Daily Nitrate and Ammonium Concentration Models for the Chesapeake Bay Watershed

The purpose of this study was to develop daily wet ammonium and nitrate concentration models for the Chesapeake Bay Watershed (CBW). Weekly measurements of ammonium and nitrate concentrations and precipitation volumes from 28 NADP/NTN sites located in or adjacent to the CBW were used for model development. Only weekly precipitation chemistry samples that were comprised of a single precipitation event were used along with the following variables: 1) the number of days since the preceding precipitation event; 2) the volume of precipitation occurring in the preceding 3, 5, 7, 10, and 14-day periods; 3) the number of days having precipitation during the preceding 7- and 14-day periods; 4) seasonality as represented by dividing each calendar year into six distinct bi-monthly periods starting with January and February; 5) local land cover (open water, forested, residential, transportation/industrial, crop land, and vegetated wetlands) as described by the 1992 National Land Cover Data (NLCD) 30-meter grids; and 6) local emission levels of ammonia and nitrous oxides obtained from the U.S. Environmental Protection Agency, National Emissions Trends (NET) database.

Precipitation volume was the strongest predictor of both ammonium and nitrate concentrations (p<0.0001). Both concentrations were inversely related to precipitation volume, although the dilution effect remained nonlinear after logarithmic transformation of both concentration and volume. The dilution effect also exhibited both seasonal and spatial variability. A latitudinal gradient was apparent in the concentrations for both species. Log-transformed concentrations of ammonium tended to increase linearly with latitude (p<0.0001). Nitrate concentrations also tended to be higher toward the north (p=0.0004), but the tendency was non-linear and confounded with longitudinal gradients in concentration and dilution rate. Significant long-term trends in concentration were observed for both nitrogen species. Ammonium concentrations tended to increase during the 1984 to 2001 (p<0.0001); whereas nitrate concentrations tended to decline during the same period (p<0.0001). Precipitation event history was also a significant factor affecting the concentrations of both ammonium and nitrate. Concentrations of both species were directly related to the number of days since the preceding precipitation event. This effect was more pronounced for nitrate (p<0.0001) than ammonium (p=0.0011). The volume of precipitation falling during the preceding seven days exhibited a significant, but moderate, inverse relationship to wet-fall ammonium (p=0.0120) and nitrate (p=0.042) concentrations, respectively.

Both land cover composition and emissions showed significant relationships to wet-fall concentrations of ammonium and nitrate. However, the elements of these two categories of predictors tended to displace each other in the stepwise regression selection process. As expected, ammonium concentrations were directly related to area-standardized emissions of ammonia (p<0.0001), and nitrate concentrations were directly related to emissions of nitrous oxides (p<0.0001). Ammonium concentrations were better predicted by emissions rates for the individual county containing the monitoring site than by the mean levels for the nearest three counties. Conversely, nitrate concentrations showed stronger relationships to mean emission rates for the nearest three counties than for the immediate county. The observed relationships of wet-fall concentrations to land use composition were more complex and less intuitive than for emissions rates. Concentrations of both ammonium and nitrate were strongly (p<0.0001), inversely related to the extent of forest cover within 8km (5miles) of a monitoring site; however, a weaker direct relationship exists with the amount for forest cover within 0.8 km (0.5 mile). Concentration levels of ammonium were also directly associated (p<0.0001) with the extent of industrial/transportation land uses within 8 km (5 miles) of a site. The stepwise predictor selection process identified the relative extent of open water in the surrounding 8- and 0.8-km (5- and 0.5 miles) proximity as a significant predictor of nitrate concentrations (p<0.0001, and p=0.0031, respectively). The functional relationship between nitrate concentration and the prevalence of surface water is not certain, but may reflect influences of coastal air masses on precipitation chemistry. No inflation of standard errors of regression coefficients was observed with the successive addition of land cover predictors to the models for either inorganic nitrogen species. Thus, there was no evidence of multi-colinearity of predictor variables. The decision to incorporate land cover composition rather than emissions levels into the final daily ammonium and nitrate concentration models was based on the slightly better model performance and on the consistent availability of land cover composition data for use in the model.

Estimates of mean event wet depositions from the models agreed well with observed depositions at six NADP/AIRMoN sites in operation within the CBW during 1992 through 2001 for both nitrogen species. However, measured individual event concentrations and depositions often varied widely from the estimated values. In spite of these large, single-event variations, the correlations between observed and estimated event depositions remained high, generally around 0.60 (range 0.54 to 0.72 for ammonium and 0.50 to 0.72 for nitrate). Applying the daily concentration models to the daily precipitation records from1984 through 2001 for the 28 NADP/NTN precipitation chemistry sites located in or adjacent to the CBW and summing the deposition estimates into annual totals provides a comparison with the observed annual deposition at those sites. The mean error for ammonium and nitrate depositions were -0.156 kg/ha and -0.85 kg/ha, respectively. The mean percent errors were 19.0 and 15.5, respectively.

⁽¹⁾ Telephone: (814) 865-8830, E-mail: jal@psu.edu
* Corresponding Author