I have several temperature variables that are predictor variables on survival. I am trying to differentiate between them to see if one temperature variable has a greater effect on survival than another. As expected (b/c it's the same temperature time series) several of my temperature variables are correlated. I have read about the problems with multicollinearity, but are there any problems with separating the correlated predictors into different models and then using AIC to identify better/worse performing models? Thank you.