Abstract: Abstract: The quantity and quality of cultivated land plays an important role in our country's food security and the long-term stability of the community. It is important to assess the change of cultivated land quality under the changed cultivated land use pattern. This study aims to explore an effective method for detecting the changed information of cultivated land quality at county-level. Songyang County, Zhejiang province was selected as the study area to evaluate the cultivated land quality based on change vector analysis. The evaluation method was supported by Remote Sensing data and spatial analysis of GIS. There are obvious differentiation in space of cultivated land use patterns of Songyang County between 2006 and 2013. Among them, the Songgu Basin mainly planted economic crop tea, while the rice planting was scattered around the whole county. The areas where the cultivated land use pattern changed significantly were mainly concentrated in the Songgu Basin. 18 pairs of "tea-to-rice" types and 33 pairs of "rice-to-tea" were selected to analyze the changes of cultivated land quality. Combined with the characteristics of Songyang county cultivated land， the object evaluation index system was constructed from the aspects of natural factors, human factors and location factors with the principles of productivity, feasibility, typicality and dominance. In this study, the change vector was used to highlight the change information of cultivated land quality. At the same time, the proposed canonical correlation analysis (CCA)based multivariate alteration detection (MAD) method was applied to minimize the correlation effect to the greatest extent, and highlight the changing of cultivated land quality under the changed cultivated land use pattern in the form of maximizing variance. It was a feasible way for monitoring the change of cultivated land quality. In addition, expectation maximization algorithm based on Gaussian mixture model (EM-GMM) was used to label the change of cultivated land quality. The results of MAD demonstrated the change information of cultivated land quality. The closer the value of the MAD result is to 0, the smaller the difference between the cultivated land quality between 2006 and 2013. The greater the difference of the index of cultivated land quality between different phases, the more likely the quality of cultivated land changes. The variables of MAD3 and MAD1 were used to analyze the effects of changed cultivated land use pattern on the quality of cultivated land, such as rice-to-tea and tea-to-rice. The quantitative proportion of the changed cultivated land quality level was compared by using two different methods, change vector analysis (CVA) and agricultural land classification method (national grade). The agreement between two methods in monitoring the quality of cultivated land was more than 80%-100%, the matching degree of increased cultivated land quality was 16.6%-50%, and the matching degree of declined cultivated land quality was66.7%-100%. Results demonstrated that the proposed change vector based change detection of cultivated land quality provide a quantitative data support for the rational optimization and allocation of cultivated land use pattern, which is significantly maintaining the stable quality and the sustainable development of cultivated land. Further research is needed on other scales, such as city, province. The proposed method should be further tested in other regions containing more complex conditions.