Adaptive fusion of multisensor precipitation using Gaussian-scale mixtures in the wavelet domain

The past decades have witnessed a remarkable emergence of new sources of multiscale multisensor precipitation data, including global spaceborne active and passive sensors, regional ground-based weather surveillance radars, and local rain gauges. Optimal integration of these multisensor data promises a posteriori estimates of precipitation fluxes with increased accuracy and resolution to be used in hydrologic applications. In this context, a new framework is proposed for multiscale multisensor precipitation data fusion which capitalizes on two main observations: (1) non-Gaussian statistics of precipitation images, which are concisely parameterized in the wavelet domain via a class of Gaussian-scale mixtures, and (2) the conditionally Gaussian and weakly correlated local representation of remotely sensing precipitation data in the wavelet domain, which allows for exploiting the efficient linear estimation methodologies while capturing the non-Gaussian data structure of rainfall. The proposed methodology is demonstrated using a data set of coincidental observations of precipitation reflectivity images by the spaceborne precipitation radar aboard the Tropical Rainfall Measurement Mission satellite and ground-based Next Generation Weather Radar weather surveillance Doppler radars.