Abstract: 【Objective】In this study, the influence of different site-quality types on the remote-sensing estimation of the carbon budget of broadleaved forests was analyzed. 【Mothed】Using forest resource inventory data on the management and TM images of Jiande City, Zhejiang Province, China in 2007, the carbon budget in broadleaved forests was calculated by the regional volume-biomass conversion factor continuous function method and biomass-carbon budget coefficient conversion method based on Zhejiang Province; site quality was evaluated by site class method, and then, three models for the estimation of the carbon budget of broad-leaved forests were compared and divided into “above-average”, “below-average site quality” and “no rank” groups, and the accuracy test was conducted. 【Result】The model for estimation of forest carbon storage was constructed with the first principal component as the characteristic variable; the correlation coefficient R was larger than 0.49, and performance was good. The forest carbon-storage regression model was compared using the measured values; whole model accuracy without site class was 65.89%, and accuracy of the above-average, below-average site-quality models were 82.53% and 83.09%, respectively. 【Conclusion】Discrimination between different site qualities can improve the precision of remote-sensing estimation of the carbon budget of broadleaved forests, and the results of this study provided adequate technical support for the remote-sensing estimation of forest carbon budgets.