This is due to the way we do reflections. At the moment we do reflections by inverting the camera through the water plane and then redrawing the scene. This is not particularly cheap CPU or GPU wise so we only do it at sea level since that is where most of the water is. Doing it for multiple water levels would require us to redraw the whole scene for each height of water. So, if the water is not at sea level you won’t get reflections.
In the future we may add screen space reflections which uses what you can already see to guess at what the reflection would look like. This is much quicker but doesn’t look as good because anything in the reflection you cannot already see will be missing.