PT - JOURNAL ARTICLE AU - Christian Mulder AU - Giorgi Mancinelli TI - Decontextualizing Big Data for a Better Perception of Macroecology AID - 10.1101/059915 DP - 2016 Jan 01 TA - bioRxiv PG - 059915 4099 - http://biorxiv.org/content/early/2016/06/22/059915.short 4100 - http://biorxiv.org/content/early/2016/06/22/059915.full AB - Fish species are charismatic subjects widely used for ecological assessment and modelling. We investigated the influence of electrofishing in an attempt to illuminate the extent to which datasets might be merged together. Three American Midwestern regions in Ohio were chosen as study area and the changes in the size-biomass spectra of more than 2000 fish assemblages were analysed. These communities behaved differently according to the sampling method in conjunction to the morphology of the investigated streams and rivers, as shown by decreasing predatory species and lowering of allometric slopes. There are here several lines of evidence indicating that the chosen sampling method, as determined by different habitats, acts as a pitfall and strongly influences the allometry of fish spectra. In the ongoing data-rich era, our results highlight that macroecological investigations, often performed by machine-learning systems without considering the different procedures adopted to collect original data, can easily produce artefactual allometric scalings.