Benchmark data set for wheat growth models: field experiments and AgMIP multi-model simulations

The data set includes a current representative management treatment from detailed, qualitytested sentinel field experiments with wheat from four contrasting environments including Australia, The Netherlands, India and Argentina. Measurements include local daily climate data (solar radiation, maximum and minimum temperature, precipitation, surface wind, dew point temperature, relative humidity, and vapor pressure), soil characteristics, frequent growth, nitrogen in crop and soil, crop and soil water and yield components. Simulations include results from 27 wheat models and a sensitivity analysis with 26 models and 30 years (1981-2010) for each location, for elevated atmospheric CO2 and temperature changes, a heat stress sensitivity analysis at anthesis, and a sensitivity analysis with soil and crop management variations and a Global Climate Model end-century scenario.

Texas A&M University, USA, 40 CGIAR-ESSP Program on Climate Change, Agriculture and Food Security, International Centre for Tropical Agriculture (CIAT), A.A. 6713,Cali,Colombia. * email: sasseng@ufl.edu # Dr. Nadine Brisson passed away in 2011 while this work was being carried out.

Abstract:
The data set includes a current representative management treatment from detailed, qualitytested sentinel field experiments with wheat from four contrasting environments including Australia, The Netherlands, India and Argentina. Measurements include local daily climate data (solar radiation, maximum and minimum temperature, precipitation, surface wind, dew point temperature, relative humidity, and vapor pressure), soil characteristics, frequent growth, nitrogen in crop and soil, crop and soil water and yield components. Simulations include results from 27 wheat models and a sensitivity analysis with 26 models and 30 years  for each location, for elevated atmospheric CO 2 and temperature changes, a heat stress sensitivity analysis at anthesis, and a sensitivity analysis with soil and crop management variations and a Global Climate Model end-century scenario.
Keywords: wheat, field experimental data, simulations, sensitivity analysis, climate change impact 1 ORIGINAL PURPOSE: The original purpose of this data set was a model intercomparison of the AgMIP-Wheat Pilot as part of the Agricultural Model Intercomparison and Improvement project (http://www.agmip.org/). The field experimental data were selected from local representative, high-quality field experimental data sets with a grain yield range from 2 to 8 t ha -1 . The original experiments were for specific agronomic studies and included measurements of crop-model-ready single treatments of wheat experiments at four contrasting locations, including a location in The Netherlands (Wageningen (Groot and Willigen 1991)), Argentina (Balcarce (Travasso, Rodriguez and Grondona 1995)), India (New Delhi (Naveen 1986)), and Australia (Wongan Hills (Asseng et al. 1998)) representing a wide range of growing conditions. The data were quality-checked and considered high-quality sentinel sites. The experimental data were used in observation-model comparisons and as the foundation for a sensitivity and uncertainty analysis for climate change impact assessments. The study was published by Asseng et al. (2013) and Martre et al. (2014). Details of the field and simulated experimental design are supplied in the Supplementary of Asseng et al. (2013) and Martre et al. (2014). A subset of the data was also published by Challinor et al. (2014). Observed variables include grain yield (t ha -1 ), anthesis date (DOY); maturity date (DOY), grain N (kg N ha -1 ); grains per square meter (# m -2 ); cumulative evapotranspiration (mm); cumulative N mineralization (kg N ha -1 ); plant available soil water to maximum rooting depth (mm); soil mineral N to maximum rooting depth (kg N ha -1 ). Note, that not all measurements were available across all four experiments.

SIMULATION OF FIELD EXPERIMENTS:
Simulations for the four experiments were carried out by 27 wheat models (see Supplementary of Asseng et al. (2013)). The annual simulation outputs included: grain yield (t ha -1 ); above-ground biomass at anthesis (kg ha -1 ); above-ground biomass at maturity (kg ha -1 ); maximum leaf area index (LAI, m 2 m -2 ); anthesis date (DOY); maturity date (DOY); cumulative soil N leaching (kg N ha -1 ); cumulative soil water loss (mm); total above-ground N at anthesis (kg N ha -1 ); total above-ground N at maturity (kg N ha ); plant available soil water to maximum rooting depth (mm); soil mineral N to maximum rooting depth (kg N ha -1 ). Each of the 27 wheat models was used to simulate the field experiments in two separate steps: 1) with limited in-season information from the experiments being made available to the modelers (partial calibration or 'blind' simulations), and 2) all available information being made available to the modelers (full calibration). Simulations with partially calibrated models were included to allow a more objective model assessment. For the partial calibration or 'blind model test', modelers had no access to measurements of grain yield, biomass, and crop water and N dynamics, receiving information only on soil characteristics, initial soil-water conditions, daily weather data, crop management, and flowering and maturity dates. For full calibration, modelers had access to all available measurements, including withinseason and final biomass, water and N uptake, soil water and soil N, grain yield and yield components.

MODEL SENSITIVITY ANALYSIS:
In addition to simulations of the single-year experiments, simulations were carried out with long-term measured daily climate data (solar radiation, maximum and minimum temperature, precipitation, surface wind, dew point temperature, relative humidity, and vapor pressure) using measured soil characteristics, measured initial soil water and soil N contents, crop management, measured anthesis and maturity dates from the single-year-experiments. For the baseline, daily climate data for the period 1980-2010 were used for all locations (31 years of climate data are required to simulate 30 years of yields in The Netherlands and India). For the location in India, solar radiation was obtained from the NASA/POWER dataset that extends back to 1983 (http://power.larc.nasa.gov). Missing data for 1980 to 1983 were filled in using the Weatherman tool included in DSSAT 4.5. In addition, 2-meter wind speed (km d -1 ), dew point temperature (ºC), vapor pressure (hPa), and relative humidity (%) were estimated for each location from the NASA Modern Era Retrospective-Analysis for Research and Applications (MERRA). For the location in The Netherlands, measured wind speed and vapor pressure were available. In the sensitivity analysis, daily temperatures were changed in steps of 3°C from -3 to +9°C and atmospheric CO 2 concentrations were changed in steps of 90 ppm from 360 to 720 ppm. In addition, a sensitivity analysis of heat stress impact was carried out with the 1981-2010 climate data by introducing seven days of maximum temperature of 35°C starting at the measured anthesis day for each location. In a sensitivity analysis on variations in soil and crop management, at each location, the plant-available water-holding capacity of the local soil was varied by +/-20%, the crop sowing date was varied by +/-20 days and N-fertiliser applications were varied by +/-50% and simulated with the baseline (1981-2010) and a Global Climate Model end-century scenario. Model input (cultivar information and crop management), soil description and initial conditions data for simulation set up are in Excel spreadsheets provided with the data. All simulation results are in text files (tab delimited). The file name and variable names are explained in Excel spreadsheets as part of the data set.