Currently, a model (referred to as a monthly model) employing monthly temperature and precipitation data is used by the Canadian Wheat Board to estimate spring wheat yields for the Canadian Prairies. The model uses a cumulative moisture index as an explanatory variable. In this paper, the performance of the monthly model was improved by first developing a daily model by employing daily, instead of monthly, data for the 1975-1996 period and then by developing a hybrid model which incorporated into the daily model an additional variable derived from the National Oceanic and Atmospheric Administration (NOAA)/Advanced Very High Resolution Radiometer (AVHRR)-based composited Normalized Difference Vegetation Index (NDVI) data for the 1987-1996 period. Out of the seven variables derived, two variables - the average NDVI during the heading phenological-phase and the average NDVI during the entire growing season - were found to be the best. The start and the end of the heading phase were estimated using a biometeorological time scale model. The performance of models was tested on five crop districts (1b, 3bn, 4b, 6a, and 9a) of Saskatchewan on the basis of coefficient of determination, R 2 . While using 1975-1996 data, the values for R 2 were 0.43, 0.82, 0.73, 0.71 and 0.00 in the case of the daily model as opposed to 0.20, 0.71, 0.57, 0.58, and 0.00 in the case of the monthly model for districts 1b, 3bn, 4b, 6a, and 9a, respectively. While using 1987-1996 data, the values of R 2 were 0.79, 0.96, 0.83, 0.95, and 0.39 in the case of the hybrid model as opposed to 0.13, 0.70, 0.75, 0.50, and 0.00 in the case of the monthly model for districts 1b, 3bn, 4b, 6a, and 9a, respectively. For district 9a, which experiences an adequate supply of soil moisture, the concept of cumulative soil moisture index was not found to hold well for yield estimation.