Evaluating random error in long-term, multi-plot flux-gradient measurements of N2O emissions

English Abstract:

Management practices for crop production have an effect on N₂O emissions. N₂O fluxes from agricultural land vary spatially and temporally at small (<1 day) and large (>1 month) time scales as controlled by climatic conditions, seasonality and management. The flux-gradient (FG) technique is well-suited for the long-term, non-intrusive, and continuous measurement of trace gas fluxes to capture the total accumulation of N₂O emissions at the field scale across growing seasons. A multi-plot approach (up to four 4-ha plots within an aerodynamically homogeneous area) allows for side-by-side comparison of management practices under similar climatic and soil conditions. Comparison of annual N₂O emissions obtained from this multi-plot approach requires an evaluation of the degree of uncertainty in the total N₂O emission values to identify treatment effects. Random errors in flux measurements are considered to be one source of this uncertainty. An error estimation routine was developed to determine the degree of random error in FG-measured fluxes. The filtering method of Saleskey et al. (2012) estimated the random error of each variable used in calculating the FG-derived fluxes. This error analysis was used on a dataset where a four-plot FG system measured 30-minute N₂O fluxes at 2 hour intervals semi-continuously from June 2009 to June 2011. A preliminary analysis showed that random errors in concentration measurements contributed to approximately 92% of the total random error in each 30-minute flux value in unstable and neutral conditions, and 75% of the total error in stable conditions. The remaining proportion of the error was attributable to measurements of the friction velocity. Errors associated with variables used in calculating stability corrections were negligible. Cumulating the errors over the experiment showed that the total random error for each of the four plots was less than 1% of the total cumulated flux values. This suggests that random errors are of minor concern when assessing treatment effects from long-term FG measurements.