Research ArticleAGRICULTURE

Spatial variations in crop growing seasons pivotal to reproduce global fluctuations in maize and wheat yields

See allHide authors and affiliations

Science Advances  21 Nov 2018:
Vol. 4, no. 11, eaat4517
DOI: 10.1126/sciadv.aat4517
  • Fig. 1 Explained variance of country-level yield anomalies for maize (A) and wheat (B).

    R2 values between four different LPJmL simulations (see Table 1) and observed FAO yield anomaly time series (1980–2010) are highlighted for the 10 main producer countries showing highest weather sensitivity (see tables S1 and S2 for all main producer countries). Statistical significance of the explained variance is indicated through chart symbols (large dots if P < 0.001; small dots if P < 0.1; circle if not significant, that is, P ≥ 0.1). Mean R2 values across displayed most weather-sensitive main producers (and across all main producers in parentheses) are shown in the top-right corner.

  • Fig. 2 Observed and simulated influences of extreme weather on global crop yields.

    Global average impact of heat waves (first column) and droughts (second column) on maize (top row) and wheat (bottom row) yields [1964–2007; all events recorded in EM-DAT (13)]. Composites are based on 7-year time windows of country-level yields centered on the respective event. Results are shown for observed FAO time series (gray) and for different LPJmL simulations (yellow to red; see Table 1). Dashes along the y axis indicate the 25th to 75th percentile range of both observations and simulations at the event year. The number of composited events is indicated in the top-right corner (n); for details, see the “Statistical analysis” section and table S3 for the list of considered extreme events.

  • Fig. 3 Effects of growing season adjustment on maize rainfall deficit.

    Boxplots in (A) highlight the national relative rainfall deficit during heat waves for LPJmL–Ref and LPJmL–PHU simulations, separated for all occurrences (1964–2007, n = 65) and Europe 2003 (n = 10). Rainfall deficit is calculated as the relative difference between simulated growing season precipitation and the long-term average precipitation during the reference growing season. The Germany 2003 case is indicated with asterisks (A) and detailed through (B) to (E): Effect of the growing season timing [(B), MIRCA2000 reference in black] on precipitation exposure [(C), PGFv2.1 forcing], rainfall deficits (D), and standardized yield anomalies (E) from 2000 to 2006. Results for LPJmL–Ref and LPJmL–PHU are shown in orange and red, respectively, and FAO observations are shown in gray. The fraction of the explained variance in yield anomalies (R2, in percent) is indicated in the bottom with statistical significance in parentheses (*** if P < 0.001; n.s. (not significant) if P ≥ 0.1).

  • Table 1 Experimental design.

    Global gridded crop model versions used in this study are characterized regarding assumptions on irrigation, phenological heat units (PHU), and sowing dates; see Materials and Methods for details.

    Model (code reference)Irrigation assumptionsPhenology assumptions
    LPJmL–Ref (37)Reference irrigation: Full, that is, unconstrained
    irrigation on irrigated land, rainfed conditions
    in rainfed systems
    Reference phenology: semistatic global PHU parameter
    not based on observations; sowing dates internally
    derived from climate conditions
    LPJmL–WaterLimIrr (38)Advanced irrigation: Mechanistic representation
    of irrigation systems and surface water availability
    Same as above
    LPJmL–PHU (this study)Same as aboveSpatially derived PHU requirements per crop and
    grid cell to match targeted growing seasons;
    sowing dates prescribed according to observational
    data and model simulations
    LPJmL–NoWaterStress (37)Full irrigation: No water stress in rainfed
    or irrigated systems
    Same as LPJmL–Ref

Supplementary Materials

  • Supplementary material for this article is available at http://advances.sciencemag.org/cgi/content/full/4/11/eaat4517/DC1

    Fig. S1. Root mean square error of standardized country-level yield anomalies for maize and wheat.

    Fig. S2. Observed and simulated historical maize and wheat yield anomaly time series.

    Fig. S3. Sensitivity of mean explained variance to number of countries considered.

    Fig. S4. Evaluation of available growing season inputs.

    Fig. S5. Best-performing crop calendar per country.

    Fig. S6. Evaluation of different climate inputs.

    Fig. S7. Influences of heat waves and droughts on rainfed and irrigated yields.

    Fig. S8. Observed and simulated influences of the 2003 European heat wave on maize yields.

    Fig. S9. Effects of growing season adjustment on wheat rainfall deficit.

    Table S1. Explained variances and RMSE of maize country-level yield anomalies.

    Table S2. Explained variances and RMSE of wheat country-level yield anomalies.

    Table S3. List of extreme events considered in this study.

  • Supplementary Materials

    This PDF file includes:

    • Fig. S1. Root mean square error of standardized country-level yield anomalies for maize and wheat.
    • Fig. S2. Observed and simulated historical maize and wheat yield anomaly time series.
    • Fig. S3. Sensitivity of mean explained variance to number of countries considered.
    • Fig. S4. Evaluation of available growing season inputs.
    • Fig. S5. Best-performing crop calendar per country.
    • Fig. S6. Evaluation of different climate inputs.
    • Fig. S7. Influences of heat waves and droughts on rainfed and irrigated yields.
    • Fig. S8. Observed and simulated influences of the 2003 European heat wave on maize yields.
    • Fig. S9. Effects of growing season adjustment on wheat rainfall deficit.
    • Table S1. Explained variances and RMSE of maize country-level yield anomalies.
    • Table S2. Explained variances and RMSE of wheat country-level yield anomalies.
    • Table S3. List of extreme events considered in this study.

    Download PDF

    Files in this Data Supplement:

Stay Connected to Science Advances

Navigate This Article