Specimen preservation affects herbarium-derived leaf traits, reflectance spectra, and predictive models

Riley Fortier , Ryan S Fuller, Matthew Austin

• Herbaria are increasingly used as sources of functional trait data and reflectance spectra used in models for trait prediction and species classification. However, it is currently unknown how leaf preservation protocols may affect these data.
• Using a factorial experimental design, we tested how alcohol preservation, drying intensity, and their interaction affected foliar traits (leaf mass per area (LMA), effective leaf width (ELW), thickness, area, length), leaf reflectance spectra, and predictive models. We also tested the effect on accuracy when predictive models are trained on spectra from one preservation treatment and tested on spectra from another treatment.
• Alcohol preservation significantly affected all leaf traits, causing greater shrinkage with drying, whereas drying intensity had minimal effects. Both alcohol and drying had marked effects on reflectance spectra and on the accuracy of predictive models. Transferring models between preservation treatments significantly reduced prediction accuracy for both traits and species, although this effect was modest for traits.
• Our study highlights the importance of considering specimen preservation in herbarium-based studies. By demonstrating the effect of preservation on models that use reflectance spectra to predict traits and classify species, our study can be used to refine predictive models built from herbarium-based spectra.