Annual-to-decadal variability in northern midlatitude temperature is dominated by the cold season. However, climate field reconstructions are often based on tree rings that represent the growing season. Here we present cold-season (October-to-May average) temperature field reconstructions for the northern midlatitudes, 1701-1905, based on extensive phenological data (freezing and thawing dates of rivers, plant observations). Northern midlatitude land temperatures exceeded the variability range of the 18th and 19th centuries by the 1940s, to which recent warming has added another 1.5 °C. A sequences of cold winters 1808/9-1815/6 can be explained by two volcanic eruptions and unusual atmospheric flow. Weak southwesterlies over Western Europe in early winter caused low Eurasian temperatures, which persisted into spring even though the flow pattern did not. Twentieth century data and model simulations confirm this persistence and point to increased snow cover as a cause, consistent with sparse information on Eurasian snow in the early 19th century.

<p>Annual-to-decadal variability in northern midlatitude temperature is predominantly dominated by the cold season. However, climate field reconstructions, which are essential for understanding the underlying mechanisms, are often based on tree rings. These mainly represent the growing season and allow limited insight on cold season effects. Plant and ice phenology data, on the other hand, are a rich source of cold season information that remains largely overlooked in climate reconstructions to date and could help to fill the seasonal gap. Here, we present Northern Hemispheric temperature field reconstructions for the extended cold season (October-to-May average) for 1701-1905 based entirely on phenological data. Time series of freezing and thawing dates of rivers together with a few early-spring plant observations covering a large area of the northern midlatitudes are used in a simple data assimilation framework.<br>The reconstructions allow a 320-yr perspective of climate variability and change of boreal cold season climate and unveil that the temperature of the northern midlatitude land areas exceeded the variability range of the 18th and 19th centuries by the 1940s, to which recent warming has added another 1.5 °C. We also find 5-10 year long sequences of cold northern midlatitude winters. The most prominent example lasted from 1808/9 to 1815/6. The conspicuously cooling during that period is associated with two volcanic eruptions (1808/9 and 1815), which caused cooling as a direct effect. The years between the eruptions are characterized by weak southwesterly atmospheric flow over the Atlantic-European sector in early winter. This lead to low Eurasian temperatures, which persisted into spring while the flow pattern did not. Twentieth century data and model simulations confirm this persistence and point to increased snow cover as a cause. This is consistent with independent information on Eurasian snow in the early 19th century.</p>