The existing procedures for crop yield estimation involve Crop Cutting Experiments ( CCEs) conducted during harvesting time in the plots selected based on a pre-designed sampling scheme using available ground data. These ground sampling designs do not consider the crop condition which is directly related to the yield during the season, for stratification and subsequent sample selection leading to biased distribution of plots. Moreover these experiments are capable of providing estimates only at larger areal units such as the total command area. Hence there is a need to improve the sampling design to achieve more accurate estimates. An alternate methodology exploiting the information on crop area and crop condition, derived from satellite remote sensing data on near real-time basis, for improving the ground sampling design has been proposed in this paper. The methodology is demonstrated in the Davangere and Malebennur divisions of the Bhadra project command area to estimate the average yield of paddy during Rabi 1992-93. The results obtained from conventional methodology and the improved procedure showed that the latter has increased the accuracy of estimates. The yield values obtained from CCE plots have been regressed with corresponding Normalized Difference Vegetation Index (NDVI) statistics and thus the derived paddy yield model is capable of providing the yield estimates at smaller area 1 units, such as within distributary command.