Accounting for legacy nitrogen stores does not improve the accuracy of riverine nitrate load model predictions

Legacy stores of nitrogen (N) have been proposed as an explanation for time lags between implementation of nutrient management practices and discernable declines in riverine N loads, particularly in agricultural watersheds. Process models that treat the environment as being divided into linked homogeneous pools or compartments have been used recently to estimate legacy N stores and infer time lags. However, questions have been raised regarding equifinality, i.e. the idea that acceptable estimates might be obtained using models with different structures and/or sets of parameters, creating uncertainty in model interpretation. To further advance understanding of legacy stores and lags in watershed N delivery, we developed six fate and transport models that vary in simplicity and physicochemical detail. Using published annual N surplus and riverine N load datasets, we calibrated the models to minimize summed squared errors compared with multidecadal NO₃-N (nitrate N) yield time series at 254 riverine sites across the continental United States. All models fit most datasets well, with calibration results suggesting little if any lagged contribution from either biogeochemical or hydrologic legacy stores. The highest quality model in terms of adjusted Akaike information criterion was the simplest, comprising a single, well-mixed compartment that lacks any mechanism to represent time lags between N input and output. Calibrated model results thus fail to provide evidence of influential legacy N stores. However, results using incrementally lagged watershed N inputs (themselves shown to be temporally autocorrelated), also imply that delivery lags of up to about a decade cannot be ruled out.