RainCheck Africa: A Systematic Review and Meta-Analysis of Gridded Precipitation Dataset Performance at Continental Scale

Rendani Mbuvha , John Bagiliko, Callan Kotzen, Aya Salama, Abdoulaye Diop, Dioumacor FAYE, Amal Nammouchi, Sabrina Amrouche, Fifame Edwige Akpoly

Gridded rainfall products are indispensable for water resource management, agricultural monitoring, and disaster risk reduction across Africa, yet no systematic synthesis of their performance has been conducted. We present RainCheck Africa: a PRISMA-compliant systematic review and meta-analysis of 72 peer-reviewed studies published between 2015 and 2025 evaluating satellite, reanalysis, and merged products against ground observations across Africa, covering 1,749 metric observations across 148 locations and 154 product variants. Three findings emerge. First, correlation improves by 0.23 units from the daily to dekadal scale across all product families, confirming that gridded products are more reliable for climate monitoring and drought assessment than for daily event-based applications. Second, overestimation and underestimation co-occur in 90\% and 87\% of extraction entries respectively; this is not a contradiction but two expressions of the same intensity-dependent bias structure. Third, the daily Nash-Sutcliffe Efficiency of uncorrected products across 105 observations is centred near zero, meaning raw products applied directly to rainfall-runoff models perform no better than the long-term mean: bias correction is a prerequisite. Product performance is context-dependent. Reanalysis products are consistently outperformed and should not be used as primary rainfall inputs. CHIRPS v2.0 has the broadest validation base among global products; IMERG Final is the emerging standard for daily applications, though its corpus-level performance is concentrated in Ethiopian highland contexts and carries wide uncertainty elsewhere. Nine African Union member states have no validation study in the corpus, and the Congo Basin remains uncharacterised; this circularity can only be broken by expanded observation networks and open data-sharing.