The outbreak of H7N9 epidemic in human has seasonal changes. However, up to now there is no research on the spatial-temporal variation characteristics of the relative risk of human H7N9 infection, and the covariate combination that has a greater impact on the relative risk of human H7N9 infection in different seasons. This study used China as the study area to predict the seasonal relative risk of human H7N9 infection through a Bayesian hierarchical conditional autoregressive model (BHCAR), which including five covariates (population density, number of live poultry markets, average precipitation, average temperature, and average relative humidity), seasonal random effects, and spatial random effects. Moreover, the sensitivity of the Bayesian hierarchical model (BH) to predict the seasonal relative risk of human H7N9 infection by changing the parameter settings of the BH prior distribution was analyzed. It was found that the relative risk of human H7N9 infection in spring and winter had spatial random effects, but not in summer and autumn. In spring, autumn and winter, the combination of population density and the number of live poultry markets had a significant influence on the relative risk of human H7N9 infection. In summer, however, the relative risk of human H7N9 infection was largely affected by population density, the number of live poultry markets, average precipitation and average temperature. Further, the standard deviation of the normal distribution to which the covariate coefficient in the BH was subject seemed to have an influence on the prediction and fitting effect of the seasonal relative risk of human infection with H7N9.