Overlooked climate parameters best predict flowering onset: Assessing phenological models using the elastic net

Area of Science:

• Ecology
• Botany
• Climate Change Biology

Background:

• Understanding plant phenology shifts is crucial for predicting climate change impacts on flora.
• Existing phenological datasets often lack diversity and duration, hindering accurate predictor identification.
• Previous studies used limited climate parameters, potentially missing key drivers of plant phenology.

Purpose of the Study:

• To assess the impact of 25 climate parameters on flowering time across 2,468 North American angiosperm taxa.
• To compare the predictive power of phenological models using herbarium records versus in situ observations.
• To introduce elastic net regularization as a novel regression approach for phenological modeling.

Main Methods:

• Utilized 894,392 digital herbarium records and 1,959 in situ observations.
• Applied elastic net regularization, a regression technique, for predictive modeling.
• Compared model performance using herbarium-derived and in situ phenological data.

Main Results:

• Phenological models from both data sources showed similar predictive capacity (R² = 0.27).
• Key predictors of flowering time, besides mean maximum temperature (TMAX), include the number of frost-free days (NFFD) and precipitation as snow (PAS).
• These top predictors (NFFD, PAS) have been historically underutilized in phenological models.

Conclusions:

• This study provides a comprehensive assessment of climate drivers on flowering phenology across a large number of taxa.
• The findings highlight the importance of NFFD and PAS in predicting plant flowering times.
• The research demonstrates a robust new approach to phenological modeling using extensive data and advanced statistical methods.