Problems using ordinal traits with continuous measures of functional diversity

Continuous indices of functional diversity are popular in studies examining community structure and ecosystem function across a wide range of subfields from paleontology to range management. These indices were designed to replace the use of more arbitrary, discrete functional groups or guilds; however, the effect of typical methodological decisions on these new continuous measures has not been fully investigated. To test the effect of using ordinal traits in functional diversity analysis, I first calculated relative functional diversity index values in real plant communities with real continuous trait data and Euclidean distances. I then compared these original values to "treatment" functional diversity index values obtained by discretizing the trait data and using Gowers distance. Agreement between original and treatment values was highly unpredictable and often abysmal. Small methodological choices, such as whether to treat a functional trait as continuous (mm) or ordinal ("small", "medium", "large"), could completely change a perceived functional diversity relationship along an environmental gradient. Some parameter combinations returned results that were no better than random noise. Because simple methodological choices can have such a large impact on continuous functional diversity indices, it is ambiguous whether analyses using ordinal traits are actually measuring an underlying functional diversity relationship between communities or just reflecting the arbitrary parameter choices of researchers.