Rainfall Variability and Droughts in the Sebou Basin (Morocco): ERA5 Bias Correction by Quantile Mapping and Trend Analysis over 1980–2024

Reconstructing long-term continuous precipitation series is essential for characterising rainfall variability, detecting hydroclimatic trends, and supporting water resource management in semi-arid Mediterranean regions. This study analyses rainfall variability in the Sebou watershed (northern Morocco, ~40,000 km2) over 1980–2024, using eleven stations spanning an altitudinal gradient from 5 m (Kénitra) to 1478 m (Aït Khebbach). To reconstruct continuous monthly series over 44 hydrological years, ERA5 reanalyses were bias-corrected using seasonal Quantile Mapping (QM) versus station data, available only until 2001. Raw ERA5 data show a generalised positive bias increasing with altitude (from +8.9% at Kénitra to +112.9% at Pont du Mdez), with severe NSE degradation in complex relief areas (NSE = −3.834 at Pont du Sker). After QM correction, the mean bias drops to −0.1% and mean NSE improves from −0.32 to +0.68. Cross-validation (100 random-splitting iterations) confirms correction robustness, yielding a mean NSE of 0.594 ± 0.076. The stationarity hypothesis at the 2001/2002 boundary is validated for all eleven stations through the Pettitt test applied directly to the ERA5–OBS bias series (p = 1.000). Mann–Kendall trend analysis reveals no significant trend for most stations, reflecting strong natural decadal variability of the semi-arid Mediterranean regime, which may limit trend detection over the study period. SPI-12 analysis identifies four major drought episodes (1981–1985, 1991–1995, 1999–2002, 2018–2022) and notable wet surpluses (1995–1996, 2008–2010). Aïn Timédrine (650 m) shows notable chronic drought vulnerability (9 years with SPI-12 < −1.0). This study provides the first continuous 44-year monthly precipitation series for the Sebou basin, quantifies the ERA5 elevation mismatch (Δz) as the dominant bias predictor (r = −0.828, p = 0.002), and delivers a multi-scale drought characterisation (SPI-3, SPI-6, SPI-12), constituting a valuable dataset for long-term hydrological modelling.