Site index is an effective method that has been widely used to evaluate site quality at present. The study of the correlation among site indexes of different tree species stands and establishment of correlation model of different site indexes in the same site for several tree species could be useful for predicting one site index (dependent variable) on condition that the other site index (independent variable) is known, and is propitious to evaluate site quality. For example, when a Pinus massioniana stand would be harvested for some reasons and replanted by Cunninghamia lanceolata, we should predict the growth potential of Cunninghamia lanceolata stand. The correlation model of site indexes of Cunninghamia lanceolata and Pinus massioniana provides a helpful routine to solve this issue. Firstly, the correlation model of site indexes of Cunninghamia lanceolata and Pinus massioniana should be established, and then the site index of Pinus massioniana site based on the investigation appraoch should be obtained. Finally, the site index of Cunninghamia lanceolata could be predicted. In this study, we used the site index data of Cunninghamia lanceolata and Pinus massoniana collected at the same plots in Xuefeng mountain to establish three linear models (i.e. general linear model, dual regression model and measurement error model, respectively).The precision and accuracy tests of three model were performed and the efficiency of these models was compared.The results show that the precision of three models is high.The relative error was 5.39% for general linear model, 5.45% for dual regression model and 5.39% for measurement error model, respectively. Both the Cunninghamia lanceolata Site-index and Pinus massoniana Site index yielded measurement error. The linear measurement error model and dual regression linear model are more appropriate than general linear model，because that both independent variable and dependent variable had the measurement error in measurement error and dual regression models，no measurement error was found for the dependent variable of the general linear model.Furthermore，the relative error of dual regression model was higher than that of measurement error model，suggesting that linear measurement error model is more appropriate than the other models.This result showed that when Cunninghamia lanceolata site index (or the Pinus massoniana site index) is known in one site, site index of another tree species could be predicted by using the linear measurement error model. Correspondingly, when the growth ability of Cunninghamia lanceolata stand (or Pinus massoniana) is known, the growth potential of another tree species stand could be estimated. The method for evaluating the site quality of Cunninghamia lanceolata stand and Pinus massoniana stand would be also effective for other trees species as well.