Nutrient enrichment is a major issue to many inland and coastal waterbodies worldwide, including Chesapeake Bay. River water quality integrates the spatial and temporal changes of watersheds and forms the foundation for disentangling the effects of anthropogenic inputs. However, many water-quality studies are focused on limited portions of the watershed or a subset of potential drivers. We demonstrate with the Chesapeake Bay Nontidal Monitoring Network (84 stations) that advanced machine learning approaches – i.e., hierarchical clustering and random forest – can be combined to better understand the regional patterns and drivers of total nitrogen (TN) trends in large monitoring networks. Cluster analysis revealed the regional patterns of short-term TN trends (2007-2018) and categorized the stations to three distinct clusters, namely, V-shape (n = 25), monotonic decline (n = 35), and monotonic increase (n = 26). Random forest models were developed to predict the clusters using watershed characteristics and major N sources, which provided information on regional drivers of TN trends. We show encouraging evidence that improved nutrient management has resulted in declines in agricultural nonpoint sources, which in turn contributed to water quality improvement. Additionally, water-quality improvements are more likely in watersheds underlain by carbonate rocks, reflecting the relatively quick groundwater transport of this terrain. However, TN trends are degrading in forested watersheds, suggesting new sources of N in forests. Finally, TN trends were predicted for the entire Chesapeake Bay watershed at the scale of 979 river segments, providing fine-level information that can facilitate targeted watershed management, especially in unmonitored areas. More generally, this combined use of clustering and classification approaches can be applied to other monitoring networks to address similar questions.