Improvements in the representation of the Indian summer monsoon in the NCEP climate forecast system version 2

Abstract

A new triggering mechanism for deep convection based on the heated condensation framework (HCF) is implemented into the National Centers for Environmental Prediction climate forecast system version 2 (CFSv2). The new trigger is added as an additional criterion in the simplified Arakawa–Schubert scheme for deep convection. Seasonal forecasts are performed to evaluate the influence of the new triggering mechanism in the representation of the Indian summer monsoon in the CFSv2. The HCF trigger improves the seasonal representation of precipitation over the Indian subcontinent. The new triggering mechanism leads to a significant, albeit relatively small, improvement in the bias of seasonal precipitation totals. In addition, the new trigger improves the representation of the seasonal precipitation cycle including the monsoon onset, and the probability distribution of precipitation intensities. The mechanism whereby the HCF improves convection over India seems to be related not only to a better representation of the background state of atmospheric convection but also to an increase in the frequency in which SAS is triggered. As a result, there was an increase in convective precipitation over India favored by the availability of moist convective instability. The increase in precipitation intensity leads to a reduction in the dry bias.

Appendix: Precipitation changes in the tropics

Figure 14 shows the JJAS precipitation bias for the entire tropics as well as the difference between HCF and CTRL. As mentioned before, the new triggering mechanism causes only small changes in the CFSv2 precipitation bias (Fig. 14a, b). Besides changes in the Indian summer monsoon, there are significant differences in precipitation over the Central America, the Amazon, tropical Atlantic Ocean, and tropical Africa (Fig. 14c). Although the new trigger improves precipitation over the Indian region, the magnitude of the bias increases over the tropical Atlantic and over tropical Africa (Fig. 14c). Consistent results (slightly smaller differences) are observed when evaluating the simulations for different periods, such as May through July (not shown). This is an example of a well-known phenomenon in developing climate models. When making model modifications it is common to observe the improvement of some regions and the detriment of others. The most important aspect of this research is the fact that we are making the parameterization more physically realistic. In addition, biases can be compensated in the future from the other parameterizations that were tuned to have the best performance in relation to the prior convective trigger.

Fig. 14

Global JJAS precipitation (mm/day) bias for a CTRL and b HCF and c mean JJAS precipitation difference between HCF and CTRL. Contour indicates regions where the difference is statistically significant at 5 % level according to a t test. The figure also shows information about precipitation average, spatial RMSE, and spatial correlation (in relation to TRMM) for the entire domain

Presumably, the changes in the precipitation biases in the Indian region could be due to changes in convection elsewhere, like the tropical Pacific, due to teleconnections. However, we did not identify significant circulation changes showing wave-like patterns between HCF and CTRL simulations (not shown). Rather, the changes in circulations are very local. Therefore, although remote influences could be playing a role in the changes in the precipitation biases over India, the mechanisms discussed in the conclusions are likely the dominant mechanisms.