Long-term trends of wet inorganic nitrogen deposition in Rocky Mountain National Park: Influence of missing data imputation methods and associated uncertainty. Academic Article uri icon


  • Excess reactive nitrogen (Nr) deposition is occurring in Rocky Mountain National Park and impacting sensitive ecosystems. In 2006, the National Park Service, State of Colorado, and Environmental Protection Agency established the goal to reduce Nr deposition to below the ecosystem critical load by 2032. Progress is tracked using 5-year averages of annual wet inorganic nitrogen (IN) deposition measured at Loch Vale, Colorado, by the National Atmospheric Deposition Program (NADP). This remote high alpine site is challenging to operate, and large fractions of the annual precipitation, at times >40%, had invalid IN concentrations. Annual wet IN deposition is calculated using the NADP protocol, which replaces missing concentrations with the annual precipitation-weighted mean (PWM) concentration of valid samples. This protocol does not account for seasonal variations in IN concentrations and the inverse relationship between concentration and precipitation amounts. Invalid samples occurred more frequently in the winter and at high and low precipitation amounts, and the NADP protocol generally overestimated annual deposition rates, by as much as 20%. Here, a new method for imputing missing weekly IN concentrations that accounts for their seasonal and precipitation dependence is introduced. Using a bootstrapping analysis shows that the new method reduced the errors in the annual deposition rates by about 30% compared to the NADP protocol and the biases were near zero. The overall trend in the wet IN deposition rates was found to be flat from 1990 to 2017, but the nitrate contribution decreased about 33%, which was offset by a nearly equal increase in ammonium wet deposition. These trends are consistent with known changes in nitrate and ammonium precursor emissions. The long-term trends in the annual IN deposition rates were similar using both data imputation methods, but the 2013-2017 average was about 10% smaller using the new method.Published by Elsevier B.V.

publication date

  • October 2019