The exchange of water and solutes between reservoirs of the river corridor is essential to maintaining many ecosystem services. While many previous studies have investigated the primary controls of river corridor exchange, contradictory relationships between controls and exchange responses have been reported in the literature and are likely a result of small sample sizes and methodological inconsistencies across studies. In this study, we present a large-scale, systematic investigation of the controls of river corridor exchange across the naturally occurring gradient of geologic and hydrologic conditions of a common river network. We conducted conservative solute tracer tests across a diversity of geomorphic and hydrologic conditions and used a basic time series analysis to describe solute transit time distributions. We then used generalized linear models and multiple machine learning techniques to predict time series with respect to remotely sensed topographic and surface water concavity indices. To determine how performance may improve by increasing observations, we simulated a large dataset and measured model performance along a gradient of sample size. Results of linear modelling exposed weak relationships between predictor and response variables, although explanatory power varied considerably with respect to response variable predicted. While predictive capacity remained poor, it appeared that hydrologic indices contributed disproportionately to explanatory power of top models. In comparing results of various machine learning and linear modelling approaches, we found no significant difference in performance among techniques. The low predictive power observed in our modelling efforts indicates a high degree of variability in exchange flux and transient storage not well explained by our existing suite of topographic, surface water concavity, and basic hydrologic indices. Results of our simulation efforts indicate that expansion of the existing dataset may lead to improved model performance in future studies.