Human activities such as overfishing, coastal development and pollution have altered bay ecosystems and eroded their capacity to provide benefits now and in the future. A systematic diagnosis of bay ecosystem health for sustainable ecological management is urgently needed. Sustainable management aims at maintaining the flow of a broad range of benefits from the bay and also requires a comprehensive and quantitative method to measure and monitor the health of bay at the level of ecosystem. A variable fuzzy evaluation model is introduced for the comprehensive evaluation of bay ecosystem health. According to the environmental pollution conditions and ecosystem characteristics in Dongshan Bay, Fujian, China, an integrated index system comprising water, sediment environment, residual toxicity in organisms and marine biota were created for evaluation of bay ecosystem health. The judgment criteria regarding the level of bay ecosystem health could be roughly classified as "excellent, good, moderate, poor, and bad". The five levels could be further sub-classified into 13 ranks. Ecosystem health assessment of Dongshan Bay in spring and autumn based on variable fuzzy evaluation model was presented in this paper. The results showed that the score of ecosystem health index for Dongshan bay was 2.36 in spring and 2.44 in autumn, respectively. The status of Dongshan bay ecosystem health could be generally classified as "between good and moderate, mostly good", with better level in spring than in autumn. In more detail, the scores of ecosystem health index ranged from 2.31 to 2.47 and all stations were classified as "between good and moderate, mostly good" in spring. In autumn, the scores of ecosystem health index ranged from 2.29 to 2.74. Only three stations (D11、D13 and D14) were classified as "between good and moderate, mostly good" and the other stations were classified as "between good and moderate, mostly moderate". Several factors showed their positive effects on the ecosystem heath status, such as heavy metals in seawater, sediment and organism, oil in seawater and sediment, zooplankton biomass, phytoplankton cells density and diversity as well as benthos diversity. The scores of health index for these indicators were all lower than 1.5 in spring and autumn. However, the health status of the ecosystem was potentially made to change to "between good and moderate, mostly moderate" by the negative factors including fish eggs and larva density (average score of heath index in spring and autumn was 4.95), nutrition level (the score of heath index in autumn was 4.47), and benthos biomass (the score of heath index in spring was 3.59). The scores of health index for these three indicators were all higher than 3.5. Case study indicates that the ecosystem health assessment based on variable fuzzy evaluation model also provides a standardized and quantitative approach for measuring overall health condition of bay ecosystems, which can perform well and accurately evaluate the comprehensive status of the bay ecosystem heath through the change of variable model parameters (the criterion and distance parameter, α and p), combining the linear and nonlinear models. Therefore, the model result is more reliable than before.