Nonlinear Dynamics in Treatment Wetlands: Identifying Systematic Drivers of Nonequilibrium Outlet Concentrations in Everglades STAs

Abstract

AbstractConstructed treatment wetlands intercept nutrients in surface runoff to protect downstream ecosystems. Conventionally, nutrient removal is considered from a chemical kinetics perspective that assumes outlet concentrations tend toward equilibrium. We propose a broader view that includes management intervention as part of the system, and we explore the possibility that complex concentration dynamics emerge endogenously from nonlinear interactions. In particular, through application of a novel nonlinear time series diagnostic framework, we empirically test whether phosphorus removal at two Stormwater Treatment Areas (STAs) in South Florida might be enhanced through improved management of hydraulic and phosphorus loads. We isolated low‐dimensional, deterministic signals in STA input/output dynamics (flows and total phosphorus concentrations); mapped causal interactions among covariates; and developed parsimonious data‐driven dynamic models to simulate and characterize key interactions. At STA‐3/4, inflow dynamics systematically drove outlet concentrations, suggesting that outlet concentration variability might be reduced through enhanced hydraulic load management. We also detected management feedback in the reverse direction—STA‐3/4 outlet concentrations driving inflows—and identified a possible inefficiency in this management response. At STA‐2, neither flows nor inlet concentrations systematically drove outlet concentrations, suggesting that the wetland effectively removed incoming load signals. However, STA‐2 inflows were evidently influenced by inlet concentrations (management feedback). The differing management feedback structures between STAs suggest that STA‐3/4 performance might be improved by managing hydraulic loads more strongly in response to inlet concentrations. More broadly, this work demonstrates that management‐relevant insights into dynamic cause‐and‐effect relationships can be gleaned from observed treatment dynamics.