This study is intended to investigate a way of considering temperature and vehicle weight as environmental factors for long-term bridge health monitoring by means of the Bayesian regression. Long-term monitoring data observed at a seven-span plate-Gerber bridge were examined. Coefficients of the autoregressive model of accelerations taken from the bridge are used as a damage-sensitive feature for health monitoring of the bridge. Observations showed that the Bayesian regression considering both temperature and vehicle weight led to more accurate results than those considering only temperature. Validity of using the data observed at a specified time to reduce the influence of traffic can be confirmed.