Simulation models are extensively used to predict agricultural productivity and greenhouse gas emissions. However, the uncertainties of (reduced) model ensemble simulations have not been assessed systematically for variables affecting food security and climate change mitigation, within multi-species agricultural contexts. We report an international model comparison and benchmarking exercise, showing the potential of multi-model ensembles to predict productivity and nitrous oxide (N O) emissions for wheat, maize, rice and temperate grasslands. Using a multi-stage modelling protocol, from blind simulations (stage 1) to partial (stages 2-4) and full calibration (stage 5), 24 process-based biogeochemical models were assessed individually or as an ensemble against long-term experimental data from four temperate grassland and five arable crop rotation sites spanning four continents. Comparisons were performed by reference to the experimental uncertainties of observed yields and N O emissions. Results showed that across sites and crop/grassland types, 23%-40% of the uncalibrated individual models were within two standard deviations (SD) of observed yields, while 42 (rice) to 96% (grasslands) of the models were within 1 SD of observed N O emissions. At stage 1, ensembles formed by the three lowest prediction model errors predicted both yields and N O emissions within experimental uncertainties for 44% and 33% of the crop and grassland growth cycles, respectively. Partial model calibration (stages 2-4) markedly reduced prediction errors of the full model ensemble E-median for crop grain yields (from 36% at stage 1 down to 4% on average) and grassland productivity (from 44% to 27%) and to a lesser and more variable extent for N O emissions. Yield-scaled N O emissions (N O emissions divided by crop yields) were ranked accurately by three-model ensembles across crop species and field sites. The potential of using process-based model ensembles to predict jointly productivity and N O emissions at field scale is discussed.

No abstract

No abstract

Simulation models represent soil organic carbon (SOC) dynamics in global carbon (C) cycle scenarios to support climate‐change studies. It is imperative to increase confidence in long‐term predictions of SOC dynamics by reducing the uncertainty in model estimates. We evaluated SOC simulated from an ensemble of 26 process‐based C models by comparing simulations to experimental data from seven long‐term bare‐fallow (vegetation‐free) plots at six sites: Denmark (two sites), France, Russia, Sweden and the United Kingdom. The decay of SOC in these plots has been monitored for decades since the last inputs of plant material, providing the opportunity to test decomposition without the continuous input of new organic material. The models were run independently over multi‐year simulation periods (from 28 to 80 years) in a blind test with no calibration (Bln) and with the following three calibration scenarios, each providing different levels of information and/or allowing different levels of model fitting: (a) calibrating decomposition parameters separately at each experimental site (Spe); (b) using a generic, knowledge‐based, parameterization applicable in the Central European region (Gen); and (c) using a combination of both (a) and (b) strategies (Mix). We addressed uncertainties from different modelling approaches with or without spin‐up initialization of SOC. Changes in the multi‐model median (MMM) of SOC were used as descriptors of the ensemble performance. On average across sites, Gen proved adequate in describing changes in SOC, with MMM equal to average SOC (and standard deviation) of 39.2 (±15.5) Mg C/ha compared to the observed mean of 36.0 (±19.7) Mg C/ha (last observed year), indicating sufficiently reliable SOC estimates. Moving to Mix (37.5 ± 16.7 Mg C/ha) and Spe (36.8 ± 19.8 Mg C/ha) provided only marginal gains in accuracy, but modellers would need to apply more knowledge and a greater calibration effort than in Gen, thereby limiting the wider applicability of models.

The objective of this study was to measure enteric CH4 emissions using a new portable automated open-circuit gas quantification system (GQS) and the sulfur hexafluoride tracer technique (SF6) in midlactation Holstein cows housed in a tiestall barn. Sixteen cows averaging 176 ± 34 d in milk, 40.7 ± 6.1 kg of milk yield, and 685 ± 49 kg of body weight were randomly assigned to 1 out of 2 treatments according to a crossover design. Treatments were (1) ad libitum (adjusted daily to yield 10% orts) and (2) restricted feed intake [set to restrict feed by 10% of baseline dry matter intake (DMI)]. Each experimental period lasted 22d, with 14 d for treatment adaptation and 8d for data and sample collection. A common diet was fed to the cows as a total mixed ration and contained 40.4% corn silage, 11.2% grass-legume haylage, and 48.4% concentrate on a dry matter basis. Spot 5-min measurements using the GQS were taken twice daily with a 12-h interval between sampling and sampling times advanced 2h daily to account for diurnal variation in CH4 emissions. Canisters for the SF6 method were sampled twice daily before milking with 4 local background gas canisters inside the barn analyzed for background gas concentrations. Enteric CH4 emissions were not affected by treatments and averaged 472 and 458 g/d (standard error of the mean = 18 g/d) for ad libitum and restricted intake treatments, respectively (data not shown). The GQS appears to be a reliable method because of the relatively low coefficients of variation (ranging from 14.1 to 22.4%) for CH4 emissions and a moderate relationship (coefficient of determination = 0.42) between CH4 emissions and DMI. The SF6 resulted in large coefficients of variation (ranging from 16.0 to 111%) for CH4 emissions and a poor relationship (coefficient of determination = 0.17) between CH4 emissions and DMI, likely because of limited barn ventilation and high background gas concentration. Research with improved barn ventilation systems or outdoors is warranted to further assess the GQS and SF6 methodologies.

