The Lower Sebou basin is unique in that it lies in the north of Morocco at the limit of the influence of the North Atlantic Oscillation (NAO) and under the oceanic influence, with continental effects towards the east. It also has the potential to play an important role in the socio-economic life of the region, which is dominated by agriculture, and to explain the impact of climate change in recent decades. The time series of climate data is dominated by precipitation in winter and spring and trends in December and March, determined by the Mann-Kendall test. The NAO and Mediterranean Oscillation (MO) have clearly shown their moderate to strong influence on precipitation by focusing on the NAO as a potential cause of precipitation variability, using the EOF technique. Based on climate data (1984-2016), mainly precipitation and temperature data, and taking into account the limitations and advantages of each drought index, four indices were developed, including the Standardized Precipitation Index (SPI) based solely on precipitation, the Standardized Precipitation Evapotranspiration Index (SPEI), the Drought Recognition Index (DRI) and the selfcalibrated Palmer Drought Severity Index (sc-PDSI) including potential evapotranspiration (PET), it has been confirmed that the SPI is the most suitable for detecting the duration and intensity of drought, particularly in sub-humid climate regions. However, the SPI underestimates drought changes by not taking the PET parameter into account. The four indices showed similar trends throughout the period studied, although the sc-PDSI tends to overestimate drought conditions due to its sensitivity to temperature and precipitation. Finally, using the SPI and SPEI indices to analyse seasonal trends showed significant trends, with an increase in temperature indicating increased warming in summer and a decrease in precipitation in winter. These trends are linked to the North Atlantic and Mediterranean oscillations, which are strongly influenced by cyclonic systems in the north-east Atlantic and which favour Mediterranean cyclogenesis. Using remote sensing and meteorological indices to assess agricultural drought and vegetation health on the one hand and to investigate the relationship between cereal yield and drought on the other, it was shown that crop health since the beginning of the 21st century has declined and that cereal yields have shown a strong response to inter-annual drought variability. Temperature-based drought indices were more correlated and sensitive to cereal yields than precipitation-based indices, suggesting that yields are more sensitive to changes in temperature than humidity. To predict these yields, province-wide empirical models using multi-source data, including indices based on remote sensing and meteorological data, and machine learning algorithms such as multiple linear regression, artificial neural network and random forest, showed that combining data from different sources led to better results than models based on a single source. Indeed, the models exploiting the RF and ANN algorithms were able to predict cereal yields as early as the winter months, with satisfactory statistical measures (0.7 < R² < 0.8 and 0.3 < RMSE < 0.5 t. ha-1 ).