Aim: To investigate temperature drivers of the spring phenology of 12 flowering events and 6 leafing events. Boreal phenology has previously exhibited only a modest response to temperature but record breaking temperatures in March 2020 led to some extreme phenological timing and provided the opportunity for a more rigorous look at the nature and complexity of the relationship between temperature and phenology.

Location: Boreal habitat in the Tatarstan Republic of Russia

Time period: 1989-2020

Major taxa studied: 18 plant species

Methods: We examined for changes over time in the timing of phenological events and the relationship with temperature drivers using a range of regression techniques. Extreme temperatures caused some extreme phenology and visually suggested temperature thresholds which were investigated using segmented regression (broken stick models). The performance of these models was subsequently compared to more complex alternatives based on accumulated daily temperatures.

Results: Temperatures in March 2020 were the warmest for this month in a record covering 200 years and were 5.9°C above the 1989-2020 average. Significant advances over time were detected for only seven of the events, but all events demonstrated significant relationships with temperature variables. Segmented regression identified significant temperature thresholds for 13 of the events with substantially stronger temperature relationships above these thresholds. Segmented regressions outperformed models based on accumulated daily temperatures for 12 of 18 events.

Main conclusions: The threshold models were a significant improvement over a linear response to temperature for 13 of the 18 events. The presence of thresholds partly explains why the response of boreal phenology to temperature previously seemed lower than in, for example, Western Europe. Responses to temperature above the identified thresholds were closer to those widely published from milder locations.

Dates of first observed flowering or leafing on study transects.

No missing values

Data are days of the year, e.g. Jan 1 = day 1