Reservoirs affect the natural streamflow in river basins in various ways. The flow variability downstream is reduced due to storing of winter floods or extreme precipitation amounts and their release during low flow situations. Releases for irrigation in summer and the repression of melt water from mountainous snow packs may even reverse the natural seasonal pattern. Two conflicting objectives have to be considered when implementing rules for reservoir operation. For flood protection purposes the water level in the reservoir is aimed to remain below a so-called critical level which ensures a reasonable amount of empty space for storing future flood waters. Conservations purposes call for maximizing the amount of water for recreation, freshwater extraction or hydropower production. Reservoirs may be operated for one of these objectives or a combination of both. A set of release rules is usually prescribed for reservoirs to meet public demands. We found that release rules for reservoirs in the Odra watershed do not reflect the actual outflow. Such a situation is particularly awkward for numerical simulations as the outflow of reservoirs forms the boundary conditions for numerical models. For testing purposes or short-term operational forecasts measured outflow data may be used, but for water budget studies on a larger temporal scale or scenario simulations generally applicable release rules have to be known which describe the mean seasonal variation of the outflow and of the water volume in reservoirs. The reportedly prescribed rules for the control of reservoirs in the Odra watershed resulted in an almost constant reservoir water content over the year which is in contrast to the actual water content. The latter varied periodically with a maximum in spring and a minimum in autumn. Instead of applying the prescribed rules an empirical sinusoidal functional relationship for the reservoir water content is deduced from historical time series and implemented into the hydrological model SEROS. SEROS comprises a combination of the land surface scheme SEWAB and a horizontal routing scheme. The effect of the reservoir parameterization on the discharge downstream is analyzed by comparing three scenarios. Scenario I uses the observed outflow of all 11 reservoirs as input to the model. In scenario II the outflow of 11 reservoirs is parameterized according to equation 9. For scenario III we assumed that there is no reservoir at all. The inflow into the respective sub-catchment and the runoff from the land-surface model make up the outflow. Scenario III could be a realistic option if time scales are considered long enough not to be affected by any reservoir control or in case no information on the reservoir outflow is available. With the new reservoir control parameterization the goodness of streamflow simulations could be improved but depends on the location of the specific gauge within the whole watershed. The impact of the inclusion of the reservoir control module on the water balance of the Odra watershed is investigated.