The sustainable management of water resources is of high relevance with regard to overall socioeconomic development and environmental protection. Water quality monitoring plays a key role in this context as it provides the necessary information on the status of water resources and on the impact of human alterations of the hydrological cycle and hence forms an important basis for decision making. Current legislative approaches to water management like the European Water Framework Directive place a high importance to water quality monitoring. Monitoring systems have to provide relevant data in an efficient manner and are at the same time under budget constraints. This makes a case for optimization strategies for water quality monitoring networks, where the location of sampling stations and monitoring frequencies play a central role. Scientific methods to optimize water quality monitoring systems have been extensively described in literature. However, they are hardly ever applied since most of them depend on a-priory knowledge of the spatial and temporal variability of water quality parameters – information which is seldom available. The objective of this dissertation is to develop a method which allows estimating long term variability of water quality parameters. The parameter nitrate is chosen as an example parameter and the Aconcagua watershed in Chile is selected as a case study. The variability of nitrate concentrations over space and time is modelled on the basis of available hydrological, land use and point source data for the time period 1986 – 2006. For estimating nitrate exports to surface water the export coefficient method was used. The results are validated with measured nitrate concentrations of the same period. Results show that the model represents nitrate concentrations well for the upper and lower part of the watershed while low agreement between modelled and observed values was found for the lower part of the watershed, probably due to an insufficient representation of the hydrology of that zone but it could also be related to shortcomings of the current sampling methods at that particular monitoring station. Criteria for the location of monitoring stations and for the selection of monitoring frequencies were developed and applied together with the modelling results to develop recommendations for an optimized monitoring system. The main conclusions were on one hand that the current monitoring frequency of four samples per year is much too low recommending biweekly sampling instead; on the other hand one station could be omitted from the network as correlation between two stations was detected. The described method can serve as a general approach to support optimizing monitoring design if a minimum of data is available in order to estimate variance of a water quality parameter. This refers to daily information on discharges and to reliable estimates of point and diffuse pollution loadings to surface water. Thus, the method can be transferred to other watersheds and to other parameters.