The aim of this study is to derive operation rules for small reservoirs during dry events. The approach is illustrated by the case study of Ghézala Dam, located in the north of Tunisia. The following data were available: daily rainfall for the period from 1968 to 2005 and hydrological data from 1985 to 2002. As a first step, events of dry periods were analysed, according to a predetermined threshold. A dry event was considered to consist of a series of dry days separated by rainfall events, each defined as an uninterrupted series of rainfall days comprising at least one day with precipitation exceeding a threshold of 4 mm. A specific procedure was followed to generate synthetic sequences of dry and wet events, of lengths corresponding to a hydrological year. These data were routed through a simple rainfall—runoff model to obtain synthetic streamflow series. Each generated data set was subject to deterministic optimization using incremental dynamic programming (IDP). Operation rules were derived by means of multiple regression analysis, and appeared to perform most satisfactorily for different predetermined levels of available reservoir storage. The practical utility of these rules is demonstrated for simulated dry, intermediate and wet years, respectively.