Agricultural runoff carrying excess nitrogen (N) and phosphorus (P) represents a leading driver of freshwater and coastal eutrophication globally. Constructed wetlands (CWs) offer a nature-based, low-energy solution for intercepting nutrient-laden agricultural drainage before it reaches sensitive receiving waterbodies. This study evaluated the nutrient removal performance of subsurface flow (SSF) and surface flow (SF) constructed wetland systems treating agricultural tile drainage under contrasting hydraulic loading rates (HLRs: 2, 4, 8, and 16 cm d⁻¹) across three growing seasons at a temperate Midwestern United States agricultural watershed. Pilot-scale wetland cells (12 m × 4 m each) were planted with native macrophyte assemblages (Typha latifolia, Phragmites australis, Scirpus acutus, and Carex spp.) and monitored for inflow and outflow nutrient concentrations, hydraulic performance, plant biomass productivity, and associated biodiversity metrics. Results demonstrated that SSF-CW systems achieved superior total nitrogen removal (mean 68.4% ± 8.2%) compared to SF-CW systems (mean 52.1% ± 11.3%) across all HLRs, with optimal performance at HLR = 4 cm d⁻¹. Total phosphorus removal was highest in SSF-CW at low HLR (2 cm d⁻¹), achieving 79.3% removal efficiency, with declining performance at higher flow rates. Plant biomass harvesting in autumn significantly enhanced P removal by 18.6% compared to unharvested controls, implicating biomass P accumulation as an important removal pathway. Macroinvertebrate diversity (Shannon-Wiener index) increased significantly within all wetland types relative to drainage channels, supporting multiple ecosystem service provisioning. Life cycle assessment indicated that CW systems achieved net carbon sequestration of 1.2–2.8 t CO₂-eq ha⁻¹ yr⁻¹, partially offsetting agricultural GHG emissions. These findings provide design guidance for optimizing constructed wetland performance as cost-effective, multifunctional agricultural water management infrastructure.