bias in mean state SST (JJA): confined warm pool and extended cold tongue.

3. Click to edit Master title style Mean state representation #2 (circulation) Obs Model Obs Model Trade winds too strong in the western Pacific in GloSea4. Mean Walker circulation is not too bad although slight westward bias (particularly in descending regions).

4. Click to edit Master title style Initially, hindcast precipitation response over India does not look unreasonable. However, largest signals are west of the coast and over the Himalaya and may not be reflected in observed AIR. As we shall see, response in some individual cases is not ideal. Regressions onto Niño-3 #1

5. Click to edit Master title style Regressions onto Niño-3 #2 Obs Model Evidence of westward bias of SST and perturbations to the large scale circulation.

6. Click to edit Master title style 1997 and 2002 chosen owing to their diversity and unusual effects on the monsoon (hence a tough test of the seasonal forecast model). 1997: large, east Pacific El Niño, normal monsoon (102% AIR). 2002: more moderate, central Pacific El Niño, monsoon drought (81% AIR). Case studies for 1997 and 2002

7. Click to edit Master title style Case studies: 1997 (SST & vp200 anomalies) Model Obs Warm El Niño anomalies extend ~30° too far west. Anomalous ascent over warm SST pushes into the Maritime Continent, shifting anomalous subsidence west over India.

8. Click to edit Master title style Case studies: 2002 (SST & vp200 anomalies) Model Obs Warm El Niño anomalies again extend too far west. Anomalous subsidence over Indian is shifted too far west, as far as East Africa.

9. Click to edit Master title style Case studies: 2004. More unusual case of El Niño followed rapidly by small La Niña. Anomalous ascent in East Pacific: responding to La Niña instead of El Niño! Obs Model

10. Click to edit Master title style Clear but not unexpected improvement in mean state using June start. 1997 case study using different start dates (1 April, 1 May, 1 June)

11. Click to edit Master title style Summary While dynamical teleconnections are slightly better represented than AIR, this might reflect larger domain used; rainfall teleconnection is still poorly represented. Coupled biases are shown in mean state 850hPa wind (easterlies penetrate too far west) & SST (equatorial cold tongue extends too far west), known to be of detriment to teleconnection simulation in coupled models (Turner et al., 2005; Annamalai et al., 2007). SST biases develop rapidly in the equatorial Pacific following initialisation. Anomalous descent situated too far west in both 1997 and 2002 case studies. Future experiments: test sensitivity to mean state in these cases using wind stress corrections.