Effect of the mixed phase precipitation scheme on HadAM3 climatology

In deciding to use the mixed phase precipitation scheme (MPP) in integrations of HadAM3 we first ran an experiment comparing the climatology with and without the scheme. The climatologies were constructed by integrating the model for 15 years with climatological observed (Reynolds) SSTs (1982-1999). We are mainly interested in the model's performance during the peak rainfall season over southern Africa (January-March, JFM).

The MPP scheme was developed at the Hadley centre, UK and the main effects of the scheme on the global climatology of HadAM3 is given by Wilson, 2000 (Q. J. R. Meteorol. Soc., 126, pp. 1281-1300). The results below are for the untuned version of the scheme and are shown to indicate the primary changes of the climatology and predicted anomalies in the African sector.

The primary reason for undertaking the analysis was a noticeable cyclonic bias in the model to the north-east of Madagascar. This bias can be seen in the figure below which shows the moisture flux climatology for the OND and JFM seasons in both the NCEP reanalysis and model datasets. Areas of convergence are marked blue and areas of divergence are marked red. The strong model convergence over the central Indian Ocean is also apparent.

Figure 2 shows the effect of MPP on the 500 hPa JFM climatology. The top map shows the bias of the standard version of the model with respect to NCEP and the bottom map shows the effect of MPP on the standard version of the model. If MPP is to improve the model performance the changes seen in the bottom figure should counteract the biases seen in the top figure. Most notable for southern Africa is a reduction in the low height bias of the mid-latitude westerlies in the southern hemisphere when MPP is introduced into the model. The general increase in heights over the globe is associated with an increase in global tropospheric temperature (not shown). However the reduction in the height bias is smaller than the bias itself.

Figure 3 demonstrates the effect of MPP on the 200 hPa velocity potential climatology. The strong convection in the western Indian Ocean in Figure 1 can be seen manifested as a region of upper-level divergence at 200 hPa close to Madagascar. MPP reduces this bias by approximately 50%. The reduced upper level divergence is associated with a reduction in the cyclonic bias seen in Figure 1, thereby improving the low-level dynamics in the western Indian Ocean. Again it should be noted that the bias still remains.

Effect of the mixed phase precipitation scheme on HadAM3 anomalies

A simple test of the effect of MPP on the model dynamics can be seen in Figure 4 which shows the difference in 850 hPa eddy geopotential height anomalies between two El-Nino (1992,1998) and two La-Nina (1997,1999) years (the data is taken from a 15 year, 1985-1999, 5 member ensemble hindcast using observed SSTs). It can be seen that both the standard version of HadAM3 (STD) and MPP capture the low-level dynamics over the central-eastern Pacific as evidenced in the NCEP reanalysis data. There are subtle differences between MPP and STD over the eastern Indian Ocean near Australia with MPP capturing the strength of the high pressure anomaly. Closer to southern Africa MPP better captures the high pressure anomaly south of the continent, though it is further east than demonstrated in the NCEP data. These features are important for southern African rainfall as they are associated with tropical temperate troughs which are responsible for the majority of rainfall variance over the continent.

Figure 5 shows the equivalent of Figure 4 for precipitation. As expected the CMAP data (satellite and rain gauge merged dataset) shows a negative anomaly over southern Africa. However the STD version of the model fails to capture this variability. With MPP some of the variability over the eastern part of the subcontinent is simulated.

Figure 6 shows the first EOF of JFM rainfall (left) and outgoing longwave radiation (right) for the period 1985-1999 for the two versions of the model and CMAP. There is a clear lack of variability in the rainfall over the continent in the standard version of the model. However MPP displays a variance structure similar to the CMAP data and its OLR pattern better matches that observed by satellite. The timeseries of EOF1 scores is given in Figure 7 and it is notable that the standard version of the model fails to capture the negative anomalies during 1992 which saw one of the worst droughts in the region.