Estimating flowering transition dates from status-based phenological observations: a test of methods

Area of Science:

• Ecology
• Botany
• Data Science

Background:

• Phenological research is expanding with digitized herbarium data and citizen science contributions.
• Status-based phenological observations (presence/absence of phenophases) require methods to estimate transition dates.
• Accurate transition date estimation is vital for modeling plant life cycles from dormancy to reproduction.

Purpose of the Study:

• To compare the accuracy of eight different estimators for plant phenological transition dates.
• To evaluate estimators at both population and individual plant levels using a high-frequency dataset.
• To identify optimal methods for utilizing large-scale status-based phenological data.

Main Methods:

• Monte Carlo analysis was employed to test eight estimators.
• A dataset of in-situ flowering observations for *Echinacea angustifolia* with known transition dates (±3 days) was utilized.
• Estimators were tested across varying sample sizes and proportions of flowering presence.

Main Results:

• The Weibull distribution method demonstrated superior accuracy for population-level onset and end of flowering estimates.
• Alternative estimators may be more suitable when absence observations significantly outweigh presence observations.
• For individual plant estimates, a method using the midway point between first flower presence and prior absence (within 7 days) was optimal, provided sample size was not limited.

Conclusions:

• The Weibull distribution method is a robust choice for population-level phenological transition date estimation.
• Specific methods are recommended for individual plant estimates based on data characteristics and sample size.
• These findings enhance the utility of extensive status-based phenological datasets for ecological research.