On a cold morning when the mud puddles are all frozen, step in one: the ice is all on top with nothing but air underneath. Of course the puddle commences freezing at the top, but it’s hard to visualize how the ice on the top somehow draws all the water up to itself.
Why is the ice not at the bottom of the puddle?
the surface of the puddle freezes. water molecules have to be in contact to make ice, the liquid is not drawn up. the liquid water is drunk by sea monkeys in the puddle. other liquid water might drain into the ground during the rest of the night.
Ice floats and evaporates slower than the remaining water.
The rest of the water has flowed away, been absorbed into the ground, or evaporated.
Note that this doesn’t happen in a truly large puddle (aka pond, lake). There, the ice hasn’t enough strength to support itself over a large span, and must float on the water, even when the water level drops.
Unlike most substances, water actually expands when it freezes. Thus, ice is slightly less dense than liquid water, it floats. In fact, IIRC my high school physics/chemistry, water begins expanding as it gets colder at about 4C, and this continues to about -40C or so.
At normal summer temps, about 70F/20C or so, the hot layer expands, is less dense, stays warm, and the cooler, lower denser water stays in the bottom. A cup of coffee cools at the top (exposed to cool air, evaporation) and the cooled coffee/water then drops to the bottom of the cup. with milk or cream added, you can see these convestion cells, a tiny version of the same sort of effect that drives weather in the atmosphere, hot rising, cool descending.
Once you get to 4C or below, what you see is the surface is the coldest; once the temperature drops below freezing, the surface layer gets colder and colder. At 0C (32F) the freezing point, as it loses heat, it solidifies. Water at freezing (0C) needs to lose (IIRC) 80cal/gram to transition from liquid to solid. Obviously, the colder the air, the faster this happens, since it loses this heat by heating the air it is in contact with. Once it is frozen, the surface layer can continue to lose heat and decrease in temperature below 0C/32F.
How deep the freezing goes depends on how cold the air above is, and for how long. The boundary layer between ice and water may grow, by transferring heat (that 80cal/g) to the ice, where by conduction this heat works its way up to the surface and out into the ambient air above. As the water underneath loses this heat, eventually it freezes.
If there was air in the water, and it undissolves once trapped under the ice, it is not uncommon to get air bubbles.
As for bigger air pockets, the water could soak into the ground or run away after the surface ice has formed.
Ice is a poor conductor of heat, and also the 80cal/g “heat of fusion” is quite high compared to most materials. This is why ice cubes don’t necessarily melt really fast in cold water. In a small puddle, there is a secondary effect that the ground underneath could lose heat by infrared radiation, ice and water being somewhat transparent, and heat from below when the sun comes up. In deeper ponds and lakes, the heat from the surround ground could also help to keep the lower depths of the pond liquid. Regardless, leave a shallow puddle outdoors on a really cold night and after a while it will freeze all the way through.
In northern Canada, where temperatures might reach -40C/F, and stay there for weeks on end, it is not unusual to have ice form up to 3 or 4 feet thick on the lakes. (There’s a series on one of the channels about ice road truckers in the far north).
Concur - when puddles on non-absorbent surfaces freeze over, they tend to remain full up to the underside of the ice - it’s only generally puddles on muddy earth and otherwise slightly-draining surfaces that end up with a void under the ice surface.