Evaluating different methods for retrieving intraspecific leaf trait variation from hyperspectral leaf reflectance

Leaf mass per area (LMA), leaf dry matter content (LDMC) and water content/ equivalent thickness (EWT) are commonly used functional plant traits in ecology. Whereas spectroscopy has recently proven to be a powerful tool collect such trait information across large scales, it remains unclear whether these reflectance-based predictions accurate enough reliably model variation at the intraspecific level (i.e. individuals of one species). We explored potential hyperspectral methods predict LMA, LDMC EWT for two herbs (Hieracium umbellatum Jacobaea vulgaris) shrubs (Rosa rugosa Rubus caesius), based on 2400 samples. More specifically we tested i) inversion PROSPECT-D radiative transfer model, ii) generic PLSR approach using multibiome LMA iii) data-specific species level. For latter furthermore assessed both transferability trade-off between sample size accuracy. Although were relatively (R2 > 71 76%, respectively, however NRMSE = 33–47%), their performance was lower < 61%, 28–50%). slightly less successful retrieving 70%, 16–43%), unsuccessful through combining results 10%, 9–192%). The highest correlation accuracy obtained all three with species-specific models 10%). If high predictive is needed, thus suggest use models. training comes cost needed 100–160 leaves however, depending trait. seems limited overall, our potentially herbaceous species.