Effects of point and nonpoint source controls on total phosphorus load trends across the Chesapeake Bay watershed, USA

Reduction of total phosphorus (TP) loads has long been a management focus of Chesapeake Bay restoration, but riverine monitoring stations have shown mixed temporal trends. To better understand the regional patterns and drivers of TP trends across the Bay watershed, we compiled and analyzed TP load data from 90 Non-Tidal Network stations using clustering and random forest (RF) approaches. These stations were categorized into two distinct clusters of short-term (2013-2020) TP load trends, i.e., monotonic increase (n = 35) and monotonic decline (n = 55). RF models were developed to identify likely regional drivers of TP trend clusters. Reductions in point sources and agricultural nonpoint sources (i.e., fertilizer) both contributed to water-quality improvement in our period of analysis, thereby demonstrating the effectiveness of nutrient management and the importance of continuing such efforts. In addition, declining TP trends have a larger chance to occur in carbonate areas but a smaller chance in Coastal Plain areas, with the latter likely reflecting the effect of legacy P. To provide spatially explicit information, TP trend clusters were predicted for the entire watershed at the scale of river segments, which are more directly relevant to watershed planning. Among the 975 river segments, 544 (56%) and 431 (44%) were classified as "monotonic increase" and "monotonic decrease", respectively. Furthermore, these predicted TP trend clusters were paired with our previously published total nitrogen (TN) trend clusters, showing that TP and TN both declined in 185 segments (19%) and neither declined in 337 segments (35%). Broadly speaking, large-scale nutrient reduction efforts are underway in many regions to curb eutrophication. Water-quality responses and drivers may differ among systems, but our work provides important new evidence on the effectiveness of management efforts toward controlling point and nonpoint sources.

Keywords: Water quality, Point source, Nonpoint source, Watershed management, Cluster analysis, Random forest