Modeling ammonia volatilization following urea application to winter cereal fields in the United Kingdom by a revised...

Urea accounts for over half of the global ammonia volatilization from synthetic nitrogen fertilization. Process-oriented models, such as the widely applied DeNitrification DeComposition 9.5 version (DNDC95), are expected to predict ammonia volatilization from agricultural urea applications under various conditions. However, compared to the wind-tunnel ammonia observations of a nationwide, seven-site/ten-case network in the United Kingdom, the DNDC95 greatly underestimated the ammonia volatilization from winter cereal fields applied with urea at a constant dose. The model was revised in this study mainly through (i) newly parameterizing the effects of clay fraction, plant standings and canopy wetting on ammonia release from the soil solution, and (ii) re-calibrating the parameters of the urea hydrolysis function and the original parameterizations of wind, soil temperature and moisture effects on ammonia volatilization using the observations in one case of the network. Next, the performances of the revised model (DNDC95_NH₃) in simulating grain yield, full crop nitrogen uptake, cumulative ammonia volatilization (CAV) and daily ammonia fluxes were evaluated using the observations of these variables in the other nine network cases. The DNDC95_NH₃ performed considerably better than the DNDC95. Particularly, the slopes of the zero-intercept regression for the observations versus simulations decreased from 1.97 to 0.86 and 1.65 to 0.75 for the CAV and daily ammonia fluxes, respectively. In addition, the number of the cases with significant discrepancies between simulated and observed CAV was reduced from six of the nine network cases to only two. The simulations by the revised model for these two cases showed a significant underestimation and an extremely large overestimation, respectively. However, reasons especially for the significant overestimation still remain unclear. Future studies still need to diagnose and solve the other existing problems so as to further improve the model's performance.