PhenoFormer Dataset

Companion Dataset for the Article  
"Deep Learning Meets Tree Phenology Modeling:
PhenoFormer vs. Process-Based Models"  
Garnot et al., 2025  

This archive contains the dataset used in the numerical experiments for the article. 
It includes:  
- Phenological observations from the Swiss Phenology Network 
for 9 woody plant species across 175 sites over 70 years.  
- Daily meteorological variables from the DaymetCH dataset 
for each observation site and year.  

Dataset format:  
- learning-models-data – Formatted for use with Python code 
for deep learning and machine learning models.  
- process-models-data – Formatted for use with R code 
for process-based models.  

Code repository:  
https://github.com/VSainteuf/PhenoFormer  

Citation:  
@article{phenoformer,  
  title={Deep learning meets tree phenology modeling: PhenoFormer vs. process-based models},  
  author={Garnot, Vivien Sainte Fare and Spafford, Lynsay and Lever, Jelle and 
  Sigg, Christian and Pietragalla, Barbara and Vitasse, Yann 
  and Gessler, Arthur and Wegner, Jan Dirk},  
  journal={{Methods in Ecology and Evolution}},  
  year={2025}  
}  

Data source and processing:  
- Data source: Federal Office of Meteorology and Climatology (MeteoSwiss)  
- Meteorological data processing: Swiss Federal Institute for Forest, 
Snow and Landscape Research (WSL)