The "January thaw" is a statistical phantom

The existence of the purported “January thaw,” i.e., systematic anomalous warming in average daily temperatures at northeastern U.S. stations during the last week of January, is investigated quantitatively. A key idea in this investigation is that winter temperatures are intrinsically more variable, and this property must be accounted for when judging unusualness of excursions of daily temperatures. Thus average temperatures for the date are expressed nondimensionally by dividing by daily temperature standard deviations that vary through the year. The warm excursion in observed records for late January is rarely the most extreme such excursion in the nondimensionalized data, even when the definition of “excursion” is optimized to emphasize the late January “event.” Hypothesis tests based on sampling distributions generated using time-series models with smoothly varying climatologies (i.e., no anomalous features such as the “thaw”) are used to quantitatively evaluate the statistical significance of the observed “January thaws.” The synthetic series produce many apparent events of similar magnitude, although occurring randomly throughout the year and equally divided between warm and cool deviations. It is thus concluded that the observed “January thaw” is a mere sampling variation with no statistical or physical significance.