Precipitation trend uncertainty from product choice rivals or exceeds the magnitude of actual trends

Abstract Gridded precipitation products are widely used to assess long-term rainfall and snowfall change, yet their trend reliability remains uncertain. We evaluate 19 (quasi-)global products for three annual statistics: mean precipitation, daily maximum, and wet-day frequency, screening for temporal homogeneity and benchmarking trends against daily and monthly rain-gauge observations. Most products exhibit spatially coherent discontinuities linked to observing-system changes, particularly satellite-instrument transitions in the late 1990s to early 2000s and 2014–2015. Notably, only 58–64% of non-gauge-corrected products match the reference trend sign on average. Even including gauge-corrected products, the interquartile range of product trends is 1.1–1.9 times the reference magnitude, so product choice rivals or exceeds the magnitude of the trends themselves. For 1983–2025 annual-mean trends over land, the eight long-record products span and extend beyond the climate model ensemble envelope, implying observations cannot cleanly disentangle model error from observational uncertainty. This uncertainty propagates into infrastructure design, agricultural risk management, and climate finance, particularly in the tropics and Global South.