Supercooled fog and instrument icing
Liquid water can persist well below zero if there is nothing for it to freeze onto. In the atmosphere this is routine: fog droplets at -5C or -10C are commonly still liquid, because freezing needs an ice nucleus and effective nuclei are scarce.
Introduce a surface and the situation resolves immediately. This is why supercooled fog ices everything it touches — instrument masts, guy wires, aircraft leading edges, anemometer cups. The accretion is one-sided and grows upwind, and on a cup anemometer it changes the mass distribution enough to bias readings low before it stops rotation entirely.
Field stations in cold, humid, cloud-immersed sites deal with this constantly. Heated sensors work but consume power. Some designs shed ice mechanically. Others accept the outage and mark the data suspect, which is at least honest, and is why cold-region wind records tend to have seasonal gaps that are not random.
The related hazard for anyone outdoors is that supercooled fog can glaze rock and boardwalk in minutes with a coating that is nearly invisible.