Complex drought patterns robustly explain global yield loss for major crops

Abstract

Multi-purpose crops as maize, rice, soybean, and wheat are key in the debate concerning food, land, water and energy security and sustainability. While strong evidence exists on the effects of climate variability on the production of these crops, so far multifaceted attributes of droughts-magnitude, frequency, duration, and timing-have been tackled mainly separately, for a limited part of the cropping season, or over small regions. Here, a more comprehensive assessment is provided on how droughts with their complex patterns-given by their compound attributes-are consistently related to negative impacts on crop yield on a global scale. Magnitude and frequency of both climate and yield variability are jointly analysed from 1981 to 2016 considering multiscale droughts, i.e., dry conditions occurring with different durations and timings along the whole farming season, through two analogous and standardized indicators enabling comparison among crops, countries, and years. Mainly winter wheat and then spring wheat, soybean and the main maize's season reveal high susceptibility of yield under more complex drought patterns than previously assessed. The second maize's season and rice present less marked and more uncertain results, respectively. Overall, southern and eastern Europe, the Americas and sub-Saharan Africa presents multi-crop susceptibility, with eastern Europe, Middle East and Central Asia appearing critical regions for the most vulnerable crop, which is wheat. Finally, yield losses for wheat and soybean clearly worsen when moving from moderate to extreme multiscale droughts.

Figure 1

Magnitude and frequency for co-occurring lower-than-normal both moisture and yield. SPEI duration-timing matrix of merged p-values measuring the overall aggregated significance among: (i) Asymmetry of contingency-tables fully skewed towards the co-occurrence of lower-than-normal both yield and moisture (LY_D); (ii) differences in SYI magnitudes under SPEI ≤ − 1 vs. SPEI > − 1; and iii) differences in SPEI magnitudes under SYI ≤ − 1 vs. SYI > − 1. Colors refer to p-value classification as reported in the legend. Lower p-values represent more significant asymmetry towards LY_D (i), significantly lower SYI values under drought vs. non-drought conditions (ii) and significantly lower SPEI values identified in cases of lower-than-normal vs. normal or higher-than-normal yields (iii). Grey cells represent no minimum sample reached to do either SPEI or SYI magnitude analyses or no asymmetry fully skewed toward LY_D. Diagonal lines indicate the case of overall higher SYI (or SPEI) under lower-than-normal moisture (or yield). SPEI ending (in Months before the Harvesting month, MBH) represents the timing and indicates the final month of the consecutive months period for SPEI durations longer than 1 month.

Figure 2

Country susceptibility to drought-low yield association. Average occurrence, for each country, of SPEI duration-timing combinations with SPEI ≤ − 1 across all the years with SYI ≤ − 1. The values obtained from all cropping systems are classified into susceptibility classes considering five quantiles while zero is presented separately (light yellow) and distinguished from “no data” land areas (dark grey). Pie diagram represents the susceptibility classes’ share (within cultivated surface) among the coloured countries (i.e., excluding no data areas); the global cultivated area is reported below the pie in Km².

Figure 3

Yearly crop susceptibility to drought-low yield association. Average global occurrence, for each year and cropping system, of SPEI duration-timing combinations with SPEI ≤ − 1 across all the occurrences of SYI ≤ − 1. Single points (some of which overlapping) represent each one year, the box limits indicate the 25th and 75th percentile while the internal horizontal line is the median. The whiskers represent the minimum and maximum after exclusion of outliers (i.e., values outside 1.5 times the interquartile range from the 25th and 75th percentile, respectively). Red line is the 2nd quantile among occurrence values assumed, as in Fig. 2, as the limit between Low and Medium susceptibility.

Figure 4

SYI values under drought related SPEI classes. Distribution of SYI values when SYI ≤ − 1 (left panel, yellow box plots), SYI ≤ − 1.5 (central panel, blue box plots) and SYI ≤ − 2 (right panel, green box plots) for − 1.5 < SPEI ≤ − 1, − 2 < SPEI ≤ − 1.5 and SPEI ≤ − 2. The box boundaries indicate the 25th and 75th percentile, the whiskers the 10th and 90th percentiles, and the red bar the median. Numbers enter parentheses indicate the sample available under each SPEI class and SYI range, across the years, countries and SPEI timings and durations considered. The letters with the same style (bold capital, underlined capital, or lowercase) indicate (if different) that in the pair-wise comparison between SPEI classes the related SYI values are significantly different at the 95% confidence level (p-value ≤ 0.05), the asterisk meaning instead a confidence level between 90 and 95% (0.05 < p-value ≤ 0.1).

Figure 4

SYI values under drought related SPEI classes. Distribution of SYI values when SYI ≤ − 1 (left panel, yellow box plots), SYI ≤ − 1.5 (central panel, blue box plots) and SYI ≤ − 2 (right panel, green box plots) for − 1.5 < SPEI ≤ − 1, − 2 < SPEI ≤ − 1.5 and SPEI ≤ − 2. The box boundaries indicate the 25th and 75th percentile, the whiskers the 10th and 90th percentiles, and the red bar the median. Numbers enter parentheses indicate the sample available under each SPEI class and SYI range, across the years, countries and SPEI timings and durations considered. The letters with the same style (bold capital, underlined capital, or lowercase) indicate (if different) that in the pair-wise comparison between SPEI classes the related SYI values are significantly different at the 95% confidence level (p-value ≤ 0.05), the asterisk meaning instead a confidence level between 90 and 95% (0.05 < p-value ≤ 0.1).

Figure 5

Example (yellow cells) of the sub-set of SPEI durations and timings considered if assuming harvest in August and a 5-month season length (i.e., planting in April). The superscripts PM and HM represent the planting and harvesting month, respectively. In each yellow cell, the small letters represent the consecutive months considered in the SPEI-d calculation and in particular: f = February, m = March, a = April, m = May, j = June, j = July, a = August.

Figure 6

Contingency table among SPEI (columns) and SYI (rows) (macro)classes, see Tables 2 and 3 for acronyms in bold. Small letters in the orange side supported more in-depth analyses of the combination between LY and D macro classes.