Complementing near-real time satellite rainfall products with satellite soil moisture-derived rainfall through a Bayesian Inversion approach

Abstract

Abstract This work investigates the potential of using the Bayesian-based Model Conditional Processor (MCP) for complementing satellite precipitation products with a rainfall dataset derived from satellite soil moisture observations. MCP – which is a Bayesian Inversion approach – was originally developed for predictive uncertainty estimates of water level and discharge to support real-time flood forecasting. It is applied here for the first time to precipitation to provide its probability distribution conditional on multiple satellite precipitation estimates derived from TRMM Multi-Satellite Precipitation Analysis real-time product v.7.0 (3B42RT) and the soil moisture-based rainfall product SM2RAIN-CCI. In MCP, 3B42RT and SM2RAIN-CCI represent a priori information (predictors) about the “true” precipitation (predictand) and are used to provide its real-time a posteriori probabilistic estimate by means of the Bayes theorem. MCP is tested across Italy during a 6-year period (2010–2015) at daily/0.25\u202fdeg temporal/spatial scale. Results demonstrate that the proposed methodology provides rainfall estimates that are superior to both 3B42RT (as well as its successor IMERG-early run) and SM2RAIN-CCI in terms of both median bias, random errors and categorical scores. The study confirms that satellite soil moisture-derived rainfall can provide valuable information for improving state-of-the-art satellite precipitation products, thus making them more attractive for water resource management and large scale flood forecasting applications.