Assessment of models for estimation of land-derived nitrogen loads to shallow estuaries

The performance of several models used to estimate land-derived N loads to shallow receiving estuaries are compared. Models included in the comparison differed in complexity and approach, and predicted either loads or concentrations in estuary water. In all cases, model predictions were compared to measured loads or concentrations, as appropriate. Measured N loads to 9 estuaries on Cape Cod, MA, were obtained as the product of mean concentrations in groundwater about to seep into estuaries multiplied by the annual recharge of groundwater. Measured annual mean N concentrations in estuaries were obtained by extensive sampling surveys. The validity of this procedure to measure loads was verified by comparison against seepage meter data. Responsiveness of model predictions was generally good: predictions increased significantly as measured values increased in 8 of the 10 models evaluated. Precision of predictions was significant for all models. Three models provided highly accurate predictions; correction terms were calculated that could be applied to predictions from the other models to improve accuracy. Four of the models provided reasonable predictive ability. Simulations were run with somewhat different versions of two of the models; in both cases, the modified versions yielded improved predictions. The more complex models tended to be more responsive and precise, but not necessarily more accurate or predictive. Simpler models are attractive because they demand less information for use, but models with more comprehensive formulations, and emphasis on processes tended to perform better. Different models predicted widely different partitioning of land-derived N loads from wastewater, fertilizers, and atmospheric deposition. This is of concern, because mitigation options would be based on such partitioning of predictions. Choice of model to be used in management decisions or for research purposes therefore is not a trivial decision. # 2002 Elsevier Science Ltd. All rights reserved.