Delimiting ecological red lines is the key to answering spatial-allocation questions for land resources. At present, most studies on the delimitation of ecological red lines have been based on the evaluation of suitability conditions of ecological land, without considering the process of historical change. Therefore, it cannot be guaranteed that the zoning results will adapt to the trends in ecological land change in the area, leading to frequent designation and adjustment of ecological red lines. In this study, to guarantee the stability of ecological red lines, we examined a delimitation method for ecological red lines based on a Bayesian network model, which included both factors concerning ecological suitability conditions and dynamic change of the land. Next, the proposed model was applied to E-Zhou, a city near the mid-reaches of the Yangtze River. In this proposed model, ecological suitability factors, which represented the quality of the ecological land, and dynamic change factors, which illustrated the history of the ecological land, were obtained. Using the ecological land potential value as the target variable, we defined the structure of the network using expert knowledge. The Bayesian network was trained by the maximum likelihood method with 20000 random sample points. The results showed that only 65.5% of ecological land remained stable from 2004 to 2013 in E-Zhou City. For the two dynamic factors, the influence of farmland occupancy was the greatest, and accounted for 43.6%, whereas the influence of urban encroachment accounted for 10.2%. The results of the sensitivity analysis also indicated that farmland occupancy had the greatest influence on the potential value of the ecological land, with the highest variance reduction of 29.5%, followed by eco-environmental sensitivity, and the importance of ecosystem services, which exhibited variance reductions of 8.55% and 1.84%, respectively. The variance reductions for the distance to a water body and the ecological protection of a forest were greater than that of the distance to a road or a railway, change in traffic facilities, and urban construction, which all had little effect on the ecological land potential. A causal link between influence factors and the target variable was obtained during backward propagation in diagnostic analysis. Under the condition that the value of ecological land potential was "yes", the probability of the "extremely important" factors of ecosystem services and water conservation increased by 6% and 2.7%, respectively. This indicated that the contributions of ecosystem service values of a forest ecosystem and aquatic ecosystem increased. Furthermore, the probabilities of "highly sensitive" and "extremely sensitive" factors of ecological sensitivity decreased by 13.4% and 1.8%, respectively. This verified the improvement of the natural condition of the land by ecological land protection. Then, using the values for the ecological suitability factors in 2013, we obtain the relative potential value of each ecological land parcel as the ecological red lines through forward reasoning of the trained network, and delimiting ecological land with the potential value of the target variable. The results showed that this delimiting method worked in accordance with the trends in urban development in the area and spatial distribution of ecological land in E-Zhou City. This method can improve the stability of ecosystem services, as well as ensure the quantity and quality of ecological red lines. This new and effective model could serve as a support for other cities.