Mud Lake is part of a wildlife refuge located in southeastern Idaho and is operated by a power company to maximize power consumption, while providing for delivery of irrigation flows. Mud Lake is used as a sediment trap for the water flowing from the Bear River into the adjacent Bear Lake. This study explores the use of multivariable relevance vector machine (MVRVM) modeling to predict suspended fine sediment and other water quality constituent concentrations, and their spatial and temporal distribution. In this article, the first of two, we describe an experimental design and data collection program for the observations of water quality constituent data and hydraulic parameters to support creation of the MVRVM model. We briefly describe the MVRVM modeling approach, describe the development of the experimental program and the resulting observations, and finally discuss the suitability of the data to modeling with the MVRVM. Details of the MVRVM model and its application are provided in a second article. Success of the MVRVM will confirm the ability of statistical learning tools to predict sediment concentrations, and will lead the way for scientists to expand the use of the MVRVM for modeling of suspended fine sediment and water quality in other complex natural systems.