Premium
Reliability analysis of fish traits reveals discrepancies among databases
Author(s) -
CanoBarbacil Carlos,
Radinger Johannes,
GarcíaBerthou Emili
Publication year - 2020
Publication title -
freshwater biology
Language(s) - English
Resource type - Journals
SCImago Journal Rank - 1.297
H-Index - 156
eISSN - 1365-2427
pISSN - 0046-5070
DOI - 10.1111/fwb.13469
Subject(s) - trait , categorical variable , reliability (semiconductor) , consistency (knowledge bases) , biodiversity , ecology , biology , affect (linguistics) , statistics , psychology , computer science , mathematics , power (physics) , physics , quantum mechanics , artificial intelligence , programming language , communication
Trait‐based approaches are commonly used in ecology to understand the relationship between biodiversity and ecosystem functioning, environmental filtering or biotic responses to anthropogenic perturbations. However, little is known about the reliability of assigned traits and the consistency of trait information among different databases currently in use. Using 99 native and alien Iberian inland fish species, we investigated a total of 27 biological and ecological traits for their consistency among 19 different databases and identified less reliable traits, that is, traits with high disagreement among databases. Specifically, we used generalised linear models and inter‐rater reliability statistics (Krippendorff's α ) to test for differences in trait values among databases. We also identified well‐studied versus data‐deficient traits and species. Our results show notable discrepancies and low reliability for several biological and ecological traits such as microhabitat preference, omnivory, invertivory, rheophily, and limnophily. Least reliable traits were mainly categorical (rather than continuous) and established by expert judgment and without a clear definition or a common methodology. Interestingly, categorical traits such as rheophily or limnophily, which showed significantly lower reliability, concurrently showed higher data availability and use than continuously scaled traits. Such uncertainties in trait assignments could affect bioassessment and other ecological analyses. Species with smaller distributional ranges and those that have been described more recently, presented lower coverage and data availability in trait databases. We encourage further standardisation of fish trait measurement protocols to help improve the robust application of bioassessment indices and trait‐based approaches.