Adaptability of Irrigation to a Changing Monsoon in India: How far can we go?

Abstract:

Agriculture and the monsoon are inextricably linked in India. A large part of the steady rise in agricultural production since the onset of the Green Revolution in the 1960’s has been attributed to irrigation. Irrigation is used to supplement and buffer crops against precipitation shocks, but water availability for such use is itself sensitive to the erratic, seasonal and spatially heterogeneous nature of the monsoon. We provide new evidence on the relationship between monsoon changes, irrigation variability and water availability by linking a process based hydrology model with an econometric model for one of the world’s most water stressed countries. India uses more groundwater for irrigation than any other country, and there is substantial evidence that this has led to depletion of groundwater aquifers. First, we build an econometric model of historical irrigation decisions using detailed agriculture and weather data spanning 35 years. Multivariate regression models reveal that for crops grown in the wet season, irrigation is sensitive to distribution and total monsoon rainfall but not to ground or surface water availability. For crops grown in the dry season, total monsoon rainfall matters most, and its effect is sensitive to groundwater availability. The historical estimates from the econometric model are used to calculate future irrigated areas under three different climate model predictions of monsoon climate for the years 2010 – 2050. These projections are then used as input to a physical hydrology model, which quantifies supply of irrigation water from sustainable sources such as rechargeable shallow groundwater, rivers and reservoirs, to unsustainable sources such as non- rechargeable groundwater. We find that the significant variation in monsoon projections lead to very different results. Crops grown in the dry season show particularly divergent trends between model projections, leading to very different groundwater resource requirements.