Uncertainty in flux - gradient nitrous oxide flux from agricultural fields

Abstract

Micrometeorological methods are ideally suited for continuous measurements of N2O fluxes but gaps in the time series occur due to low turbulence conditions, malfunction of measuring instrumentation, power failures, and adverse weather conditions. Currently, there is no standardized gap-filling technique for N2O emissions, while there are established methods for carbon dioxide fluxes. Two gap filling methods including linear interpolation and artificial neural networks (ANNs) were utilized to reconstruct missing N2O flux data over six years and evaluate the impact on annual N2O emissions. Two approaches of training the ANN model were applied including a multiyear ANN fit across the entire six year period and training each year separately. The performance of gap filling techniques was examined by generating 44 scenarios with artificial gaps (ranging in length from 1 to 14 continuous days, as well as combinations of different sized gaps).The single-year ANN method is recommended since this method captured flux variability better than the linear interpolation method (average R2 of 0.41 vs. 0.34). Annual N2O emission and annual bias resulting from linear and singleyear ANN were compatible with each other when there were few and short gaps (i.e. percentage of missing values were less than 30%). However, with longer gaps (>20 d) the bias error in annual fluxes varied between 0.082 and 0.344 kg N2O-N ha-1 for linear and 0.069 and 0.109 kg N2O-N ha-1 for single-year ANN. Hence, the single-year ANN with lower annual bias and stable annual sums over various years is recommended. The performance of ANN in gap filling was comparable for growing and non-growing seasons and better than over a whole year. Uncertainties in annual N2O flux was investigated, considering random errors for components of flux-gradient technique, error due to gap filling of N2O flux, and uncertainty coming from the perturbation of flux input variables. For random error, the main source of error compared to other variables for the estimation of N2O flux was the concentration difference. The uncertainties from gap filling was highest at 93% of total uncertainty while the uncertainty from perturbation of input variables was lowest (1.1%).