Evaluating Satellite and Reanalysis Rainfall Estimates for Climate Services in Agriculture: A Comprehensive Methodology

ABSTRACT High‐resolution rainfall estimates from satellite and reanalysis data sources (SRE) could play a major role in improving climate services for agriculture. This is particularly relevant in regions that predominantly rely on rain‐fed farming but lack a dense network of ground‐based measurements to provide localised historical climate information, as in most of the Global South. However, there is a need for a framework which practitioners can use to determine the suitability of these estimated data for specific agricultural applications. This paper presents a comprehensive methodology for evaluating the ability of SRE to provide historical rainfall information for agricultural applications, primarily through comparison with ground‐based measurements. The methodology comprises five main steps: data selection and pre‐processing, spatial and temporal consistency checks, quantitative SRE‐gauge comparisons, bias correction, and application specific summaries. The methodology makes use of graphical summaries, standard comparison metrics, and Markov chain models. We describe how users can apply this methodology to evaluate rainfall estimates for specific applications, complementing existing validation studies. Evaluation cases are presented to demonstrate the methodology using five widely used satellite and reanalysis rainfall products and ground‐based measurements from 12 stations in Africa and the Caribbean. The case studies demonstrate how the methodology can be applied to examine multiple aspects of the rainfall estimates. While previous validation studies ask “Does the SRE estimate the true rainfall well?”, this methodology provides means of establishing “To what extent can an SRE be used for this specific purpose?” and a comprehensive framework for achieving this. This meets a major need for location specific rainfall information to improve climate information services for millions of small‐holder farming households.