The potential of spatial aggregation to extract remotely sensed sun-induced fluorescence (SIF) of small-sized experimental plots for applications in crop phenotyping

Abstract

Airborne measurements of sun-induced chlorophyll fluorescence (SIF) are a promising tool for monitoring plant functioning on different scales. However, currently operational airborne imaging spectrometers for SIF measurements still have limited spatial resolution and pointing accuracy. This is challenging in terms of the practical use of SIF maps for crop breeding and plant phenotyping. We developed and tested two spatial aggregation approaches to make airborne SIF data usable in experimental settings with a high number of small experimental plots. The two aggregation approaches generating representative SIF values for experimental plots demonstrated the potential to be used in crop phenotyping. The first aggregation approach (Approach A) aggregates pixel values directly on SIF maps, whereas the second approach (Approach B) aggregates at-sensor radiance before SIF retrieval. The statistical analysis showed that Approaches A and B led to significantly different SIF products for single experimental plots (p < 0.001). To evaluate the usability of the two approaches, aggregated SIF products were fitted against ground-based reference measurements. We found that Approach B provided a better representation of ground truth SIF760 (R2 = 0.61, p < 0.001) than Approach A (R2 = 0.55, p < 0.001) when combined with weighted averaging and robust outlier detection. Furthermore, our results suggest that a slight decrease in the spatial resolution of the image data improves accuracy of aggregation.