Search

We estimated net anthropogenic phosphorus inputs (NAPI) to 18 Lake Michigan (LM) and 6 Lake Erie (LE) watersheds for 1974, 1978, 1982, 1987, and 1992. NAPI quantifies all anthropogenic inputs of P (fertilizer use, atmospheric deposition, and detergents) as well as trade of P in food and feed, which can be a net input or output. Fertilizer was the dominant input overall, varying by three orders of magnitude among the 24 watersheds, but detergent was the largest input in the most urbanized watershed. NAPI increased in relation to area of disturbed land (R = 0.90) and decreased with forested and wetland area (R = 0.90). Export of P by rivers varied with NAPI, especially for the 18 watersheds of LM (R = 0.93), whereas the relationship was more variable among the six LE watersheds (R = 0.59). On average, rivers of the LE watersheds exported about 10% of NAPI, whereas LM watersheds exported 5% of estimated NAPI. A comparison of our results with others as well as nitrogen (N) budgets suggests that fractional export of P may vary regionally, as has been reported for N, and the proportion of P inputs exported by rivers appears lower than comparable findings with N. [ABSTRACT FROM AUTHOR]

Catchment nitrogen (N) budgets are a valuable tool to assess relative magnitude of N inputs and predict losses via riverine export. However, a range of computational approaches may be chosen, potentially affecting the modeled relationship between inputs and exports. To determine the influence of various assumptions and computational details on the effectiveness of N input estimates in predicting riverine N export, we compared eight separate net anthropogenic N input budgets and one soils compartment budget for each of 18 Lake Michigan catchments. N input estimation methods that took into account seasonal fluctuations in livestock numbers and estimated crop N-fixation by legume yield rather than area harvested best predicted river N export. The average annual river export of N from the 18 catchments ranged from less than 300 kg N km−2 year−1 in forested areas to more than 800 kg-N km−2 year−1 in agricultural catchments and 1,580 kg-N km−2 year−1 in small urban catchments. Using the most effective model ( R 2 = 0.95, median prediction error = 1.8%) riverine N exports were found to account for 21% of N inputs. Other methods predicted riverine N exports less well ( R 2 = 0.61–0.73), bias was greater, and the fractional export of N inputs by rivers decreased to ~13%. The soil N budget also was a less effective predictor of river export. This comparison demonstrates that N budgeting that incorporates more detailed description of agricultural N sources can substantially improve prediction of riverine N exports from catchments with a wide range of landscape characteristics. [ABSTRACT FROM AUTHOR]