One man, 73 years, and 25 species. Evaluating phenological responses using a lifelong study of first flowering dates

K. Bolmgren, D. Vanhoenacker, A. J. Miller-Rushing

Phenological shifts linked to global warming reflect the ability of organisms to track changing climatic conditions. However, different organisms track global warming differently and there is an increasing interest in the link between phenological traits and plant abundance and distribution. Long-term data sets are often used to estimate phenological traits to climate change, but so far little has been done to evaluate the quality of these estimates. Here, we use a 73-year long data series of first flowering dates for 25 species from north-temperate Sweden to evaluate (i) correlations between first flowering dates and year for different time periods and (ii) linear regression models between first flowering date and mean monthly temperatures in preceding months. Furthermore, we evaluate the potential of this kind of data to estimate the phenological temperature sensitivities (i.e. number of days phenological change per degree temperature change, β60) in such models. The sign of the correlations between first flowering dates and year were highly inconsistent among different time periods, highlighting that estimates of phenological change are sensitive to the specific time period used. The first flowering dates of all species were correlated with temperature, but with large differences in both the strength of the response and the period(s) of the year that were most strongly associated with phenological variation. Finally, our analyses indicated that legacy data sets need to be relatively long-term to be useful for estimating phenological temperature sensitivities (β60) for inter-specific comparisons. In 10-year long observation series only one out of 24 species reached ≥80 % probability of estimating temperature sensitivity (β60) within a ±1 range, and 17 out of 24 species reached ≥80 % probability when observation series were 20 years or shorter. The standard error for β60 ranged from 0.6 to 2.0 for 10-year long observation series, and 19 out of 24 species reached SE < 1 after 15 years. In general, late flowering species will require longer time series than early flowering species.