What I'm sure of here is what you can easily observe: lakes with a
layer of ice on top of liquid water. In that very common case, the top
layer obviously has frozen before the lower layers, although I don't
know what's happened on the bottom in the middle of the lake.
argument might sound reasonable, but it can't trump direct observation.
And in fact there are some flaws in it. When water gets almost down to
freezing temperature, it actually expands rather than contracts on
further cooling. So very near freezing, it's the COLDEST water which
would tend to rise to the top. In addition there are all sorts of
complications which I don't know much about, such as different salt
concentrations in different parts, comparative rates of freezing and of
temeperature-driven circulation, etc.
That's why it's nice to
know the answer by observation. And this is an important answer- if you
thought the process went the other way, you'd assume that when the top
was ice, it was supported by ice all the way down. Many people have
died from making that assumption.
Just to add a
couple of details: Most lakes freeze over because the ambient air
temperature gets below freezing. That, combined with evaporation from
the surface and radiation of heat energy to space makes heat loss from
the surface greater than heat loss through the bottom. In fact, because
the earth generates thermal energy due to nuclear decay, the ground
warms up the water rather than cooling it off, at least in the
wintertime. So the ice forms on the surface because that's where the
heat loss is, rather forming first on the bottom and then floating
(republished on 07/25/06)