COS 268-5 - Changes in spectral signature of leaves after desiccation: Implications for leaf traits and plant-soil interaction prediction.

Abstract: What drives changes in leaf reflectance and the impacts on the accuracy of spectra-trait predictability are of interest to ecologists due to the consequences on ecosystem function inference. One of these drivers is the process of leaf desiccation, a technique used to transport leaf samples out of the field. The desiccation process drastically changes the shape of spectral signatures and influences the behavior of the leaf radiative transfer models (RTMs) used to predict biochemical and physiological traits. The present work investigates the implications of leaf desiccation on leaf spectral reflectance within the VIS-SWIR (400 – 2500 nm) spectrum and the impact on the inference of leaf traits using the PROSPECT leaf-level RTM. In addition, we evaluate the capacity of leaf spectra in both leaf states to predict soil-leaf covariation between soil properties and emergent leaf biochemical and structural traits using leaf traits inferred from fresh and dry leaves. Thus, to address these questions, we collected three leaves from 55 woody plants growing at the Enid A. Haupt Conservatory at the New York Botanical Garden and measured the spectral reflectance of fresh leaves with a leaf clip attached to a spectroradiometer. Then, we dried the leaves following field techniques and measured the reflectance of the same leaves when dry. Also, we collected soil samples from the rhizosphere of eight trees and measured the pH, cation exchange capacity, and macro and micronutrients. The results showed that spectral reflectance and the first and second derivatives changed the most in the NIR wavelengths (i.e., 800-1100nm) and that these changes were not just due to water loss but also changes in the structural integrity of other molecular leaf components. In addition, we also found that changes in the detection uncertainties varied depending on the leaf trait. PROSPECT predicted the structural parameter N and total chlorophyll content per area with the same uncertainty values with fresh and dry leaves. The uncertainty values increased for carotene and water content per area. But, uncertainty values decreased for the leaf's dry mass and brown pigment content per area. For the analysis of soil-leaf properties covariation, the results showed that the linear models built with leaf traits predicted from dry leaves were better at predicting relationships with soil components such as soil chemical properties and macro and micronutrients compared to models built with leaf traits inferred from fresh leaves. In conclusion, dry leaf spectroscopy is cost-efficient for ecological research.