Comparison of large-scale citizen science data and long-term study data for phenology modeling

The USA National Phenology Network provides a relatively new phenology dataset with a lot of promise. But there are potential issues that come with all citizen science datasets. Before using the dataset in my research I decided to do a comparison with older LTER datasets to see if they gave the same answers to questions. I built process based phenology models for specific species twice. Once using the USA-NPN dataset and again using LTER data. From those models I then compared their parameters, predictions, and errors. Parameters were very different for most models between the two datasets, and also between the same species within LTER datasets. Predictions and errors were surprisingly similar between USA-NPN and LTER derived models, though the best performing model was always one that had been built using the same dataset which was being tested.

Available open access at https://doi.org/10.1002/ecy.2568.
Preprint
Code