Simulation models quantify the impacts on carbon (C) and nitrogen (N) cycling in grassland systems caused by changes in management practices. To support agricultural policies, it is however important to contrast the responses of alternative models, which can differ greatly in their treatment of key processes and in their response to management. We applied eight biogeochemical models at five grassland sites (in France, New Zealand, Switzerland, United Kingdom and United States) to compare the sensitivity of modelled C and N fluxes to changes in the density of grazing animals (from 100% to 50% of the original livestock densities), also in combination with decreasing N fertilization levels (reduced to zero from the initial levels). Simulated multi-model median values indicated that input reduction would lead to an increase in the C sink strength (negative net ecosystem C exchange) in intensive grazing systems: -64 ± 74 g C m yr (animal density reduction) and -81 ± 74 g C m yr (N and animal density reduction), against the baseline of -30.5 ± 69.5 g C m yr (LSU [livestock units] ≥ 0.76 ha yr). Simulations also indicated a strong effect of N fertilizer reduction on N fluxes, e.g. NO-N emissions decreased from 0.34 ± 0.22 (baseline) to 0.1 ± 0.05 g N m yr (no N fertilization). Simulated decline in grazing intensity had only limited impact on the N balance. The simulated pattern of enteric methane emissions was dominated by high model-to-model variability. The reduction in simulated offtake (animal intake + cut biomass) led to a doubling in net primary production per animal (increased by 11.6 ± 8.1 t C LSU yr across sites). The highest NO-N intensities (NO-N/offtake) were simulated at mown and extensively grazed arid sites. We show the possibility of using grassland models to determine sound mitigation practices while quantifying the uncertainties associated with the simulated outputs.

Anthropogenic nitrogen inputs cause major negative environmental impacts, including emissions of the important greenhouse gas N2O. Despite their importance, shifts in terrestrial N loss pathways driven by global change are highly uncertain. Here we present a coupled soil-atmosphere isotope model (IsoTONE) to quantify terrestrial N losses and N2O emission factors from 1850-2020. We find that N inputs from atmospheric deposition caused 51% of anthropogenic N2O emissions from soils in 2020. The mean effective global emission factor for N2O was 4.3 ± 0.3% in 2020 (weighted by N inputs), much higher than the surface area-weighted mean (1.1 ± 0.1%). Climate change and spatial redistribution of fertilisation N inputs have driven an increase in global emission factor over the past century, which accounts for 18% of the anthropogenic soil flux in 2020. Predicted increases in fertilisation in emerging economies will accelerate N2O-driven climate warming in coming decades, unless targeted mitigation measures are introduced.

Nitrous oxide (N 2 O) is a potent greenhouse gas that is primarily emitted from agriculture. Sampling limitations have generally resulted in discontinuous N 2 O observations over the course of any given year. The status quo for interpolating between sampling points has been to use a simple linear interpolation. This can be problematic with N 2 O emissions, since they are highly variable and sampling bias around these peak emission periods can have dramatic impacts on cumulative emissions. Here, we outline five gap-filling practices: linear interpolation, generalized additive models (GAMs), autoregressive integrated moving average (ARIMA), random forest (RF), and neural networks (NNs) that have been used for gap-filling soil N 2 O emissions. To facilitate the use of improved gap-filling methods, we describe the five methods and then provide strengths and challenges or weaknesses of each method so that model selection can be improved. We then outline a protocol that details data organization and selection, splitting of data into training and testing datasets, building and testing models, and reporting results. Use of advanced gap-filling methods within a standardized protocol is likely to increase transparency, improve emission estimates, reduce uncertainty, and increase capacity to quantify the impact of mitigation practices.