Spatial dependency of Solar-induced Chlorophyll Fluorescence (SIF)-emitting objects in the footprint of a FLuorescence EXplorer (FLEX) pixel: a SIF-downscaling perspective

<p>The assessment of large-scale vegetation functioning is essential to improve cropland productivity and monitor natural ecosystem health. The development of remote sensing (RS) technologies over decades made such assessments possible from field- to global-scale. Nevertheless, commonly used reflectance-based RS methods are often not sensitive enough to timely inform preventive or corrective actions. Recent advances on the RS of solar-induced chlorophyll fluorescence (SIF) have opened opportunities for novel approaches of earlier stress detection since SIF was found to be closely linked to photosynthesis. The forthcoming FLuorescence EXplorer (FLEX) satellite mission of the European Space Agency (ESA, to be launched) will offer timely non-aggregated global-scale SIF data at 300 m spatial resolution. Such pixel size, even though unique and accurate enough to monitor processes at biome level, may not be suitable to assess field scale processes. Therefore, the development of methodologies to downscale satellite-SIF information is currently of utmost interest since allowing to increase the spatial resolution of origin observations. A first step to comprehend the characteristics that possible approaches must meet is to understand the magnitude of the spatial variability within a coarse pixel footprint across representative vegetation types. Our study consequently aims to understand the spatial variability within the footprint of a FLEX pixel. We particularly analyze the spatial dynamics of SIF via the near infrared reflectance of vegetation (NIRv) data derived from Sentinel 2, World View- and Geo Eye- (10.0 m, 0.30, 0.40 m pixel<sup>-1</sup>, respectively) that was suggested as proxy for SIF in absence of environmental stress. With Sentinel 2 based NIRv we focus on four ecosystems, including small and large scale agriculture, pampa and savannah, with World View- and Geo Eye based NIRv, we investigate rain and coniferous forests. The very high resolution of World View- and Geo Eye was required to compute the variograms of forests since they were affected by a nugget effect when using Sentinel-2 images. Investigated ecosystems represent the most abundant vegetation types that the FLEX mission will cover. We also assessed the relation between the spatial dependencies (approximated by the lag of calculated semi-variograms) and the average object size in all the ecosystems. We found largest spatial dependencies (400-600 m) in large-scale agriculture, pampa and savannah and contrasting lower (<10 m) in forests. Spatial dependencies of small-scale agricultural scenes were in a middle position with approximately 100 m. Moreover, the spatial dependencies were found to be significantly (p = 0.023) linked to the average object size of the ecosystems. This demonstrates the importance of flexible downscaling methods, e.g. in a fractals-based direction (Quiros et al., in press).</p>