Using environmental variables and soil processes to forecast denitrification potential and nitrous oxide fluxes in coastal plain wetlands across different land uses

Abstract

We examined relationships between denitrification (DNF) and nitrous oxide (N2O) fluxes and potentially important chemical and physical predictors to build a predictive understanding of gaseous N losses from coastal plain wetlands. We collected soil, gas, and pore water samples from 48 sampling locations across a large (440 ha) restored wetland, an adjacent drained agricultural field, and nearby forested wetlands every two months over two years. In summer and fall 2007, we measured soil DNF potential (21.6–3560 mg N m−3 d−1) and N2O efflux (−4.36–8.81 mg N m−2 d−1), along with 17 predictor variables. We developed statistical models for the most comprehensive subset of the data set (fall 2007) and used another subset (summer 2007) for cross‐validation. Soil pH and total soil nitrogen were the best predictors of DNF potential (Radj2 = 0.68). A regression using carbon dioxide flux and soil temperature together with soil extractable NH4+ and DNF potential explained 85% of the variation in fall N2O fluxes. The model for DNF performed reasonably well when cross‐validated with summer data (R2 = 0.40), while the N2O model did not predict summer N2O fluxes (R2 < 0.1). Poor model performance was likely due to nonlinear responses to high temperatures and/or higher and more variable root respiration by plants during the growing season, leading to overprediction of N2O flux. Our results suggest that soil DNF potential may be modeled fairly effectively from a small number of soil parameters, that DNF potential is uncorrelated with N2O effluxes, and that successful estimation of wetland N2O effluxes will require finer‐scale models that incorporate seasonal dynamics.