Spatiotemporal downscaling under the volumepreserving constraint

The problem of prediction and simulation of point values from spatially and/or temporally aggregated block data is addressed within the change of support problem. We show that the geostatistical framework explicitly account for support differences between available data-spatiotemporal block data-and the target predictions at a higher spatial and/or temporal resolution, and yield coherent predictions (volumepreserving property). The spatiotemporal simulation allows performing uncertainty analysis associated with the downscaled values. In a case study we demonstrate that the downscaled predictions/simulated values reproduce the original spatiotemporal block data values, and the summary statistics of the original data, such as the mean and the spatio-temporal covariance structure, using both real world data and simulated spatiotemporal values.