A Bayesian hierarchical framework for fusion of remote sensing data: An example with solar-induced fluorescence

Solar-induced chlorophyll fluorescence (SIF) has emerged as an effective indicator of vegetation productivity and plant health. The global quantification of SIF and its associated uncertainties yields many important capabilities, including improving carbon flux estimation, improving the identification of carbon sources and sinks, monitoring a variety of ecosystems, and evaluating carbon sequestration efforts. Long-term, regional-to-global scale monitoring is now feasible with the availability of SIF estimates from multiple Earth-observing satellites. These efforts can be aided by a rigorous accounting of the sources of uncertainty present in satellite SIF data products. In this paper, we introduce a Bayesian Hierarchical Model (BHM) for the estimation of SIF and associated uncertainties from Orbiting Carbon Observatory-2 (OCO-2) satellite observations at one-degree resolution with global coverage. The hierarchical structure of our modeling framework allows for convenient model specification, quantification of various sources of variation, and the incorporation of seasonal SIF information through Fourier terms in the regression model. The modeling framework leverages the predictable seasonality of SIF in most temperate land areas. The resulting data product complements existing atmospheric carbon dioxide estimates at the same spatio-temporal resolution.