SUMMARY
During 2018, many cities in Ivory Coast (Tiébissou, Bouaké, Niakaramandougou, Odienné, etc.) experienced prolonged episodes of drought, thus mobilizing the efforts of the Ivorian government to find solutions emergency. Thus, in order to effectively combat the effects of climate change which increasingly affect the cities of our country, it is necessary to go through adaptation schemes. This is why this study aims to model the average monthly minimum annual flows (QMNA) in the Marahoué watershed (Bandama) in order to define a system of early low water alerts (alert threshold, tolerance threshold and crisis threshold). As a result, the monthly flow data collected over the period 1961-2017 at the Bouaflé hydrometric station were used. The methodology used is based on the one hand on the frequency analysis in order to identify the law which best adjusts the QMNA and on the other hand, the determination of low-water thresholds making it possible to give early warnings. A distribution laws preselection was made and allowed to retain three laws (Gamma law, Weibull law and Exponential law). The empirical frequencies were calculated from the Hazen relation. The parameters of the different laws were determined by the weighted moments method. The adjustment was appreciated from graphical representations and the ?2 test. Two criteria (Akaike criterion and Bayesian criterion) were chosen to decide on the choice of the best models. The main results obtained showed that the QMNA of Marahoué were better adjusted by the Gamma law. It is therefore recommended to use this statistical model for the determination of low water levels in the Marahoué watershed (Bandama).