Tropospheric ozone causes widespread damage to vegetation, however, monitoring of ozone-induced damage is usually reliant on manual leaf inspection. Reflectance spectroscopy of vegetation can identify and detect unique spectral signatures of different abiotic and biotic stressors. In this study, we tested the use of hyperspectral leaf reflectance to detect ozone stress in alder, beech, birch, crab apple, and oak saplings exposed to five different ozone regimes (ranging from daily target maxima of 30ppb ozone to 110ppb). Hyperspectral reflectance varied significantly between ozone treatments, both in whole spectra analysis and when simplified to representative components. Ozone damage had a multivariate impact on leaf reflectance, underpinned by changes in pigment balance, water content, and structural composition. Vegetation indices derived from reflectance which characterised the visible green peak were able to differentiate between ozone treatments. Iterative normalised difference spectral indices across the hyperspectral wavelength range were correlated to visual damage scores to identify significant wavelengths for ozone damage detection. We propose a new Chronic Ozone Damage Index (COzDI), which characterises the reflectance peak in the shortwave infrared region and is highly correlated to ozone damage. These results pioneer hyperspectral reflectance as a high-throughput method of ozone damage detection in a range of common broadleaf species.

This experiment took place at the UK Centre for Ecology & Hydrology's air pollution facility at Abergwyngregyn, North Wales (53.2◦N, 4.0◦W). Ozone treatment was delivered in solardomes, dome-shaped glasshouses of 3 m diameter and 2.1 m height. From June to October 2022, alder, beech, birch, oak, and crab apple saplings were grown within the ozone treatment solardomes. Each month hyperspectral leaf reflectance measurements were taken.  Reflectance spectra of the adaxial leaf surface from 350-2500nm were collected using an HR-1024i with leaf clip, Spectra Vista Corp, USA including an active light source. 

The spectrometer was referenced using the incorporated leaf clip reflective standard every 5 minutes during measurements. Dark signal baseline correction was applied by the HR-1024i automatically, a dark spectrum was taken before each reflectance measurement. Once per month during ozone treatment, reflectance spectra were taken of three leaves from three replicates of each species per ozone treatment. The three leaves were chosen as one from each third of the vertical span of the sapling's leaf cover, including young, mature, sun, and shaded leaves.

Reflectance spectra were imported and processed as a spectral library in Python 3 using the SpecDAL package. The overlapping regions of the three component spectrometers in the HR-1024i were stitched. Reflectance measurements were interpolated to correspond to 1nm interval wavelengths. Absolute reflectance was derived from the relative reflectance by multiplying by the known reflectance of the reference panel. Noisy regions caused by water vapour absorption at 1350-1460nm and 1790-1970nm were smoothed using a Savitsky-Golay Filter. Spectra were grouped by attributes including species, ozone treatment, and date of measurement.