Abstract
Decision makers need to know the drivers of surface water phosphorus (P) concentrations, the environmental factors that mediate P loading in freshwater systems, and where pollution sources and mediating factors are co-located to inform water quality restoration efforts. To provide this information, publicly available spatial data sets of P pollution sources and relevant environmental variables, like temperature, precipitation, and agricultural soil erodibility, were matched with >7,000 stream and lake total P observations throughout the conterminous United States. Using three statistical approaches, consisting of (a) correlation, (b) regression, and (c) machine learning techniques, we identified likely drivers of P concentrations. Surface water concentrations in streams were more strongly correlated and effectively predicted by annual fertilizer and manure input rates and agricultural legacy sources compared to that of lakes. This observation suggests that streams may be more immediately responsive to improvements in agricultural nutrient management. In contrast, lake concentrations, though still positively associated with agricultural input and surplus variables, may be more influenced by historic erosional inputs, internal lake recycling, and other environmental factors. Thus, lake TP concentrations may not be as immediately responsive as streams to improvements in phosphorus management. Both stream and lake P concentrations will potentially increase because of warming temperatures and forest recovering from past acidification, putting even further pressure on existing water quality restoration efforts to meet nutrient loading reduction targets. The identified spatial data sets and relationships elucidated in this effort can inform the placement and development of watershed restoration strategies to reduce excess P in aquatic systems.