A New Framework for Robust Retrieval and Fusion of Active/Passive Multi-Sensor Precipitation

Abstract:

This study introduces a new inversion approach for simultaneous retrieval and optimal fusion of multi-sensor passive/active precipitation spaceborne observations relevant to the Global Precipitation Measurement (GPM) constellation of satellites. This approach uses a modern Maximum a Posteriori (MAP) Bayesian estimator and variational principles to obtain a robust estimate of the rainfall profile from multiple sources of observationally- and physically-based a priori generated databases. The MAP estimator makes use of a constrained mixed and -norm regularization that warranties improved stability and reduced estimation error compared to the classic least-squares estimators, often used in the Bayesian rainfall retrieval techniques. We demonstrate the promise of our framework via detailed algorithmic implementation using the passive and active multi-sensor observations provided by the microwave imager (TMI) and precipitation radar (PR) aboard the Tropical Rainfall Measuring Mission (TRMM) satellite. To this end, we simultaneously obtain an observationally-driven retrieval of the entire precipitation profile using the coincidental TMI-PR observations and then optimally combine it with a first guess derived from physically-consistent a priori collected database of the TMI-2A12 operational product. We elucidate the performance of our algorithm for a wide range of storm environments with a specific focus on extreme and light precipitation events over land and coastal areas for hydrologic applications. The results are also validated versus the ground based observations and the standard TRMM products in seasonal and annual timescales.