I think you've got it. The only option would be to create a material where you "fake" the water loss using a property vs temperature curve. You might be able to get away with the rule of mixtures if you know what the properties of the 2 materials are at the 2 temperatures and the volume/mass fraction of each at those points.
To get a realistic model, you're going to need a physical experiment to validate your method. But once you've done that, you should be able to use the technique on future models.