Identifying leaf traits that signal stress in TIR spectra

Plants under constant water and temperature stress experience a chain of reactions that in the long term alter their leaf traits (morphology, anatomy and chemistry). The use of these traits as proxies for assessing plant stress was so far mainly based on conventional laboratory methods, which are expensive and time-consuming. Remote sensing methods based on spectral changes can detect changes in pigments and productivity using the visible and near infrared. However, the use of infrared spectra, where changes in the spectra are associated with physical changes of the leaf, is still incipient.

In this study plants of Rhododendron cf. catawbiense, were exposed to low temperatures and low soil water content during a six months experiment. The spectral response in the infrared region 1.4–16 μm, microstructural variables, leaf water content, leaf area and leaf molecules such as lignin and cellulose concentrations were measured in individual leaves after the period of stress. This study revealed that under cold conditions plants have most changes in leaf water content, lignin and cellulose concentrations and leaf area, while under drought conditions the most striking change is water loss. These leaf trait modifications are also correlated with changes in thermal infrared spectra, showing their potential as proxies for detecting plant stress in this species. A multinomial model allows the estimation of the stress treatments imposed on these plants from their infrared spectra. This model reveals a group of 15 bands in the SWIR and MWIR between 2.23 and 7.77 μm, which show relatively large changes, and had an overall accuracy of 87%.

Finally, individual partial least squares regression models show that lignin, cellulose, leaf water content and leaf area are the leaf traits reacting significantly to long-term stress and that are also generating measurable changes in the infrared spectra. Although these models are based on laboratory data, the congruence of the identified bands with the fundamental molecular vibrations used in remote sensing, shows the potential of these findings in the assessment of plant stress.