Multi-Model Assessment of Mitigation Options for GHG Emissions in Croplands
Carozzi, M., Ehrhardt, F., Brilli, L., Bellocchi, G., Bathia, A., De Antoni Migliorati, M., Bregon, J.D., Dorich, C., Doror, L., Fitton, N., Grace, P., Grant, B., Giacomini, S.J., Leonard, J., Loubet, B., Massad, R.S., Mula, L., Pattey, E., Sharp, J., Smith, P., Smith, W., Zhang, Q., Recous, S. 2018. GRA affiliated presentation: Multi-Model Assessment of Mitigation Options for GHG Emissions in Croplands. 20th N Workshop and Side event, 25-27 June, 2018, Rennes, France
During a first step of the project (Ehrhardt et al., 2018), current biogeochemical models were assessed and calibrated through a blind test on five long-term field experiments with rotations of annual crops (including wheat, soybean, canola, maize, triticale, rice, oat) and assessing outputs as GHG emissions, C stocks and biomass production. Models calibration procedure consisted of five stages, gradually providing all the observed outputs to the modellers. The calibrated models were then used to evaluate the effect of mitigation options on GHG emissions and biomass production. A set of options with gradual intensities were tested with respect to site-specific N fertilisation regimes (from maximum to the minimum doses ordinarily used per crop), irrigation amounts (from +50 to -50% of the baseline values) and management of crop residues (exported or returned in the field), while keeping the parameter settings resulting from site-specific calibration. To do so, a simulation protocol provided to the modellers a set of maximum 60 scenarios per site, combining three strategies of management. The experimental data and simulated practices were applied to multiple years rotations from five sites (India, France, Australia, Canada and Brazil). A set of 13 process-based models contributed to the simulations: APSIM 7.6, Agro-C 1.0, CERES-EGC, DailyDayCent, DAYCENT (v4.5.2013, v4.5.2010, v4.5.2006), DNDC CAN, DSSAT GHG, EPIC 0810, FASSET v2.5, INFOCROP v2.1 and STICS v.831. Multi-model ensemble medians are used to interpret the individual responses and improve the accuracy of the assessment.