Abstract
In 2022, the Indian state of Bihar experienced its sixth driest year in over a century. To document the consequences and farmer responses to the meteorological drought, real-time survey data was collected across 11 districts of Bihar. We then developed a causal machine learning model to quantify drought impacts on rice production and to characterize how access to affordable irrigation from electric pumps mitigated productivity losses. This model addresses the empirical challenge of conducting a counterfactual causal analysis when a factor like drought affects nearly all sampled farmers. In the 2022 event, drought led to rice acreage reduction, transplanting delays, damage to seedling nurseries, and higher use rates of supplemental irrigation. For fields that were planted, average yield losses from water stress were estimated as 0.94 t/ha (∼23 % yield loss) with these losses reduced by 0.3 t/ha in fields with access to electric tubewells. Agronomic management practices such as earlier transplanting were also identified as complementary strategies that increased the adaptation value of investments in irrigation. To reduce the impact of drought in Bihar, additional investments in electric irrigation infrastructure are needed along with focused extension efforts and decision support systems that empower farmers to make economically and sustainably rational use of available water resources to maintain yield and profitability.