An intercomparison of four gridded precipitation products over Europe using an extension of the three-cornered-hat method

Abstract. Precipitation is arguably one of the most relevant surface variables impacting human lives on the planet, but global-coverage, high-resolution and good-quality observations are not readily available. In particular, gridded observational datasets are much needed for model development and forecast quality assessment. Here, we compare the quality of four types of gridded precipitation products over Europe, namely a rain gauge interpolation, a satellite-derived product, a radar composite and a reanalysis. Each product has its own strengths and weaknesses, and since each precipitation estimate uses different measuring techniques, we can employ a triangulation method to estimate the error variance of each product with respect to the unknown true values. Results show that (a) the satellite product has limited quality over Europe and may be problematic to use in quantitative forecast evaluation and diagnostics; (b) the radar composite has spurious features that need to be considered when used in verification; (c) all products struggle in topographically complex areas; (d) the rain gauge interpolation is not free of errors, despite rain gauges often being treated as ground truth in the literature; and (e) the reanalysis dataset produces, in some cases, the best available estimates, particularly over the European near-coastal waters.