Effects of Environmental and Management Factors on Worldwide Maize and Soybean

Tzu-Shun Lin, Yang Song, Atul K Jain, Peter Lawrence, and Haroon S Kheshgi 11 Department of Atmospheric Sciences, University of Illinois, Urbana, IL 61801, USA 12 2 Department of Hydrology and Atmospheric Sciences, University of Arizona, Tucson, AZ, 13 85721, USA 14 National Center for Atmospheric Research, Boulder, CO 80305, USA 15 ExxonMobil Research and Engineering Company, Annandale, NJ 08801, USA 16 17


21
Text S1. Bias Correction of Future Climate 22 The bias correction method proposed by Hawkins et al. (2013) is used in this study to correct the 23 six atmospheric variables, including downward shortwave and longwave radiation, surface 24 pressure, wind speed, surface temperature and specific humidity from CESM outputs. This method 25 considers the mean and variability differences between CRU-NCEP reanalysis and CESM output

35
The correction of future precipitation is calculated using the following equation: 36 P cor = P raw × P obs,m P raw,m ̅̅̅̅̅̅̅̅̅̅ ̅̅̅̅̅̅̅ (SE2) 37 whereP cor is bias-corrected model precipitation, P raw stands for uncorrected corresponding 38 precipitation, P obs,m P raw,m ̅̅̅̅̅̅̅̅̅̅ is the ratio of monthly mean precipitation from CRU-NCEP to that from 39 CESM for the same reference period (Déqué et al., 2007). This method reduces mean biases 40 between models and observations; however, it does not take coefficient of variance of the modeled 41 precipitation into account.
historcal and under two future scenarios at the same rates as the N fertilizer amount. Then we add 91 spatial and time varying N fertilizer N manure, and N deposition amounts to estimate N input.

93
Text S4. Estimation of Irrigation Water Amount 94 The irrigation amount (Wirrig) is estimated as the soil moisture deficit between irrigation target (Wt) 95 and soil water content (Wliq) within the root-zone as follows: where Wt is defined as soil moisture content without water stress (i.e., WS is 1.0) for crop 98 photosynthesis during the growing period.

99
The irrigation scheme adds Wirrig directly to the topsoil layer, analogously to drip irrigation. 100 The estimated Wirrig is withdrawn first from surface runoff and drainage. If surface runoff plus 101 drainage is less than Wirrig, the remaining requirement of irrigation is extracted from river water 102 storage (i.e., accumulated total runoff). To estimate these water we have incorporated a river 103 transport module (RTM) into ISAM (Sharma et al, 2018). RTM is used to distribute total runoff 104 from the land surface model to the downstream systems such as rivers and oceans. In RTM the 105 water is routed from each grid cell to the neighboring grid cell by using a linear transport scheme 106 (Oleson et al., 2013). However, the water demand for crops may still not be met with these sources 107 in some dry seasons or areas, because the water resources and management, such as dynamic  (Leng et al., 2015). In this study we assume that the deficit of irrigation 111 demand is met by the groundwater.

112
Irrigation is applied at each time step when the leaf area index (LAI) of the crop is greater than 113 zero and root-zone soil water stress (WS) is less than 1.0 (i.e., water is limiting photosynthesis).

114
The WS is expressed as an index ranging from 0 to 1in ISAM (Song et al., 2016 The CV is calculated for temperature (K) and precipitation (mm) respectively as follows: where y is the year, CV is the variation coefficient of temperature or precipitation in year y, N is 143 the number of years, σ is the standard deviation of temperature or precipitation in year y, and μ 144 the annual mean temperature or precipitation in year y. where T air7d is the 7-day running mean of daily air temperature, T soil8d 8-day running mean of 175 daily root-zone average soil temperature, P d8 accumulated precipitation of the past eight 176 consecutive days, and P m_avg 50-year average of monthly precipitation. Other crop-specific 177 variables are defined in Table S1.

178
The second method to calculate the planting day (DOYphu) assumes the day when the 179 accumulated phenological heat unit (PHU0) reaches a certain minimum threshold (β*PHUmin where PHUmin is the mean of annual PHU0 for the period 1901-1950. β is the fractional parameter 185 for maize (0.1) and soybean (0.12). 186 We evaluate simulated planting time with the data compiled by AgMIP (Elliott et al., 2015).

187
This data is compiled using planting time data from two global crop calendars SAGE (Sacks et al., where Rseed: seeding rate; Rseed_ref: reference seeding rate; and Cstorage_ref: referenced carbon storage 211 in seed.

212
In the previous version of ISAM, Rseed, Rseed_ref, and Rseed_ref kept constant at each grid cell 213 according to the collected data in the United States (US) for maize and soybean (  where ε is 0.04 mol mol -1 for maize (Arora, 2003); and 0.08 for Soybean (Sellers et al., 1996) 270 is quantum yield of electron transport, α (=4.6 µmol J -1 ) is conversion from photosynthetically active radiation to photosynthetic photon flux, and ø (W m -2 ) is absorbed photosynthetically active radiation, which is calculated from solar fluxes using the two-stream approximation and varies 273 between sunlit and shaded leaves for photosynthesis (Song et al., 2013).

274
To address this issue, we added curvature to the light response curve in J for the revised version Text S10. The Calculation of the Percent Bias (PBIAS) 336 The PBIAS is calculated to compare modeled yield with measured yield at regional and global 337 scales: where Y i,j o and Y i,j m are the observed and modeled yearly yields for the available data i in the year j.

340
M is the number of available data points. N is the number of years for each available data. To assess the performance models against FACE site data in Table S4, AgMIP project selected

359
But temperatures are held constant for the AgMIP models experiment results reported in Table S6.     Table S6. Maize and soybean yields (t/ha) at global and regional scales averaged over the period