In a context of increasing climate variability, accurately assessing drought has become essential, particularly in semi-arid regions such as the Souss watershed. The scarcity of meteorological stations, short observation periods, and missing rainfall data complicate the reliable characterization of drought at the local scale. In this regard, satellite precipitation products (SPPs) offer a promising alternative, although their direct use is limited by the presence of biases, especially during extreme events. This study proposes an innovative bias correction approach based on multiple linear regression to generate a new, more accurate hybrid product. Unlike previous studies that used uncorrected data or basic methods, our approach integrates multiple SPPs, including ERA5 and CHIRPS, significantly improving data quality. The new product achieves an average R² of 0.72, compared to 0.67 for ERA5 and 0.57 for CHIRPS. The corrected product is used to compute the Standardized Precipitation Index (SPI) at 6- and 12-month scales to evaluate the effect of temporal scale on drought characterization. SPI-06 and SPI-12 results for the 1970–2018 period show strong agreement, despite some minor discrepancies. SPI-06 tends to emphasize the intensity of drought episodes compared to SPI-12. The analysis reveals that the standard SPI classification thresholds are not fully applicable to a climatic context such as ours, which led us to propose new, context-specific thresholds (corrected SPI). We also observed a noticeable lengthening of moderate drought episodes starting in 1998, including a three-year event from 1998 to 2000, as well as a marked intensification of severe droughts after 2008. Despite two exceptionally wet years in 2009 and 2010, the overall climate regime has become increasingly irregular. Finally, drought events appear to be occurring more frequently, with shorter intervals between successive episodes. Keywords: Satellite precipitation, CHIRPS, ERA5, corrected SPI, SPI-12, SPI-06, drought, multiple linear regression