Watershed restoration in the Florida Everglades: Agricultural water management and long-term trends in nutrient outcomes in the Everglades Agricultural Area

Water quality degradation from agricultural runoff remains a pressing problem worldwide. A major challenge for restoring water quality is the need for long-term evaluation of governance and management interventions. In agricultural contexts, the primary interventions are best management practices designed to minimize nutrient losses by reducing fertilizer application, soil erosion, and drainage. Most studies are undertaken over short time scales or a few farms, which makes it difficult to connect management to water quality outcomes at larger scales. This paper addresses these gaps by examining 22 years of water quality trends at monthly time scales across the entirety of the 166 small, artificial drainage basins in the Everglades Agricultural Area, the sugarcane-growing region of Florida, USA. The Everglades Forever Act mandated the adoption of best management practices to reduce phosphorus loads but devolved implementation to farms collectively rather than requiring individual compliance. We examined the effect of biophysical and management drivers on long-term trends for two outcomes: a ratio of pumping-to-rainfall, which measures drainage decisions, and total phosphorus load per acre. We analyzed the magnitude and consistency of observed trends using Theil-Sen and Mann-Kendall analysis respectively across wet and dry seasons. Statistically significant downward trends were more common for decreases in magnitude than in consistency for both variables, indicating important management shifts have occurred but that some have not been continually improved over time. However, we also found statistically significant upward trends in a small number of basins for both variables. These results suggest that devolving management to farms has led to a widespread shift in management but that incentives for ongoing improvement would be valuable. Findings on biophysical and management drivers were limited, indicating that more fine-grain data may be needed to better detect their effects.