Simulation of nitrous oxide emissions from wheat-cropped soils using CERES

Estimation of nitrous oxide (N2O) emissions from arable soils, in relation to crop fertilization, is essential to devise strategies to mitigate the impact of agriculture on global warming. This paper presents the development and test of a N2O model resulting from the linkage of a dynamic soil-crop simulation model (CERES) with two sub-models of N2O production and reduction in soils. These sub-models (NOE and NGAS) account for both the nitrification and denitrification pathways. The resulting models (CERES NOE and CERES-NGAS) were tested against experimental data collected on three contrasting wheat-cropped soils representative of the Beauce agricultural region in France. Although the input variables for the N2O modules were correctly simulated, CERES-NGAS was over-responsive to soil water content in a Haplic Calcisol, and strongly over-estimated the N2O fluxes as a result. On the other hand, CERES-NOE predicted correct mean N2O emission levels for all sites, but failed to simulate the peak fluxes observed in the weeks following fertilizer application in the most N2O-productive soil. Both models achieved root mean squared errors in the 23-26 g N-N2O ha(-1) day(-1) range, significantly higher than the average experimental error on the measurements. On the other hand, their mean deviations were acceptable, being lower than 2.2 g N-N2O ha(-1) day(-1), compared with a mean observed flux of 7.9 g N-N2O ha(-1) day(-1). Overall, the response of CERES-NOE to soil type was more accurate, but this came at the cost of costly, site-specific characterization on the soils' biological properties. The development of pedo-transfer functions to infer these parameters from basic soil characteristics appears as a pre-requisite for the use of CERES-NOE on a wider scale.