Aquatic ecosystems are subjected to many chemical stressors, including nutrients and emerging contaminants like pharmaceuticals. While pharmaceutical concentrations in streams and rivers are often below the thresholds for acute toxicity, they nonetheless disrupt ecology through changes to organisms' physiology, metabolism, and behavior. However, analyzing samples for the wide range of manufactured pharmaceuticals is often prohibitively expensive for many monitoring efforts. As such, the ability to predict pharmaceutical concentrations over space and time using easier-to-monitor water quality parameters would expand our understanding of the scope and consequences of pharmaceutical contamination in aquatic ecosystems. We applied random forest models to data from the Baltimore Ecosystem Study to investigate how well routinely monitored water quality parameters could be used to predict concentrations of nutrients and pharmaceuticals. We found that concentrations of nutrients were accurately predicted by these models, but models for predicting concentrations of pharmaceuticals had high error rates and low predictive ability. Differences in our ability to predict concentrations of nutrients as opposed to pharmaceuticals could be due to differences in their sources, chemistries, or behavior in the environment. More concerted efforts to monitor pharmaceutical concentrations over time in aquatic ecosystems may help to resolve environmental drivers of their concentration and improve our ability to predict them.