Spatial variation of canopy PRI with shadow fraction caused by leaf-level irradiation conditions

Photochemical reflectance index (PRI) is one of the best proxies to estimate the light use efficiency and photosynthetic activity of vegetation from remote sensing observations, especially if diurnal variations can be monitored. The calculation of PRI from leaf-level spectral reflectance measurements is unambiguous. Interpretation of the value of this index is more complicated, as it is affected by leaf structure and its carotenoid and chlorophyll content. Generally, a change in leaf-level PRI indicates a change in its photosynthetic capacity. At the scales of the canopy and beyond, various non-physiological factors modulate the leaf-level PRI signal inducing large angular and spatial variations in PRI. Specifically, previous studies have shown that within-canopy illumination variations and shadowing effects directly affect the PRI of a canopy. When observing a forest with a resolution finer than the size of a tree crown, large areas of shaded and sunlit foliage become visible. The spectral distribution of irradiance in these canopy regions is different from the average top-of-canopy irradiance used in the calculation of the canopy PRI. Thus, the leaf and canopy PRI can become decoupled. To date, no thorough analytical and empirical analysis of how the spectrally variable within-canopy light conditions cause apparent, non-physiological variation in canopy PRI has been published. In this study, we propose a new method to assess these PRI variations in structured vegetation from high spatial resolution (pixel size smaller than 1 m) imaging spectroscopy data. We used airborne imaging spectroscopy of boreal forest stands to evaluate the spectral irradiance in different locations inside the canopy and calculated a correction term for the canopy PRI estimates defined using top-of-canopy irradiance. We determined the maximum value of the correction term by sampling the most sunlit and shaded road surface locations adjacent to tree crowns. Results indicated that under the particular illumination-view geometry, irradiance variations decreased the canopy PRI by as much as 0.06 (relative change > 100%). The correction depended only slightly on atmospheric correction parameters. Finally, we reduced the illumination-related apparent variation in canopy PRI using the two-leaf canopy photosynthesis modeling scheme, canopy shadow fraction and the maximum correction term. In a test scene, the average illumination-corrected PRI was 0.027 smaller than non-corrected canopy PRI and showed no correlation with the shadow fraction, indicating a lack of down-regulation at the time of measurement. In theory, approach can be applied to all canopy level PRI measurements from towers, aircrafts and satellites under any observation geometry. However, further validation, preferably using in situ leaf reflectance data from different biomes, would be required before the algorithm can be routinely applied