A test of six simple indices to display the phenology of butterflies using a large multi-source database

Biological recording at broad temporal and spatial scales produces large volumes of species occurrence data. Multi-source datasets, which include opportunistic records, are unstructured and contain bias, mainly due to uneven and unknown observation effort, but they also provide meaningful information about species phenology. Butterflies are well known and well represented in citizen-science programs and national inventories, which makes them an interesting case for phenological studies. This work aims to find a simple, flexible, fast-rendering phenology index, which has to prove reliable when compared to standard knowledge. Six indices (two non-corrected and four corrected for observation effort) were built and implemented on butterfly records. They were analysed against blind expert opinion and a set of monitoring data. Surprisingly, all indices produced mostly realistic phenological patterns and non-corrected indices were as good as corrected ones. The number of species records divided by the number of records of all species of the group collected during the same period is the only index that should be avoided, because of an over-correction of recording intensity. Additional work is needed, in particular to refine the analysis by testing the sensitivity of the index to the amount of data, as well as by employing statistical models that are also useful for exploring trends and seasonal shifts.