Practice makes the expert: The importance of training volunteers in the generation of phenological data from photographs of biodiversity observation platforms.

Phenology studies the time at which events in the life cycle of a species occur sand how they are related to environmental cues. Patterns of change in phenology at different scales can be used as an indicator of ecosystem changes and climate change, but the data necessary to detect these changes can be difficult to obtain due to their temporal and regional dimensions. Citizen science can contribute to generate large amounts of data on phenological changes at wide geographical scales that would be almost impossible for professional scientists to generate, but the quality and reliability of these data are often questioned. The objective of this study was to evaluate the use of a biodiversity observation citizen science platform based on photographic information as a potential source of large-scale phenological information, and to identify the key benefits and limitations of this type of information source. We used the Naturalista photographic databases for two invasive species in a tropical region: Leonotis nepetifolia and Nicotiana glauca. The photographs were classified into different phenophases (initial growth, immature flower, mature flower, dry fruit) by three groups of volunteers: a group of experts, a trained group with information on the biology and phenology of both species, and an untrained group. The degree of reliability of the phenological classifications was estimated for each group of volunteers and each phenophase. The degree of reliability of the phenological classification of the untrained group was generally very low for all phenophases. The group of trained volunteers showed accuracy levels for the reproductive phenophases that equaled the degree of reliability among the expert group, regardless of species, and was consistent across phenophases. We conclude that volunteer classification of photographic information contained in biodiversity observation platforms can provide phenological information with high geographic coverage and an increasing temporal coverage on general phenological patterns of species with wide distributions but has limited applicability in the identification of exact start and end dates. and peaks of the different phenophases.