Meeting Abstract

42.3  Sunday, Jan. 5 10:45  Crowdsourced morphometrics: a novel method to overcome bottlenecks in collecting phenotype data CHANG, J*; RABOSKY, DL; ALFARO, ME; Univ. of California, Los Angeles; Univ. of Michigan; Univ. of California, Los Angeles jchang641@gmail.com

The “bioinformatics revolution” has assembled massive genomic data sets; however, our ability to collect phenotypic data at a similar scale has not kept pace. Phenotypic data collected for a species or family are quite common, but comprehensive data sets across large radiations are rare. We present a method to distribute geometric morphometric landmarking tasks to untrained users over Amazon Mechanical Turk, an online crowdsourcing platform. We use this method to explore how shape diversity has evolved across the ray-finned fishes, the largest radiation of vertebrates. We asked remote workers to digitize landmarks of over 1000 lateral fish photographs, and evaluate the repeatability and accuracy of these crowdsourced data. We also construct a morphospace and examine the major axes of body shape variation, and compare these results to previous studies. Compared to the traditional approach of a single trained digitizer, our method gathers crowdsourced data at low cost and competitive accuracy, with significant speed increases for larger workloads. We will also discuss plans to increase the sampling density of this study, and its potential to reveal patterns of shape diversification and convergent evolution in the ray-finned fish radiation.