Spatially downscaling sun-induced chlorophyll fluorescence leads to an improved temporal correlation with gross primary productivity

Sun-induced chlorophyll fluorescence (SIF) is known to relate directly to leaf and canopy scale photosynthesis. Retrieving SIF from space can thus provide an indication on the temporal and spatial patterns of the terrestrial gross primary productivity (GPP). Recent studies have successfully demonstrated the serendipitous retrieval of SIF from satellite remote sensing instruments originally destined to atmospheric studies. However, the finest spatial resolution achieved by these products is 0.5°, which remains too coarse for many applications, including the early detection of drought impacts on vegetation and the integration with ground GPP measurements from flux-towers. This paper proposes a methodology to spatially disaggregate the information contained within each coarse SIF pixels by using a non-linear model based on the concept of light use efficiency (LUE). The strategy involves the aggregation of high-resolution (0.05°) remote sensing biophysical variables to calibrate the downscaling model locally and independently at each time step, which can then be applied to non-aggregated data to create a new layer, denoted SIF*, with a spatial resolution of 0.05°. A global SIF* dataset is generated by applying this methodology globally to 7 years of monthly GOME-2 SIF data. SIF* is shown to be a better proxy for GPP than the original coarse spatial resolution product according to flux-tower eddy covariance measurements. Its performance is comparable to dedicated GPP products despite that (unlike SIF*) these are calibrated based on the same flux towers, driven by meteorological data and not hampered by the large noise caused by the SIF retrieval. To further illustrate the added-value of the global SIF* product, this paper also presents: (1) an ecosystem level assessment showing a considerable reduction of noise with respect to the original SIF; (2) a spatio-temporal inter-comparison with existing GPP products; and (3) estimations of global terrestrial productivity per selected vegetation types based on SIF*.