Data Reliability in a Citizen Science Protocol for Monitoring Stingless Bees Flight Activity

Simple Summary This work aims to validate a citizen science protocol for monitoring the flight activity of stingless bees. The count of flight activity (entrance, exit, and entrance carrying pollen) filmed in 30 s videos was compared among three different groups: "original" citizen scientists (group that filmed and performed the count in their own videos), "replicator" citizen scientists (group of citizen scientists who performed flight activity counts on videos shot by other citizen scientists), and experts (researchers who work with bees and who performed the counts on videos shot by citizen scientists). The analysis was divided into two levels: perception (detection of activity in videos) and counting. The results of this analysis revealed that citizen scientists and experts have similar perception and count of bee entrance and exit activity, as no statistical differences were found in these two items. However, replicator citizen scientists noticed more bees carrying pollen than original citizen scientists and experts. Despite this, considering only the videos in which the groups agreed on the presence of pollen, the count was similar for both. These results enabled the validation of the protocol and indicated high quality of data produced by individuals who participate in scientific practices following a citizen science approach. Although the quality of citizen science (CS) data is often a concern, evidence for high-quality CS data increases in the scientific literature. This study aimed to assess the data reliability of a structured CS protocol for monitoring stingless bees' flight activity. We tested (1) data accuracy for replication among volunteers and for expert validation and (2) precision, comparing dispersion between citizen scientists and expert data. Two distinct activity dimensions were considered: (a) perception of flight activity and (b) flight activity counts (entrances, exits, and pollen load). No significant differences were found among groups regarding entrances and exits. However, replicator citizen scientists presented a higher chance of perceiving pollen than original data collectors and experts, likely a false positive. For those videos in which there was an agreement about pollen presence, the effective pollen counts were similar (with higher dispersion for citizen scientists), indicating the reliability of CS-collected data. The quality of the videos, a potential source of variance, did not influence the results. Increasing practical training could be an alternative to improve pollen data quality. Our study shows that CS provides reliable data for monitoring bee activity and highlights the relevance of a multi-dimensional approach for assessing CS data quality.