Sparse monitoring data and error inherent in water quality models make the identification of waters not meeting regulatory standards uncertain. Additional monitoring can be implemented to reduce this uncertainty, but it is often expensive. These costs are currently a major concern, since developing total maximum daily loads, as mandated by the Clean Water Act, will require assessing tens of thousands of water bodies across the United States. This work uses the Bayesian maximum entropy (BME) method of modern geostatistics to integrate water quality monitoring data together with model predictions to provide improved estimates of water quality in a cost-effective manner. This information includes estimates of uncertainty and can be used to aid probabilistic-based decisions concerning the status of a water (i.e., impaired or not impaired) and the level of monitoring needed to characterize the water for regulatory purposes. This approach is applied to the Catawba River reservoir system in western North Carolina as a means of estimating seasonal chlorophyll a concentration. Mean concentration and confidence intervals for chlorophyll a are estimated for 66 reservoir segments over an 11-year period (726 values) based on 219 measured seasonal averages and 54 model predictions. Although the model predictions had a high degree of uncertainty, integration of modeling results via BME methods reduced the uncertainty associated with chlorophyll estimates compared with estimates made solely with information from monitoring efforts. Probabilistic predictions of future chlorophyll levels on one reservoir are used to illustrate the cost savings that can be achieved by less extensive and rigorous monitoring methods within the BME framework. While BME methods have been applied in several environmental contexts, employing these methods as a means of integrating monitoring and modeling results, as well as application of this approach to the assessment of surface water monitoring networks, represent unexplored areas of research.