| Abstract: |
Anthropogenic nutrient enrichment of estuaries is a problem dramatically transforming coastal ecosystems worldwide. Recently, studies have suggested that many eutrophied estuaries, such as the Chesapeake Bay, the Gulf of Mexico, and in Danish coastal waters, have exhibited an unexpected response to nutrient reduction: hypoxic volume has continued to increase while nutrient loading has plateaued or decreased. One possible explanation for this phenomenon is that internal nutrient loading from the sediments is compensating for the decrease in external nutrient loading. In order to understand how the dynamics of sediment nutrient cycles respond to changing conditions, a standalone version of the Sediment Flux Model (SFM) used in the Chesapeake Environmental Modeling Package (CBEMP) was calibrated and compared to data in the Sediment and Oxygen Nutrient Exchange (SONE) database. Originally, SFM was calibrated and developed using data collected from 1985-1988 at eight stations. Since that time, the SONE database has grown to include almost 300 stations and in some cases almost twenty years of sediment flux measurements. In order to expand the model-data comparison to new stations and years a method of approximating depositional flux of organic matter was necessary. Depositional fluxes are very difficult to measure and only one site in the Chesapeake Bay has an appreciable time series of depositional flux data from sediment traps. In this post-audit, we calculated organic matter depositional flux from surface chlorophyll, carbon:chlorophyll ratio, and seasonally varying settling velocity. This method has the dual advantages of using more data for input and outputting useful information such as what fraction of the organic matter is incorporated in the sediment. Using data from the Chesapeake Information Management System (CIMS) for overlying water column nutrient and oxygen concentrations, we demonstrate that the SFM was capable of simulating fluxes of ammonia, dissolved inorganic phosphorus, and oxygen during years and locations for which it was never calibrated, while nitrate fluxes were generally over predicted and silica fluxes largely under predicted. For example, the correlation coefficient for a mid-bay station (R-64) between modeled and observed ammonia and phosphorus flux was 0.7 and 0.8, respectively over a ten year period (1986-1996). The model also performed well over a 20 year time series at multiple stations in the Patuxent River estuary, where substantial changes in carbon deposition and overlying water nutrient concentrations resulted from nutrient management actions and other external changes. Model fitting on an intra-annual scale is sensitive to the choice of organic matter settling velocity as this represents the time lag between algal blooms and subsequent deposition. The identification of years, stations, and processes that were poorly simulated offers the opportunity to improve model kinetics and/or add model processes, but overall model performance more than adequately captured intra- and interannual trends as well as demonstrating that this mass balanced model is capable of accounting for sediment nutrient memory. |
