Spatio-temporal Statistical Inference and Data Fusion and their Applications to Decadal Survey Missions (Invited)

The Decadal Survey recommends a suite of missions that will produce data that are both overlapping and complementary. These data will be heterogeneous: they will have different sampling characteristics, geometries, observing technologies, and accuracies. To exploit their synergy, we suggest a statistical formalism for modeling relationships between true quantities of interest and noisy observations of them or of related variables. Our model accounts for sources of error and observing differences that affect the input data, and produces statistically optimal estimates with formal uncertainties. These inferences can be made from a single data set, or from multiple data sets. When a single data source is used, the process is typically called spatial or spatio-temporal interpolation. When multiple data sources are used, we call the process spatial or spatio-temporal data fusion. In both cases the underlying statistical model is the same. This talk discusses that model, and how it may be applied to data from existing missions and future ones in the Decadal Survey era.