Improved Measures of Atmospheric Deposition Have a Negligible Effect on Multivariate Measures of Risk of Water-Quality Impairment: Response from Brown and Froemke
In their comment on our article on nonpoint-source pollution threats to water quality (Brown and Froemke 2012), Fenn and colleagues pointed out that we used a simplified measure of atmospheric deposition as one of our nine water-quality stressors. Our atmospheric stressor ignored dry nitrogen (N) deposition and the contribution of ammonium to total N. Furthermore, we used a simplified approach for combining nitrate and sulfate to capture acidic deposition.
As we stated in our article, we avoided the complications of computing more refined measures of our stressors because prior research had shown that the additional precision afforded by using more refined measures added little to the resultant multivariate measure of risk—principally because of the generally high correlation between a simplified measure and its related more refined measure. An additional reason for not including dry deposition was that the more comprehensive and spatially detailed estimates, from the Community Multiscale Air Quality model (CMAQ; www.cmaq-model.org), appeared to rely on a greater degree of modeling than we were comfortable with and, furthermore, were available for only one year: 2002 (see Baron et al. 2011).
Nevertheless, it is useful to determine whether including dry deposition and wet ammonium deposition in the measures of atmospheric N deposition and refining our measure of acidic inputs would significantly alter our estimates of the relative risk of water-quality impairment. In addition, it would be useful to see the effect of using PRISM (Parameter-elevation Regressions on Independent Slopes Model) precipitation data to distribute deposition estimates across the landscape, which would allow for greater spatial precision than what is available with the National Atmospheric Deposition Program (NADP) interpolation procedure that we relied on (Baron et al. 2011, Latysh and Wetherbee 2012).
In our article, we used subsets of nine stressors to estimate a relative risk value for each of three water-quality problems (sediment, nutrients, and toxics) for each of the 15,272 fifth-level watersheds covering the coterminous United States. The risk values for the three problems were then combined to yield a single risk value for each watershed. Wet nitrate () deposition was one of the stressors affecting the nutrient problem, and the sum of wet nitrate plus wet sulfate () deposition was one of the stressors affecting the toxics problem. The data for these measures were taken from the NADP data set for the years 2000–2006. New measures we explore here are, for the nutrient problem, nitrate plus ammonium () wet deposition (both incorporating PRISM precipitation data) plus dry deposition, all expressed in kilograms (kg) of N per hectare (ha) and, for the toxics problem, nitrate plus sulfate deposition expressed in equivalents per ha. The ammonium data were taken from the NADP Web site for the years 2000–2006, and the dry deposition data were taken from the CMAQ site mentioned above for 2002.
Adding dry N to wet N deposition has a substantial effect, increasing the median total N deposition across the watersheds from 3.5 kg per ha for wet only to 7.7 kg per ha for wet plus dry deposition. Across the full set of watersheds, the correlation of the new N measure with our original measure is r = .86. The correlation of acidic inputs in terms of equivalents with our original measure is r = .98. Finally, the correlation of the original scale values of risk with those computed when the new atmospheric deposition measures are incorporated is r = .996. The effect on the risk values of the change to the new deposition measures is very small, largely because in this framework, dry deposition affects only one of five stressors of one of three problems. Details on the new measures are available from the authors.
Therefore, we agree that our original measures of atmospheric deposition lacked completeness and precision, but we confirm that moving to more complete and refined deposition measures has a negligible impact on our multivariate, multiproblem characterization of the risk of water-quality impairment.
We thank Jill S. Baron and Ray Hulse for guidance on interpreting atmospheric deposition data.