Abstract: Objective To understand the influence of site quality in the remote sensing Chinese fir biomass estimation. Method Based on the forest resource management inventory data and TM image of Jiande city obtained in 2007, the biomass of Chinese fir was calculated by forest volume-biomass conversion factor continuous function method and the site quality was evaluated by site class method. Four biomass estimation models for different site classes (good, moderate, poor and no ranking) were compared and the accuracy of them was tested. Result (1) The performance of regression model based on the first principal component analysis of TM remote sensing image is the best, the determination coefficients R2 is higher than 0.69 and the maximum is 0.855. (2) Verifying the model accuracy by reserved independent samples, the whole model accuracy without site class is 87.78% and the accuracies of good, moderate, poor site quality models are respectively 97.37%, 95.82%, and 98.23%. Conclusion Distinguishing different site qualities could improve remote sensing estimation precision of Chinese fir biomass. The research results provide with an improved method for the remote sensing estimation of forest biomass, and a reference for improving the remote sensing estimation accuracy of forest biomass and carbon storage.