Citizen science is an increasingly popular method in ecology to engage the public in science participation and data gathering. However, despite the prevalence of manipulative experiments in the work of professional scientists, citizen science projects remain overwhelmingly observational. To understand the landscape factors that influence pollinator visitation and crop production differences between urban community gardens, we attempted a hybrid model, combining non-observational citizen science with a traditional manipulative experiment run by professional scientists. We distributed cherry tomato starts with three experimental pollination treatments (excluding, open, hand-supplemented) to 106 citizen scientists at 29 community garden sites in Seattle, WA, supplemented by our own 10 replicate plants at 10 selected sites. We were interested in understanding (1) if enough data could be generated by citizen scientists doing a manipulative experiment to robustly answer our experimental questions, (2) what would influence patterns of citizen scientist data submission through time, and (3) what would influence patterns of citizen scientist data submission by site.
About 20% (21/106) of participants and 50% (15/29) of sites submitted data to the project; field interns working with our replicates collected about four times the data as the citizen scientists, and with higher consistency site over site. With these participation rates, we could not have collected enough data at enough sites using citizen scientists alone. We predicted that a higher proportion of citizen scientists would submit data at least one time, but that submission would wane as the season went on. However, we found that the mean number of submissions per plant and the range of harvest dates submitted did not differ between interns and citizen scientists. Though variation in submission rate was higher for citizen scientists, these results suggested that those citizen scientists that submitted data at least once were likely to follow through with the experiment. Surprisingly, we also found that participants at sites with our potted replicates were more likely to submit data, and submitted significantly more data than participants at sites without our replicates, by about 10 submissions/plant (or roughly double). Further research would be needed to understand the mechanism of this, and to further explore the viability of this model.