The role of the basic state in determining the predictability of the tropical climate system, with particular reference to the Asian summer monsoon

Andrew Turner( A.G.Turner@reading.ac.uk), Pete Inness and Julia Slingo, CGAM, University of Reading.

The basic state errors in the tropical Pacific of the UK Met Office Hadley Centre Model (HadCM3) are partially corrected by the use of a flux adjusted version of the model (HadCM3FA), originally devised by Inness et. al. (2003). The anomalous summer zonal temperature gradient in the equatorial Pacific is reduced, correcting the excessive trade winds prevalent there in HadCM3. The strength of the Pacific trades is a well-known signature of the Asian summer monsoon circulation (see, e.g. Webster et al. 1998). Consequently, their reduction in HadCM3FA is coupled with a reduction in the strength of the broad-scale monsoon over Asia, as measured by the dynamical monsoon index (DMI), first used by Webster and Yang (1992). The index covers 5-20°N, 40-110°E and is a measure of the anomalous zonal wind-shear in summer. A reduced south-westerly monsoon flow is observed in the summer HadCM3FA climate. However, monsoon variability in the Hadley Centre model increases with flux adjustments, observed as an increase in the standard deviation of the DMI in 40 year integrations of the model: 1.20 for HadCM3, 2.19 for HadCM3FA, 1.60 for ERA40, (used as an indicator of the real-world scenario). The increased variability is also observed in strong minus weak monsoon difference composites of lower tropospheric wind, surface temperature and precipitation. Differences of strong and weak monsoon summers therefore show stronger monsoon flow into India with the flux-adjusted version of the model, and this is also represented in the equatorial Pacific with increased easterly trade winds during strong monsoon summers. Possible sampling errors were accounted for by extending integrations to 100 years, to no effect on the overall circulation variability or summer mean climate. That the use of a seasonal cycle of flux-adjustments to force sea surface temperatures in a coupled GCM increases the interannual variability is somewhat surprising. The adjustments have altered the ENSO cycle to the extent that the standard deviation of the Nino-3 timeseries has increased (0.87, 1.29, 0.85 for HadCM3, HadCM3FA, and ERA40 respectively). Correlations between the Nino-3 timeseries and the DMI for ERA40 and the 100 year model integrations (see figure) show a stronger ENSO teleconnection in the flux adjusted model. This is to be expected with the mean state correction, extending the warm pool further eastward. The spring predictability barrier becomes evident with flux adjustments, represented by the steep nature of the HadCM3FA correlation curve around March/April time of the same year prior to the summer monsoon. Stronger ENSO forcing in HadCM3FA may explain this, but its lack in HadCM3 is not yet understood.

Figure 1: Correlation between Nino-3 and summer (JJAS) DMI plotted against lag in months. ¡ÆAug-1¡Ç indicates the August Nino-3 anomaly the year before DMI. Correlations significant in all but 5% of cases are indicated outside the dashed lines: |r|>0.31 for the 40 year dataset, |r|>0.20 for the 100 year datasets.

References