The Global Gridded Crop Model Intercomparison phase 1 simulation dataset

The Global Gridded Crop Model Intercomparison (GGCMI) phase 1 dataset of the Agricultural Model Intercomparison and Improvement Project (AgMIP) provides an unprecedentedly large dataset of crop model simulations covering the global ice-free land surface. The dataset consists of annual data fields at a spatial resolution of 0.5 arc-degree longitude and latitude. Fourteen crop modeling groups provided output for up to 11 historical input datasets spanning 1901 to 2012, and for up to three different management harmonization levels. Each group submitted data for up to 15 different crops and for up to 14 output variables. All simulations were conducted for purely rainfed and near-perfectly irrigated conditions on all land areas irrespective of whether the crop or irrigation system is currently used there. With the publication of the GGCMI phase 1 dataset we aim to promote further analyses and understanding of crop model performance, potential relationships between productivity and environmental impacts, and insights on how to further improve global gridded crop model frameworks. We describe dataset characteristics and individual model setup narratives.

Global wheat production with 1.5 and 2.0°C above pre-industrial warming

Efforts to limit global warming to below 2°C in relation to the pre-industrial level are under way, in accordance with the 2015 Paris Agreement. However, most impact research on agriculture to date has focused on impacts of warming >2°C on mean crop yields, and many previous studies did not focus sufficiently on extreme events and yield interannual variability. Here, with the latest climate scenarios from the Half a degree Additional warming, Prognosis and Projected Impacts (HAPPI) project, we evaluated the impacts of the 2015 Paris Agreement range of global warming (1.5 and 2.0°C warming above the pre-industrial period) on global wheat production and local yield variability. A multi-crop and multi-climate model ensemble over a global network of sites developed by the Agricultural Model Intercomparison and Improvement Project (AgMIP) for Wheat was used to represent major rainfed and irrigated wheat cropping systems. Results show that projected global wheat production will change by -2.3% to 7.0% under the 1.5°C scenario and -2.4% to 10.5% under the 2.0°C scenario, compared to a baseline of 1980-2010, when considering changes in local temperature, rainfall, and global atmospheric CO₂ concentration, but no changes in management or wheat cultivars. The projected impact on wheat production varies spatially; a larger increase is projected for temperate high rainfall regions than for moderate hot low rainfall and irrigated regions. Grain yields in warmer regions are more likely to be reduced than in cooler regions. Despite mostly positive impacts on global average grain yields, the frequency of extremely low yields (bottom 5 percentile of baseline distribution) and yield inter-annual variability will increase under both warming scenarios for some of the hot growing locations, including locations from the second largest global wheat producer-India, which supplies more than 14% of global wheat. The projected global impact of warming <2°C on wheat production is therefore not evenly distributed and will affect regional food security across the globe as well as food prices and trade.

Multimodel ensembles improve predictions of crop-environment-management interactions

A recent innovation in assessment of climate change impact on agricultural production has been to use crop multimodel ensembles (MMEs). These studies usually find large variability between individual models but that the ensemble mean (e-mean) and median (e-median) often seem to predict quite well. However, few studies have specifically been concerned with the predictive quality of those ensemble predictors. We ask what is the predictive quality of e-mean and e-median, and how does that depend on the ensemble characteristics. Our empirical results are based on five MME studies applied to wheat, using different data sets but the same 25 crop models. We show that the ensemble predictors have quite high skill and are better than most and sometimes all individual models for most groups of environments and most response variables. Mean squared error of e-mean decreases monotonically with the size of the ensemble if models are added at random, but has a minimum at usually 2-6 models if best-fit models are added first. Our theoretical results describe the ensemble using four parameters: average bias, model effect variance, environment effect variance, and interaction variance. We show analytically that mean squared error of prediction (MSEP) of e-mean will always be smaller than MSEP averaged over models and will be less than MSEP of the best model if squared bias is less than the interaction variance. If models are added to the ensemble at random, MSEP of e-mean will decrease as the inverse of ensemble size, with a minimum equal to squared bias plus interaction variance. This minimum value is not necessarily small, and so it is important to evaluate the predictive quality of e-mean for each target population of environments. These results provide new information on the advantages of ensemble predictors, but also show their limitations.

