Effects of Vegetation Directional Reflectance on Sun-induced Fluorescence Retrieval in the Oxygen Absorption Bands

Remote sensing of sun induced fluorescence (SIF) in absorption bands is based on the analysis of the changes in atmospheric absorption features between reflected and incident sunlight. Indeed, chlorophyll fluorescence increase the reflected light in the same amount insight and outside the absorption features, thus reducing the apparent depth of bands. However, incoming light is a mixture of direct and diffuse radiation. Diffuse light experienced a longer optical path, and have a greater absorption band depth than direct light. Furthermore incoming light is often estimated globally by measuring radiance from a reference board. As the vegetation reflectance response may be different for incoming direct and diffuse radiation, it may results in a variation of the absorption features estimated in the vegetation radiance that would impact fluorescence retrieval using conventional algorithms. These directional reflectance effects coupled to the angular distribution of incoming irradiance have been already mentioned in the literature as a possible source of artefacts in remote sensing of SIF using oxygen bands (Guanter et al. 2010). But until now, no experimental nor theoretical analyses of these effects on fluorescence retrieval have been reported. In this study, we report first experimental evidence of such a bias by comparing reflected and incoming sunlight under different illumination regimes over a senescent wheat crop containing no chlorophyll. Canopy structure was modulated by ears cutting and rows removing in order to evaluate the link between architecture, density and the measured bias. It was found that the bias dynamically evolved with daytime and illumination conditions. A maximum appears when diffuse and direct incident radiations are balanced. We propose an analytical scheme to account for directional reflectance effects in fluorescence retrieval algorithms (FLD, 3FLD, SFM) by considering different vegetation reflectance responses for direct and diffuse radiations. A numerical model based on SAIL and MODTRAN was developed to estimate these responses for different canopy architectures and illumination patterns. The simulated bias was in the same order of magnitude than the observed one, supporting the evidence of vegetation directional effects on fluorescence retrieval using commonly used algorithms. Application of this numerical model to green canopies for operational bias correction is discussed. Finally, a classification of experimental cases where directional effects can be neglected is proposed. 1. A SUSPECTED EFFECT 1.1. Historical intuition and operational observation Knowing the interest of steady state fluorescence variations relatively to illumination intensity as an index of stress status of a leaf (among others Cerovic, 1996, Flexas, 2000, Ounis, 2001) several authors had follow the steady state fluorescence over canopies during cloud shadowing using absorption band in-filling. When this approach was applied into Oxygen absorption bands some trends appears to be hardly attributable to pure physiological response. When simultaneous measurement where performs, the solar H bans shows none of the effects affecting O2 terrestrials bands. Those observations have trigged an intuition into the community of an expectable effect especially visible during sun/shade illumination transitions. This intuition was support by early numerical simulations (see FluoMODleaf report Miller et al., 1999). 1.2. Assessing the irradiance Operational retrieval of reflectance is based on hypothesis about the local irradiance of the target. Those hypotheses often include an assumption of isotropy of irradiance at ground level. 5th INTERNATIONAL WORKSHOP ON REMOTE SENSING OF VEGETATION FLUORESCENCE, 22-24 APRIL 2014, PARIS (FRANCE) 2 In the case of remote sensing of SIF, natural irradiance at ground level is one step more complex when we have to consider its anisotropy in the spectral vicinity of telluric absorption band. Depth is the ratio of irradiance intensity at maximum absorption band over the spectral continuum. The angular distribution of this ratio shows a pronounced angular and temporal variability. The driving illumination conditions are expected to be daytime, clouds cover and also atmospheric composition are key factors to understand such interception effects. As formulated by Plascyk in 1975, the reflectance is estimated in order to be ‘separate’ from fluorescence in the radiance. And the reflectance estimated under such anisotropic light should be biased, inducing a complementary bias on the fluorescence. We propose to set up a dedicated experiment for evidence of this specific bias on a natural canopy. 2. A dedicated experiment for a specific bias