When the Lake Closed
For most of the nineteenth century Lake Champlain froze over almost every winter, and somebody wrote down the day it happened. The record runs unbroken from 1816 — the year without a summer — to this past February, which makes it one of the longest continuous observations of anything in the northeast. It is kept as a plain table on a weather-service page. Drawn out, it is a climate record you do not need a model to read.
Watch a winter
The bed of the lake, north to the right, colored by depth — palest in the shallows, deepest blue down the four-hundred-foot (122 m) trench. Ice grows where the water is shallow and gives up last where it is deep, which is why the two ends and the bays go first and the broad lake off Burlington goes last, or not at all. Pick a winter and run it.
What “closed” means
Not that there was ice — there is ice most years, in the bays. Closed means the lake shut: historically, that commercial navigation stopped, and later that ice could be seen running continuously from Burlington west across to Port Kent. Since 2009 it has been called from satellite imagery. That is three different instruments over two centuries, and the record states as much; the seam between an ice-bound ferryman and a satellite is the one real weakness in the series.
It does not much matter here. The thing the chart shows is not a five-day shift that a change of method could explain. It is the difference between a lake that closed every winter for fifty years running and one that has closed nine times in twenty-six. The median closing date moved eighteen days later over the same stretch.
By half-century
| years | closed | of | median date |
|---|
Two and a half weeks later, and two thirds fewer winters in which it happens at all. Both changes are almost entirely after 1950.
Can any of it be predicted?
Bates, by way of Myer and Gruendling, gives a threshold for this lake: about 240 to 280 freezing degree days for a permanent freeze in Shelburne Bay. A model built on that idea is rebuilt here on every data run, fitted on the record above and scored on winters it was not fitted on. It does not clear its bar, and so it forecasts nothing.