Conservation is the cognitive understanding of whether or not the amount of a given (inert) solid or liquid will change despite its presentation. For example, human toddlers don't grasp conservation, and will erroneously believe that pouring water from a tall glass into a wide bowl makes more water, since it "looks" bigger in the wider container (e.g. it takes up more visual space in a wider vessel, and therefore must be "more"). No other animals possess this skill either (thank you Wikipedia ).
Combine that with the nebulous nature of gas - it's not intuitive to understand that an empty glass is, in fact, full of air - and understanding phase changes become that much more difficult.
It's kind of like how most people won't intuitively understand how discrete phases can break down at higher temperatures and pressures, or the differences between different lattice configurations of the same substance. If it's not something we encounter every day, or aren't taught at an early age, the intuition simply won't be there.
Anyway my prediction would be that many colder-climate animals have some intuitive grasp of the ice to water transition specifically because it’s so important to their lives. I would not expect them to understand water to steam since it’s not something they’re likely to encounter — even those who live near geysers have no reason to know that steam is actually water, they just need to know they should stay away from it. Many animals would understand the relationship between clouds and rain, though.
I wouldn’t expect them to understand phase changes in any material other than water, since it’s hard to think of examples where phase changes actually occur in animals’ natural habitats.
Alison Gopnik's The Philosophical Baby is full of remarkable experiments to determine what babies understand. (At least, I assume it is! I haven't read it - just saw her give a talk about the book and her research when the book came out, and she talked about lots of different experiments.) Stuff like showing babies white and black balls going in a container, then when one is picked out, the babies look longer at it if they're surprised, if it's an unlikely event. They have much better models of probability younger than anyone thought. Or they show them animations of dots shaped like Momma dot and baby dot, and see how the babies react when various things happen to the baby dot.. Totally ingenious. A lot of them, or something similar, could be done with animals; I wonder how much has.
Anyway, the first thing that came to me mind was "Causal understanding of water displacement by a crow" https://www.youtube.com/watch?v=ZerUbHmuY04
Therefore, I doubt greatly that animals would ever understand that concept. They might catch a glimmer of the truth, in the ice/water model, but not the whole concept.
I bet people were heating up ice to melt it to get water way before that...