In order to ensure water resources for multiple uses, such as, urban agriculture, aquaculture, tourism and industry, and also flood protection, reservoirs building police was adopted in Brazilian semi-arid, and especially in Ceará state, region of high annual water deficit due to the irregularity in rainfall distribution in time and space,. That multi-purpose uses and the large variety of factors involved, make the management of reservoirs a complex task. The hydrological models emerge as a valuable tool to assist the decision maker in the process of water allocation, allowing for preliminary analysis based on simulation of different scenarios. This study presents a methodology for modeling water resources in semiarid northeastern - Brazil, targeting reservoir water balance. The model uses a spatial database and the Curve Number method for estimating runoff, and Muskingum-Cung for flow in natural channels. As input, the following dataset was used: spatial database containing the layout of the reservoir system, subdivided into drainage areas up to 25 km 2 (named interbasins) and between two confluences, soil type layer, spatial distribution of rain gauges and weather stations with their measured daily water precipitation and estimated evaporation; data from modeled reservoirs, such as, elevation-area-volume model, daily water level and flow released and spatial distribution of the reservoirs upstream of the main reservoir and its release, as well as water uptake from rivers between two reservoirs (water right) in the stream network. The automatic delineation of the river network was performed by Archydro tool in ArcGIS 9.3 platform, resulting in 227 interbasins for the catchment area of the Pedras Brancas reservoir, 1811 for Banabuiú reservoir, 2634 for Castanhão reservoir and 3263 for Orós reservoir. The daily runoff water depth in each interbasin, obtained by the Curve Number, was converted to flow and transferred by the Muskingum-Cunge method all the way to the reservoir as inflow. The evaporated volume was estimated from meteorological data obtained from Automatic Weather Stations, and the outflow volumes from reservoir releases. The precipitated volume was obtained by raingage data distributed near the reservoir. Interbasin runoff and the displacement of the hydrograph to the reservoir were computed using Microsoft Excel software, with programming language Visual Basic for Application. The model was applied in the Pedras Brancas and Banabuiú reservoirs, in Ceará state, with storage capacity of 434 hm 3 and 1,700 hm 3 , and watersheds areas of 1,787 km 2 and 13,500 km 2 , respectively. The model was tested by comparing the daily observed and estimated volumes. Although the model has shown weak performance during the rainy season, suggesting overestimation of runoff using the curve number method associated with the presented methodology, the performed well during the dry season, with good estimates of evaporation.