Author Correction: The uncertainty of crop yield projections is reduced by improved temperature response functions

Nature Plants 3, 17102 (2017); published online 17 July 2017; corrected online 27 September 2017.

Erratum: The uncertainty of crop yield projections is reduced by improved temperature response functions

This corrects the article DOI: 10.1038/nplants.2017.102.

The uncertainty of crop yield projections is reduced by improved temperature response functions

Increasing the accuracy of crop productivity estimates is a key element in planning adaptation strategies to ensure global food security under climate change. Process-based crop models are effective means to project climate impact on crop yield, but have large uncertainty in yield simulations. Here, we show that variations in the mathematical functions currently used to simulate temperature responses of physiological processes in 29 wheat models account for >50% of uncertainty in simulated grain yields for mean growing season temperatures from 14 °C to 33 °C. We derived a set of new temperature response functions that when substituted in four wheat models reduced the error in grain yield simulations across seven global sites with different temperature regimes by 19% to 50% (42% average). We anticipate the improved temperature responses to be a key step to improve modelling of crops under rising temperature and climate change, leading to higher skill of crop yield projections.

Multimodel ensembles of wheat growth: many models are better than one

Crop models of crop growth are increasingly used to quantify the impact of global changes due to climate or crop management. Therefore, accuracy of simulation results is a major concern. Studies with ensembles of crop models can give valuable information about model accuracy and uncertainty, but such studies are difficult to organize and have only recently begun. We report on the largest ensemble study to date, of 27 wheat models tested in four contrasting locations for their accuracy in simulating multiple crop growth and yield variables. The relative error averaged over models was 24-38% for the different end-of-season variables including grain yield (GY) and grain protein concentration (GPC). There was little relation between error of a model for GY or GPC and error for in-season variables. Thus, most models did not arrive at accurate simulations of GY and GPC by accurately simulating preceding growth dynamics. Ensemble simulations, taking either the mean (e-mean) or median (e-median) of simulated values, gave better estimates than any individual model when all variables were considered. Compared to individual models, e-median ranked first in simulating measured GY and third in GPC. The error of e-mean and e-median declined with an increasing number of ensemble members, with little decrease beyond 10 models. We conclude that multimodel ensembles can be used to create new estimators with improved accuracy and consistency in simulating growth dynamics. We argue that these results are applicable to other crop species, and hypothesize that they apply more generally to ecological system models.

Multi-scale geospatial agroecosystem modeling: a case study on the influence of soil data resolution on carbon budget estimates

The development of effective measures to stabilize atmospheric CO2 concentration and mitigate negative impacts of climate change requires accurate quantification of the spatial variation and magnitude of the terrestrial carbon (C) flux. However, the spatial pattern and strength of terrestrial C sinks and sources remain uncertain. In this study, we designed a spatially-explicit agroecosystem modeling system by integrating the Environmental Policy Integrated Climate (EPIC) model with multiple sources of geospatial and surveyed datasets (including crop type map, elevation, climate forcing, fertilizer application, tillage type and distribution, and crop planting and harvesting date), and applied it to examine the sensitivity of cropland C flux simulations to two widely used soil databases (i.e. State Soil Geographic-STATSGO of a scale of 1:250,000 and Soil Survey Geographic-SSURGO of a scale of 1:24,000) in Iowa, USA. To efficiently execute numerous EPIC runs resulting from the use of high resolution spatial data (56m), we developed a parallelized version of EPIC. Both STATSGO and SSURGO led to similar simulations of crop yields and Net Ecosystem Production (NEP) estimates at the State level. However, substantial differences were observed at the county and sub-county (grid) levels. In general, the fine resolution SSURGO data outperformed the coarse resolution STATSGO data for county-scale crop-yield simulation, and within STATSGO, the area-weighted approach provided more accurate results. Further analysis showed that spatial distribution and magnitude of simulated NEP were more sensitive to the resolution difference between SSURGO and STATSGO at the county or grid scale. For over 60% of the cropland areas in Iowa, the deviations between STATSGO- and SSURGO-derived NEP were larger than 1MgCha(-1)yr(-1), or about half of the average cropland NEP, highlighting the significant uncertainty in spatial distribution and magnitude of simulated C fluxes resulting from differences in soil data resolution.

