The Performance of Growing Degree Day Models to Predict Spring Phenology of Herbaceous Species Depends on the Species' Temporal Niche

Abstract

The concept of growing degree days (GDDs) is commonly used to predict phenological events in plants, assuming that plants develop proportionally to the accumulated temperature. Two species‐specific parameters, T Base and t 0 (minimum temperature above which and start date when GDDs begin to accumulate), are considered for the calculation. However, species‐specific optimised thresholds of wild herbaceous species remain sparse, and therefore the reliability of the models is questionable. By employing several modelling approaches using phenological records of leaf unfolding and flowering onset of 87 wild herbaceous species collected in six European botanical gardens between 2019 and 2024, we assessed the reliability of GDD models across a diverse array of species. We further examined whether thresholds of T Base and t 0 for calculating GDD can be optimised for a large set of species and for single species. We aimed to estimate and evaluate these thresholds and the reliability of GDD models using species’ temporal niche and bud traits to see whether for specific groups of species, specific GDD models work better. Our analyses revealed that GDD models for leaf unfolding and flowering onset performed better than the null model (i.e. mean date across years and species) for 84% and 70% of the species, respectively. Our results showed that species with intermediate temporal niches were less dependent on the selection of T Base and t 0 . Overall, we found better performance of the models using a T Base around 4°C for most of the species. By considering optimised thresholds, we found that predictions of leaf unfolding dates were more accurate in early‐growing species, and regarding the start date for temperature accumulation, we found that larger values for t 0 are suitable for predictions for species with later leaf unfolding or flowering onset. Our results emphasise that simple temperature accumulating GDD models can be optimised by using the temporal niches of the studied species to approximate the underlying model parameters or by applying thresholds that are valid for many species. The use of simple but optimised GDD models can be advantageous for small datasets that would otherwise be overfitted with more complex models.