Modelling variability in N2O emissions from fertilized agricultural fields

Estimates of long-term landscape-scale N₂O emissions for greenhouse gas inventories are complicated by large temporal and spatial variability. Much of this variability is likely caused by topographic effects on surface and subsurface water flows. We hypothesized that this variability could be explained as degassing events during anaerobic soil conditions and during transitions from anaerobic to aerobic soil conditions as controlled by precipitation and subsequent water redistribution in complex landscapes. We simulated degassing events in the ecosystem model ecosys run in three-dimensional mode to simulate a fertilized agricultural field with topographic variation derived from a digital terrain map. N₂O emissions modelled from two areas within the field that had received 15.5 and 9.9 g N m⁻² as urea in May 1998 were compared with those measured by micrometeorological flux towers during June and July 1998. Modelled N₂O emissions during 1998 accounted for 2.3 and 2.0% of urea N applied at 15.5 and 9.9 g N m⁻², respectively. Degassing events in the model coincided with a key N₂O emission event measured in the field during several days after a rainfall in mid-June. During this event, modelled and measured surface fluxes rose rapidly to exceed 1 mg N m⁻² h⁻¹ for 2-3 d before declining. Emissions modelled concurrently at different topographic positions within the landscape during the emission event had coefficients of variation that varied over time between 30 and 180%. Much of the spatial variability in modelled emissions was attributed to temporal differences in the progression of emission events at different landscape positions caused by lateral water movement. The magnitude of temporal and spatial variability in N₂O emissions suggests that aggregation of flux measurements to regional scales should be based upon sub-daily measurements at representative landscape positions, rather than upon less frequent measurements at individual sites as currently done. The use of three-dimensional ecosystem models with input from digital terrain maps may provide a means for such aggregation to be conducted.