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

Area of Science:

• Ecology
• Phenology
• Citizen Science

Background:

• Large-scale citizen science data and local intensive studies offer different trade-offs for ecological research.
• Phenology studies often face choices between these data sources, impacting spatial scale, observer variance, and interannual variability.

Purpose of the Study:

• To compare phenology models built using citizen science data from the USA National Phenology Network (USA-NPN) and intensive local data from Long Term Ecological Research (LTER) sites.
• To evaluate differences in parameter estimates, phenological event predictions, and out-of-sample errors between models derived from the two data sources.

Main Methods:

• Statistical and process-based phenology models were developed for common species using both USA-NPN and LTER datasets.
• Parameter estimates, phenological event dates, and out-of-sample errors were compared between models built from each data source.

Main Results:

• Model parameter estimates for the same species showed high similarity with simple models but diverged significantly with increased model complexity.
• Estimates of phenological event dates and out-of-sample errors were consistent across both data types and model complexities.
• Models performed best when predicting data from the same source they were built from (USA-NPN models for USA-NPN data, LTER models for LTER data).

Conclusions:

• The choice of data (USA-NPN vs. LTER) depends on the research question: LTER data is best for species-specific requirements, while USA-NPN data excels at large-scale predictive modeling.
• Models built from USA-NPN data integrate parameters across large spatial scales, whereas LTER models capture fine-scale interannual variability.
• Future research should integrate strengths of both data types for robust, large-scale phenology forecasting.