Performance limits of GPM IMERG for sub-daily precipitation extremes over the Comunitat Valenciana, eastern Spain: Representativeness, attenuation, and pixel-scale displacement

We evaluate GPM IMERG Final V07 (0.1°, half-hourly) for sub-daily precipitation extremes over the Comunitat Valenciana (eastern Spain) using the dense AVAMET network (556 gauges) for 2019–2025. We present an empirical benchmark that combines direct gauge–pixel comparison, gauge-derived cell-scale controls, fixed-threshold verification, and event-based diagnostics at 30, 60, 180, and 360 min. Across 15,648 station–scale extreme-event records pooled over four accumulation scales, three results stand out. First, IMERG frequently shows an event-time wet response: the pooled event-conditioned wet-response rate is 0.870 at the station-matched cell and 0.947 under fixed 3 × 3 diagnostic tolerance. This conditional sensitivity is not evidence that IMERG reproduces extreme intensity and is therefore complemented by scale-specific and station-level summaries. Second, IMERG strongly attenuates magnitude (pooled median IMERG/AVAMET ratio 0.081 and 0.166, respectively), and fixed-threshold skill remains low even at 6 h (ETS 0.0714 and 0.0927). Third, gauge-derived cell-scale controls remain well above IMERG after partial support harmonization, indicating substantial residual mismatch in this benchmark. Among 14,795 records with a positive, uniquely located 3 × 3 maximum, only 11.5% place that maximum in the station pixel (median grid-cell-centre displacement 11.12 km), consistent with combined attenuation and local displacement. The results define limitations for sub-daily rainfall estimation; they do not constitute a validation of hydrological warning performance.