Damage accumulation always happens to civil engineering structures during their operations，degrades the structural performance，and may eventually lead to catastrophic event. Therefore，it is necessary to monitor the state of structures and alert dangerous behavior. However，during the long-period monitoring of structures，the changes of structural characteristics resulted from the variations of environmental factors（such as temperature）often mask the real structural damages，and lead to false warning. In this paper，a damage warning method based on Gaussian Mixture Model（GMM）clustering and cointegration（CI）theory is developed to remove the nonlinear environmental effects. This method uses the identified structural modal frequencies as cointegration variables. Assuming piecewise linear effect of environmental factors on structural frequencies，GMM is applied to cluster the frequency samples extracted from different environmental conditions. Then，the linear cointegration relationship between structural frequencies is established using Johansen test from the frequency samples in each cluster group. Finally，the damage is identified through the X-bar control chart using the cointegration residuals as damage indicators. The proposed method is applied to the in-situ test results of the Z24 Bridge in Switzerland，and the results show that the proposed method can effectively eliminate the nonlinear effect of temperature on the structural frequencies，and reduces the chance of false alert of structural damages.