Climate calibration of the Spring Index model for more accurate broad-scale first leaf predictions

Phenological models are needed for forecasting plant and ecosystem responses to climate change. Due to a lack of considering local adaptation induced variations in climatic requirements of plant species for phenological development, traditional uniform/non-spatial models that cover broad geographic regions are susceptible to systematic prediction biases. This study presents a climate calibration method that incorporates climate adaptation patterns of plant species into a widely used Spring Index (SI) First Leaf (FL) model. Multi-year (2009-2021) phenological observation data for a most frequently observed shrub species(common lilac Syringa vulgaris) and a most frequently observed tree species(red maple Acer rubrum) in the eastern USA from the USA-National Phenology Network (USA-NPN) were used to develop and validate the calibrated models. Climatic gradients defined by latitudinal temperature variations were used to predict varied climatic requirements of the populations of each species. Prior to calibration, SI FL predictions showed consistent geographic biases and yielded large prediction errors (especially for red maple, RMSE = 30 d). Calibrated SI FL predictions yielded reduced errors (e.g. RMSE = 16 d for red maple) and were freed from significant geographic biases (α = 0.05) in all cases. The calibration method accounted for both intraspecific and interspecific variations, leading to more accurate broad-scale first leaf predictions for the species tested. The climate-calibrated SI FL allows for more accurate tracking of the onset of spring over extensive geographic areas and would support spatially explicit natural resource and environmental conservation efforts under climate change.