干旱区生态环境知识资源中心

The annual cycle of solar radiation, together with the resulting land-ocean differential heating, is traditionally considered the dominant forcing controlling the northward progression of the Indian monsoon. This study makes use of a state-of-the-art atmospheric general circulation model in a realistic configuration to conduct "perpetual" experiments aimed at providing new insights into the role of land-atmosphere processes in modulating the annual cycle of precipitation over India. The simulations are carried out at three important stages of the monsoon cycle: March, May, and July. Insolation and SSTs are held fixed at their respective monthly mean values, thus eliminating any external seasonal forcing. In the perpetual May experiment both precipitation and circulation are able to considerably evolve only by regional internal land-atmosphere processes and the mediation of soil hydrology. A large-scale equilibrium state is reached after approximately 270 days, closely resembling mid-summer climatological conditions. As a result, despite the absence of external forcing, intense and widespread rains over India are able to develop in the May-like state. The interaction between soil moisture and circulation, modulated by surface heating over the northwestern semi-arid areas, determines a slow northwestward migration of the monsoon, a crucial feature for the existence of desert regions to the west. This also implies that the land-atmosphere system in May is far from being in equilibrium with the external forcing. The inland migration of the precipitation front comprises a succession of large-scale 35-50 day coupled oscillations between soil moisture, precipitation, and circulation. The oscillatory regime is self-sustained and entirely due to the internal dynamics of the system. In contrast to the May case, minor changes in the land-atmosphere system are found when the model is initialized in March and, more surprisingly, in July, the latter case further emphasizing the role of northwestern surface heating.