Spatial Resolution Effects on Chlorophyll Fluorescence Retrieval in a Heterogeneous Canopy Using Hyperspectral Imagery and Radiative Transfer Simulation

Increasing attention is being given to chlorophyll fluorescence (F) for global monitoring of vegetation due to its relationship with physiology. New progress has been made in the methodological and technical aspects of signal retrieval with the recently published low-resolution global maps of fluorescence. Nevertheless, little progress has been made in the interpretation of the F signal when quantified in large pixels, an important issue due to the effects of structure, percentage cover, shadows, and background. High-resolution (40 cm) airborne hyperspectral imagery is used in this letter to assess the retrieval of fluorescence by the Fraunhofer line depth method from pure tree crowns and aggregated pixels. Due to canopy heterogeneity, the F signal extracted from aggregated pixels is highly degraded. A poor relationship is obtained between fluorescence extracted from pure tree crowns (Fcrown) and that quantified from pixels aggregating pure tree crowns, shadows, and background (Faggregated) (R-2 = 0.25; p < 0.01). The relationship between F and stomatal conductance (used as a physiological indicator) decreases as a function of aggregation, yielding R-2 = 0.69 (p < 0.01) when calculated from pure tree crowns and R-2 = 0.38 (p < 0.05) from pixels containing crown, shadows, and soil. This letter demonstrates the need for methods to accurately retrieve a pure-vegetation fluorescence signal from aggregated pixels. The FluorMODleaf and FluorSAIL models were combined with the geometric forest light interaction model (FLIM) model and led to the FluorFLIM model developed for this letter. Simulations conducted with FluorFLIM obtain predictive relationships between Fcrown and Faggregated pixels as a function of percentage cover, enabling the estimation of pure-crown F from aggregated pixels (R-2 = 0.72, p < 0.01).