Yesterday's post on snow cover and temperature prompted an interesting question from reader Eric about whether the data really showed what I was claiming, which was that deeper snow cover in Fairbanks at this time of year causes lower temperatures, and conversely a lack of snow cover causes warmer conditions. Eric suggested that the data pointed to different air masses and their temperatures as the leading cause of snow variations, rather than implying that snow has a pronounced influence on temperature. As I mentioned before, this is a chicken and egg problem: which comes first, snow or cold temperatures?
To help untangle the chain of causality, I repeated the analysis with 850 mb temperatures rather than surface temperatures. The idea here is that 850 mb temperature will not be influenced by the hypothesized low-level cooling from snow cover; and therefore if 850 mb temperature shows the same connection with snow depth, then the correlation is more likely caused by weather regime variations, i.e. a pattern with warm air drawn up from the south will favor less snow, but a cold pattern will favor more snow. The chart below shows the results, which remarkably (to me) are extremely similar to the original chart for surface temperature (included below for ease of comparison):
The result seems clear: year-to-year snow depth variations in Fairbanks at this time of year are just as closely connected to the 850 mb (free atmosphere) temperature as to the surface temperature. Warm air aloft is tied to a lack of snow, and cold air aloft is associated with deeper snow cover.
To dig in a little deeper, I examined the connection between snow depth and the difference between surface temperatures and 850 mb temperatures. This is useful because if deep snow cover causes cooling, then the surface temperatures ought to be relatively low compared to the 850 mb temperature; but if snow doesn't cause cooling, then the vertical temperature difference would not depend on snow depth. Remarkably, the latter is true - see the chart below. There seems to be no significant connection between the vertical temperature difference and snow depth; the downward trend is just the climatological tendency for strengthening inversion as the sun goes away. One could argue there is a small separation at the end of October between no-snow and deep snow, but it's a small and no doubt statistically insignificant difference.
In summary, the updated results show that deeper snow in late autumn and early winter is not significantly associated with surface cooling relative to temperatures aloft, and the correlation between snow depth and temperature extends to temperatures well above the surface. And so in terms of causation, it appears that snow depth responds to the large-scale weather pattern and is not itself a major influence on Fairbanks temperatures; this is just what Eric suggested.
I'd be glad to hear suggestions of other ways to look at this problem, or other physical mechanisms that could be at play to explain the data.