This study reports the results of model simulations of the potential yield of tea in north-east India. The CUPPA-Tea model, developed using data from a high-altitude site close to the equator in East Africa, was validated against the yield data from irrigation experiments conducted on contrasting soil types at Siliguri and Tezpur in the tea growing region of north-east India. The close correspondence between observed and predicted yield and yield distribution suggests that the model is applicable in north-east India. The model was used to simulate the yield response of tea to drought and irrigation using daily weather data for seven years (1983–89) at Siliguri and 14 years (1974–85) at Tezpur. On a clay loam soil at Siliguri, with an available water capacity of about 200 mm m−1, the predicted mean reduction in yield was 1.5 kg ha−1 for each 1 mm reduction in evapotranspiration. However, there appeared to be no reduction in evapotranspiration until the soil water deficit reached about 240 mm. By contrast, at Tezpur on loamy sand, with an available water capacity of about 100 mm m−1, evapotranspiration was reduced once soil water deficits exceeded about 85 mm. There was then a mean reduction in yield of 2.2 kg ha−1 for each 1 mm reduction in evapotranspiration. At both sites, even when soil water deficits were not limiting, the predicted year-to-year variation in yields was about 500 kg ha−1. The results highlight the soil-related differences in response to irrigation and the benefits of using process-based simulation models to investigate the potential yields over long periods.