Summary prepared by: Fennessy
last update: 5 November 1994
The COLA GCM has been used simulate the difference between the anomalously dry Indian summer monsoon of 1987 and the wet Indian summer monsoon of 1988 using observed global SST as a boundary condition. Six 90 day integrations were carried out starting from the observed initial state of the atmosphere on 1,2,3 June, 1987 and 1,2,3 June, 1988, using the observed time- varying global SST. The ensemble 1988 - 1987 June-July-August seasonal mean (JJA) rainfall difference was near zero over India. However, an analysis of the seasonal mean Indian monsoon simulation in these integrations revealed serious model deficiencies, particularly in the simulation of rainfall over India.
As described by Fennessy et al. (1994), we carried out a large number of sensitivity studies with the aim of improving the Indian monsoon simulation. We found that the replacement of the enhanced silhouette orography by a mean orography more representative of the actual terrain heights produced a more realistic simulation of the Indian monsoon circulation and rainfall. The six 90 day integrations for 1987 and 1988 were repeated with the improved model. The ensemble 1988 - 1987 rainfall differences over the entire Asiatic monsoon region, and especially over India, are quite realistic. This study clearly shows that the model's ability to simulate the differences in the 1988 and the 1987 monsoon rainfall crucially depends on the realism of the simulated mean climate.
Figures 1a through 1i show nine different simulated JJA mean precipitation differences for the Indian monsoon region. In each case, the precipitation from an integration done using observed time-varying global SST for 1987 is subtracted from the precipitation from an integration done using observed time-varying global SST for 1988. Otherwise, the only other differences among the 18 integrations are the observed atmospheric conditions used for initialization, the dates of which are given in Table 1.
Despite the wide range of initial conditions used in the 18 integrations, the JJA mean rainfall anomalies simulated over India with 1988 SST minus 1987 SST are remarkably similar in eight of the nine cases. Furthermore, the JJA ensemble mean difference of the last 6 random initial condition cases (Figs. 1d - 1i, ensemble not shown), is nearly indistinguishable from the JJA ensemble mean of the first 3 cases which used realistic initial conditions (Figs. 1a - 1c, ensemble not shown).
This suggests that for this set of boundary conditions and for this model, the simulated interannual variability does not have much sensitivity to the initial atmospheric conditions.
a. Active and break simulations for 1987 and 1988
The intraseasonal variability of the COLA GCM precipitation has been analyzed in light of the abundant research in the literature on the phenomena of "active" versus "break" monsoon periods. Daily mean data from 90 day simulations using observed SST as a boundary condition and initialized from observed initial states for 1987 and 1988 are examined. Simulated active/break monsoon periods are defined based on the central India precipitation, in accordance with the precipitation deficit region associated with historical monsoon breaks.
Composite maps of the simulated precipitation and atmospheric circulation were produced for active and break periods for 1987, 1988 and the two years combined. An examination of these maps reveals the qualitative similarity of the active and break monsoon periods between these two years. The simulated break composite minus active composite precipitation difference pattern is similar to the historical break departure from normal precipitation map and is remarkably similar to the JJA mean precipitation difference between 1987 and 1988. Simulated active, break and difference composite maps of upper and lower level circulation features are in broad agreement with those associated with historical active/break monsoon periods.
The 1988 simulations contain more active days and fewer break days than the 1987 simulations, however the magnitude of the active and break periods were similar among the simulations. Thus, the simulated wet/dry monsoon of 1988/1987 is associated with more/less active monsoon days and less/more break monsoon days, rather than with a change in the magnitude of the active or break monsoon periods, in agreement with historical observations.
b. Role of land surface boundary conditions
The possible role of land surface boundary conditions in modulating the magnitude and/or frequency of the active/break monsoon cycle is also being investigated. If the active/break cycle is driven or modulated by intraseasonal heating/cooling, then the soil wetness may affect it through its impact on the land surface intraseasonal temperature variability. The role of soil wetness in the local surface heating is a key part of the low-frequency monsoon variability mechanism proposed by Webster (1983). We performed an extreme experiment designed to greatly reduce the local surface intraseasonal heating/cooling by artificially saturating the soil over India and Pakistan throughout the course of a 90-day integration. This integration (hereafter WetSoil) is compared to a control integration in which the soil wetness is predicted by the model. The resulting soil wetness anomaly ranges from 30 to 60% over most of this region.
This large soil wetness anomaly had a relatively small impact on the seasonal mean precipitation, with less than 1 mm/day WetSoil minus control differences occurring over most of India. However, the intraseasonal variability of the simulated precipitation over the region was greatly reduced in the Wet Soil integration as compared to that in the control integration. Particularly striking is the near absence of break periods in the WetSoil integration. This reduction in the intraseasonal variability of precipitation is likely related to the dramatic reduction in the intraseasonal variability of temperature in the WetSoil integration versus the control integration.
In order to determine whether the reduction in the intraseasonal precipitation variability was due solely to local effects or whether the response of the large scale circulation also played a role, we examined the area averaged time series of evaporation and vertically integrated moisture flux convergence. The evaporation in the control simulation is quite variable while that in the WetSoil integration is both higher and considerably less variable. The variability of the vertically integrated moisture flux convergence is also largely reduced in the WetSoil integration compared to the control integration.
These results imply that the surface hydrology can provide a feedback mechanism which modulates the precipitation intraseasonal variability through both local effects and by influencing the large scale circulation.
Personnel: Fennessy, Shukla
Summary prepared by: Fennessy
last update: 5 November 1994