Role of land-surface vegetation in the march of Indian monsoon onset isochrones in a coupled model

Although the Climate Forecast System version-2 model simulates an overall dry bias in boreal summer mean rainfall over Indian land, the deficiency is particularly prominent over northwest India. The prevailing dryness limits the interannual prediction skill of the Indian summer monsoon rainfall and its subseasonal variability because of poor representation of latent heating due to weak moist convection and the resulting circulation. Here, we show that land-surface vegetation plays a crucial role in determining the dry bias in the Climate Forecast System version-2 model. We replaced the land-surface model's existing vegetation type over India with that derived from recent satellite-based observations. The modifications helped improve the seasonal mean rainfall over northwest India by 6%. The improvements are especially noticeable during the monsoon season's onset (14%) and withdrawal (10%) phases. Simulations with modified vegetation advanced the onset dates over Kerala, central India, and northwest India closer to that observed. This improvement in the mean onset dates is most prominent over northwest India. Such an improvement was possible owing to a substantial reduction of long rainfall hiatus after onset over Kerala in the simulation with modified vegetation. The modification makes the spatial orientation of monsoon onset isochrones more realistic. We found that although the vertically integrated moisture flux is eastward over most of the Indian monsoon region during its onset phase, its intraseasonal components are westward. In other words, at the intraseasonal time-scale, moisture propagates against the prevailing low-level westerlies. This intraseasonal eddy moisture transport advances onset from the Bay of Bengal toward the far northwest parts of the Indian land. The representation of such intraseasonal moisture seepage in the model updated with satellite-derived vegetation types was improved. Our study indicates the necessity of greater attention to land-surface representations for improved predictions of onset dates.