An approach for statistical distribution selection: Application to the Saguenay-Lac-St-Jean hydrographic basin

The knowledge of the probability of occurrence of hydrologic events such as flood volume, flood peak, maximum rainfall, and water equivalent of snow i s essential to the optimum management of water ressources and planning of h ydropower production. In this paper, we present a practical approach for th e selection of an adequate statistical distribution to describe a hydromete orological variable. The selection of the distribution is based on statisti cal (mean absolute value of deviation between empirical and theoretical qua ntiles, Akaike criterion), qualitative considerations (parsimony, represent ation on normal probability plot), as well as hydrometeorological criteria (regional information, spatial coherence, temporal coherence). The approach is illustrated with the hydrometeorological data of the Saguenay - Lac St- Jean basins (Quebec, Canada). The proposed procedure reduces the subjectivi ty in the choice of an adequate statistical distribution but does not repre sent a completely objective methodology.