Effect of variability in soil properties plus model complexity on predicting topsoil water content and nitrous oxide emissions

Despite the importance of soil physical properties on water infiltration and redistribution, little is known about the effect of variability in soil properties and its consequent effect on contaminant loss pathways. To investigate the effects of uncertainty and heterogeneity in measured soil physical parameters on the simulated movement of water and the prediction of nitrous oxide ([N.sub.2]O) emissions, we set up the Agricultural Production Systems sIMulator (APSIM) for different soil types in three different regions of New Zealand: the Te Kowhai silt loam and the Horotiu silt loam in the Waikato region, and the Templeton silt loam in the Canterbury region, and the Otokia silt loam and the Wingatui silt loam in the Otago region. For each of the soil types, various measured soil profile descriptions, as well as those from a national soils database (S-map) were used when available. In addition, three different soil water models in APSIM with different complexities (SWIM2, SWIM3, and SoilWat) were evaluated. Model outputs were compared with temporal soil water content measurements within the top 75 mm at the various experimental sites. Results show that the profile description, as well as the soil water model used affected the prediction accuracy of soil water content. The smallest difference between soil profile descriptions was found for the Templeton soil series, where the model efficiency (NSE) was positive for all soil profile descriptions, and the RMSE ranged from 0.055 to 0.069 [m.sup.3]/[m.sup.3]. The greatest difference was found for the Te Kowhai soil, where only one of the descriptions showed a positive NSE, and the other two profile descriptions overestimated measured topsoil water contents. Furthermore, it was shown that the soil profile description highly affects [N.sub.2]O emissions from urinary N deposited during animal grazing. However, the relative difference between the emissions was not always related to the accuracy of the measured soil water content, with soil organic carbon content also affecting emissions.Additional keywords: APSIM modelling, [N.sub.2]O emissions, soil water content, soil water module complexity.Received 23 March 2018, accepted 4 September 2018, published online 11 October 2018
IntroductionLand management models are increasingly being used to inform government policy that addresses challenges such as farming within environmental limits, and climate change mitigation and adaptation (Holzworth et al. 2017). [...]