Based on the relationships between dry matter partitioning index, harvest index, and product of thermal effectiveness and PAR, a simulation model for greenhouse tomato dry matter partitioning and yield prediction was built, and validated by independent experimental data of different cultivars, substrates and locations. The coefficient of determination (R2) between simulated and measured shoot, root, stem, leaf and fruit dry matter weight based on 1:1 line was 0.95, 0.57, 0.82, 0.79 and 0.93, the root mean squared error (RMSE) between them was 647.0, 78.1, 279.0, 496.9 and 381.8 kg x hm(-2), and the R2 and RMSE between predicted and measured yield based on 1:1 line were 0.88 and 5 828.5 kg x hm(-2), respectively. Compared to 'source-sink' theory, the model developed in this study could give satisfactory prediction of the dry weight of leaf, stem, fruit and yield, with fewer parameters that could be easily obtained in practice.