Assessing the applicability of NDVI data for the design of index-based agricultural insurance in Bihar, India

Appropriate management of agricultural risks could prevent smallholder farmers in India from falling into poverty traps. Index-based insurance schemes offer policy holders a payout based on an objective indicator (e.g. rainfall). One main problem with weather-index based insurance is that the correlations between weather and yield variables can be low in some cases. Here we evaluate the potential of remotely-sensed Normalised Difference Vegetation Index (NDVI) to estimate crop yield in the state of Bihar, India. We use panel linear regression analysis to compare the relationship between rainfall and NDVI with rice, maize and wheat yield on the district level. We obtained highly significant, but low R²-values (<; 0.3). In most cases, NDVI explained crop yield variance better than cumulative rainfall. Furthermore, incorporating both NDVI and rainfall in the regression model was beneficial.