Evaluating Random Error in Long-term, Multi-plot Flux Gradient Measurements of Nitrous Oxide Emissions

Evaluating Random Error in Long-term, Multi-plot Flux Gradient Measurements of Nitrous Oxide Emissions

The multi-plot flux gradient (FG) technique is well-suited for non-intrusive measurements of agricultural N₂O emissions for individually-treated field-scale plots across growing seasons at high temporal and spatial resolution.  The degree of random error associated with N₂O FG measurements is unknown; knowledge of these errors will increase confidence in the flux measurements and strengthen comparisons of total N₂O emissions between treatments. An error estimation routine was developed to determine the degree of random error (σ) associated with FG-measured fluxes (σ_F).  The combination of a moving-block bootstrapping technique and the filtering method of Salesky et al. (2012) estimated the σ values for each variable used in calculating individual 30-min FG-derived fluxes. This error analysis was applied to a year-long dataset where a four-plot FG system measured N₂O fluxes semi-continuously in a soybean field in Southwestern Ontario, Canada, with each plot having different treatments affecting N₂O emissions.  Random errors of the eddy diffusivity (K) were on average 5% of the measured values of K and were largely controlled by the magnitude and random error of the friction velocity.   The application of stability corrections to K had a negligible effect on σ_F. Errors of the concentration differences contributed the largest proportion to the σ_F values, but the magnitude of the concentration difference did not correlate with σ_F.  The individual 30-min σ_F values did not correlate with flux values, and represented on average 69% of the magnitude of the measured 30-min flux, with this proportion being biased by the large number of fluxes close to zero.  Cumulating the errors over the experiment reduced the degree of error associated with the cumulated total N₂O-N emissions with an average value of 31.5 g N₂O-N ha^-1, which represented on average ±5.5% of the total N₂O-N emissions. This study was the first to present a method for calculating σ_F using high frequency data.  Expanding this methodology to a dataset where CO₂ multi-plot FG-fluxes were measured simultaneously with N₂O allowed for a comparison of the random errors between two different scalars measured under the same turbulence conditions.  Comparisons will also be made between the random errors of the FG measurements and the random error of eddy covariance measurement from the same field site.