Comparison of N2O emissions from soils at three temperate agricultural sites: simulations of year-round measurements by four models

Nitrous oxide (N2O) flux simulations by four models were compared with year-round field measurements from five temperate agricultural sites in three countries. The field sites included an unfertilized, semi-arid rangeland with low N2O fluxes in eastern Colorado, USA; two fertilizer treatments (urea and nitrate) on a fertilized grass ley cut for silage in Scotland; and two fertilized, cultivated crop fields in Germany where N2O loss during the winter was quite high. The models used were daily trace gas versions of the CENTURY model, DNDC, ExpertN, and the NASA-Ames version of the CASA model. These models included similar components (soil physics, decomposition, plant growth, and nitrogen transformations), but in some cases used very different algorithms for these processes. All models generated similar results for the general cycling of nitrogen through the agro-ecosystems, but simulated nitrogen trace gas fluxes were quite different. In most cases the simulated N2O fluxes were within a factor of about 2 of the observed annual fluxes, but even when models produced similar N2O fluxes they often produced very different estimates of gaseous N loss as nitric oxide (NO), dinitrogen (N2), and ammonia (NH3). Accurate simulation of soil moisture appears to be a key requirement for reliable simulation of N2O emissions. All models simulated the general pattern of low background fluxes with high fluxes following fertilization at the Scottish sites, but they could not (or were not designed to) accurately capture the observed effects of different fertilizer types on N2O flux. None of the models were able to reliably generate large pulses of N2O during brief winter thaws that were observed at the two German sites. All models except DNDC simulated very low N2O fluxes for the dry site in Colorado. The US Trace Gas Network (TRAGNET) has provided a mechanism for this model and site intercomparison. Additional intercomparisons are needed with these and other models and additional data sets; these should include both tropical agro-ecosystems and new agricultural management techniques designed for sustainability.