Measuring epistemic success of a biodiversity citizen science program: A citation study

might indicate that the differences between Vigie-Nature articles and ‘‘mainstream” ones in terms of impact become more relevant in the long-term. To test the significance of this effect, we extend the longitudinal analysis presented in the previous section (see the ‘‘Method” sec- tion for full details). As for the comparison of citation history, we compared two models in their ability to fit the data. The first one (model 1) does not consider the interaction terms between time T and origin O of the papers, and the second one introduces interaction terms. We then led a likelihood-ratio test in order to test the null hypothesis that the two models are equivalent. For the comparison between the BC and VN groups, the test gives a p-values of 0.026 (<0.05). We thus accept the hypothesis that taking into account the interactions between T and O improves the fit. This means that the VN and BC citation history profiles present dif- ferent tendencies (that is to say, curves in Fig 2 are not similar, in the sense that they are not ‘‘parallel”). The same test performed for the VN and VN journals groups gives a p-value<10−6, which confirms the previous results. By comparing this result to the evolution of the average relative VN citation rates (Fig 3), it seems reasonable to argue that these difference in general tendencies are due to a time-dependant growing ‘‘excess” of citation of Vigie-Nature articles. In other words, the relative impact of VN articles (compared to ‘‘mainstream” ones published in the same journals) might significantly grows with time (at least from 0 to 11 years).

Statistical analyses of the dispersion of the citation rates At this point, an issue can be raised about the statistical distribution of the overall ‘‘excess” of citations of Vigie-Nature articles. Does it only concern a few highly-cited articles, or is it a more equally distributed feature regarding Vigie-Nature publications? To get a more precise picture of the citation profile, we computed the value of the median of the set of Vigie-Nature articles citations at each time t post-publication (that is to say, the number X of citations such as half of the papers are more cited than X, t years after publication). It appears that from 0 to 12 years, the median value is slightly higher than the average citation rate of the articles in the ‘‘Biodiversity and Conservation” category from WoS (see Fig 4). This means that about half of the Vigie-Nature articles are annually more cited than the mainstream publications in the same area. We also computed the deciles partition of the total number of citations of Vigie- Nature articles. Table 2 presents the citation thresholds of the 10%, 20% and 50% most cited papers both for Vigie-Nature and the whole ‘‘Environment and Ecology” category (from Clari- vate Analytics) between 2009 and 2019. The 9th decile (the 10% threshold) is about 2.3 times higher in the Vigie-Nature set of articles, whereas the 5th decile (the 50% threshold) is about 3.5 times higher. Together, these results suggest that the ‘‘excess” of citations of Vigie-Nature productions is quite equally distributed among the articles. At the very least, the average cita-

tion history of Vigie-Nature papers is clearly not led by a small amount of highly cited ones. Interpretation and discussion We have computed the citation history of 123 papers published in peer-review journals (between 2007 and 2019) resulting from Vigie-Nature, a long-term and large-scale citizen sci- ence program in ecology and environment sciences. We found that Vigie-Nature articles are on average significantly more cited from 0 to 12 years after publication than the total of articles in the same domain. This average value refers not only to a few highly cited papers. In addi- tion, the difference in yearly citations rate strongly increases with time. We evaluated that between 0 and 4 years post-publication, Vigie-Nature publications are cited each year about 1.4 times more than the mainstream ones (BC and VN journals groups). After 11 years, they are on average cited 3.3 times more per year. Quite interestingly, this long-term impact strongly contrasts with what was obtained by [15] when studying the citation history of PLOS ONE A citation study of a biodiversity citizen science program PLOS ONE | https://doi.org/10.1371/journal.pone.0258350 October 11, 2021 8 / 13