Evaluation of three field-based methods for quantifying soil carbon

Three advanced technologies to measure soil carbon (C) density (g C m⁻²) are deployed in the field and the results compared against those obtained by the dry combustion (DC) method. The advanced methods are: a) Laser Induced Breakdown Spectroscopy (LIBS), b) Diffuse Reflectance Fourier Transform Infrared Spectroscopy (DRIFTS), and c) Inelastic Neutron Scattering (INS). The measurements and soil samples were acquired at Beltsville, MD, USA and at Centro International para el Mejoramiento del Maíz y el Trigo (CIMMYT) at El Batán, Mexico. At Beltsville, soil samples were extracted at three depth intervals (0–5, 5–15, and 15–30 cm) and processed for analysis in the field with the LIBS and DRIFTS instruments. The INS instrument determined soil C density to a depth of 30 cm via scanning and stationary measurements. Subsequently, soil core samples were analyzed in the laboratory for soil bulk density (kg m⁻³), C concentration (g kg⁻¹) by DC, and results reported as soil C density (kg m⁻²). Results from each technique were derived independently and contributed to a blind test against results from the reference (DC) method. A similar procedure was employed at CIMMYT in Mexico employing but only with the LIBS and DRIFTS instruments. Following conversion to common units, we found that the LIBS, DRIFTS, and INS results can be compared directly with those obtained by the DC method. The first two methods and the standard DC require soil sampling and need soil bulk density information to convert soil C concentrations to soil C densities while the INS method does not require soil sampling. We conclude that, in comparison with the DC method, the three instruments (a) showed acceptable performances although further work is needed to improve calibration techniques and (b) demonstrated their portability and their capacity to perform under field conditions.

Simulating long-term and residual effects of nitrogen fertilization on corn yields, soil carbon sequestration, and soil nitrogen dynamics

Soil carbon sequestration (SCS) has the potential to attenuate increasing atmospheric CO2 and mitigate greenhouse warming. Understanding of this potential can be assisted by the use of simulation models. We evaluated the ability of the EPIC model to simulate corn (Zea mays L.) yields and soil organic carbon (SOC) at Arlington, WI, during 1958-1991. Corn was grown continuously on a Typic Argiudoll with three N levels: LTN1 (control), LTN2 (medium), and LTN3 (high). The LTN2 N rate started at 56 kg ha(-1) (1958), increased to 92 kg ha(-1) (1963), and reached 140 kg ha(-1) (1973). The LTN3 N rate was maintained at twice the LTN2 level. In 1984, each plot was divided into four subplots receiving N at 0, 84, 168, and 252 kg ha(-1). Five treatments were used for model evaluation. Percent errors of mean yield predictions during 1958-1983 decreased as N rate increased (LTN1 = -5.0%, LTN2 = 3.5%, and LTN3 = 1.0%). Percent errors of mean yield predictions during 1985-1991 were larger than during the first period. Simulated and observed mean yields during 1958-1991 were highly correlated (R2 = 0.961, p < 0.01). Simulated SOC agreed well with observed values with percent errors from -5.8 to 0.5% in 1984 and from -5.1 to 0.7% in 1990. EPIC captured the dynamics of SOC, SCS, and microbial biomass. Simulated net N mineralization rates were lower than those from laboratory incubations. Improvements in EPIC's ability to predict annual variability of crop yields may lead to improved estimates of SCS.