| I estimated nitrogen (N) loading to 25 watersheds of the Lake Michigan Basin (LMB) from 1920 to 2002 to examine temporal and spatial variation in net anthropogenic N inputs (NANI) in relation to land use, climate, and agricultural practices and to explore how well NANI and climate are able to predict temporal and spatial variation in river export of total N. Based on my accounting of net anthropogenic N inputs (NANI) due to fertilizer application, crop fixation, net atmospheric deposition, and net trade of N as food and feed, total NANI to the entire LMB increased nearly three-fold over the 20th Century, from 136,500 Mg N/yr in 1920 to a peak of ∼375,200 Mg N/yr in 1987, and subsequently declined to 301,000 Mg N/yr in 2002. Watersheds with intense corn production in the eastern LMB experienced the largest (about six-fold) increase, from ∼770 kg-N/km2/yr to ∼4,570 kg-N/km2/yr, from 1920 to 2002.;To determine how well riverine export of total N (TN) can be predicted from N inputs to the land, I compared linear and log-linear regression models predicting riverine TN exports for 18 selected watersheds over five census years from 1974--1992, using a number of different N budgeting approaches. Various assumptions and computational details influenced model fit and prediction errors, and statistical relationships were improved, especially for small watersheds with diverse land use and farming practices, in response to specific model adjustments. NANI estimation procedures that account for seasonal fluctuations in livestock populations, and estimate crop N fixation using crop yield methods rather than area harvested, resulted in stronger models. A non-linear regression model that simultaneously incorporated spatial and temporal data for the preferred NANI model as well as annual runoff was able to account for 87% of the variation in riverine TN exports over time and space. This model, based on a more detailed description of N sources and losses and annual runoff, and incorporating both temporal and spatial variation, was found to have lower bias and higher precision in the prediction of riverine TN exports. |
