气候变化领域集成服务门户

Process-basedmodels are useful tools to integrate the effects of detailed agricultural practices, soil characteristics, mass balance, and climate change on soil N2O emissions from soil - plant ecosystems, whereas static, seasonal or annual models often exist to estimate cumulative N2O emissions under data-limited conditions. A studywas carried out to compare the capability of four models to estimate seasonal cumulative N2O fluxes from 419 field measurements representing 65 studies across China's croplands. The models were 1) the DAYCENT model, 2) the DNDC model, 3) the linear regression model (YLRM) of Yue et al. (2018), and 4) IPCC Tier 1 emission factors. The DAYCENT and DNDC models estimated crop yields with R-2 values of 0.60 and 0.66 respectively, but both models showed significant underestimation for all measurements. The estimated seasonal N2O emissions with R-2 of 0.31, 0.30, 0.21 and 0.17 for DAYCENT, DNDC, YLRM, and IPCC, respectively. Based on RMSE, modelling efficiency and bias analysis, YLRM performed well on N2O emission prediction under no fertilization though bias still existed, while IPCC performed well for cotton and rapeseed and DNDC for soybean. The DAYCENT model accurately predicted the emissions with no bias across other crop and fertilization types whereas the DNDC model underestimated seasonal N2O emissions by 0.42 kg N2O N ha(-1) for all observed values. Model evaluation indicated that the DAYCENT and DNDC models simulated temporal patterns of daily N2O emissions effectively, but both models had difficulty in simulating the timing of the N2O fluxes following some events such as fertilization and water regime. According to this evaluation, algorithms for crop production and N2O emission should be improved to increase the accuracy in the prediction of unfertilized fields both for DAYCENT and DNDC. The effects of crop types and management modes such as fertilizations should also be further refined for YLRM. (C) 2018 Elsevier B.V. All rights reserved.