R&D contributions to increasing and stabilizing yields
Agricultural R&D is a key driver in achieving high and stable yields. Yield instability is measured by CV of yield departure from time-varying average yield (i.e., yield anomaly). By definition, CV values rise with an increase in yield anomaly SD and/or a decrease in average yield. An analysis of country-level yield data for maize, rice, wheat, and soybean in 2001–2020 by the Food and Agriculture Organization (FAO) of the United Nations showed that increases in public expenditure on agricultural R&D are proportional to increases in average yield. For the four crops together, countries with large areas under cultivation (≥ 1 Mha) tended to have invested considerably more into R&D than countries with smaller areas of cropland (Fig. 1A). Although a larger yield anomaly SD was generally associated with higher average yields (Fig. 1B), the relationship between average yields and yield anomaly CV was not very strong (Fig. 1C). Many countries with higher levels of R&D expenditure were distributed in the lower half of the yield anomaly CV, i.e., below the median quantile regression curve (Fig. 1C), indicating these countries achieved more stable yields than countries with lower levels of R&D expenditure when average yields were comparable.
Recent changes in yield stability
In 2010, the combined global harvested area for the four crops together accounted for 646.2 Mha. The analysis of the grid yield dataset in this study covered two-thirds of this area, proving a spatially detailed overview. The results showed that, in 2001–2020, 19% of the global harvested area (119.2 Mha) had increased yield anomaly SD (Fig. 2B). However, because trend in average yield was positive in a large share of the global cropland (Fig. 2C), only 57% of areas with increased SD experienced a rise in CV during the same period (Fig. 2A). This is due to the fact that, by definition, yield anomaly CV decreases with increasing average yield.
Contributions of climate and management practices
The rise in yield anomaly SD is often compensated by a positive trend in average yields, giving the impression that yield instability remains unchanged or lifted, but the compensation is not systematic. In order to identify the origins of the changes in yield anomaly SD, the contributions of climate and management practices were estimated using the validated machine-learning yield anomaly models and factual and counterfactual climate scenarios (Supplementary Text S1). The effect of climate on yield anomaly SD was clarified through a comparative analysis of factual and non-warming counterfactual yields (fig. S1). Both yield series were simulated using the same random forest models associating yield anomalies with climate, soil, and management conditions (fc and ct, respectively; Materials and Methods). The results revealed that climate change in 2001–2019 adversely affected agro-climatic conditions and increased yield anomaly SD in a large share of the cropping areas. From the viewpoint of yield stability, it was estimated that unfavorable agro-climatic conditions under climate change, relative to those without human influences to the global climate systems, occurred in 47% of areas with increased SD (the percentage comes from 9-to-19, which is the ratio between 9% of unfavorable areas to 19% of areas with increased SD; Fig. 1B, Fig. 3A, C). Favorable agro-climatic conditions were observed in 24% (4-to-17) of areas with decreased SD.
We also examined the contributions of management practices to the changes in the yield anomaly SD (fig. S1). Specifically, we compared two factual yield series, i.e., a simulation with recent progress in management practices and a simulation without such improvements (fc and fc.em runs, respectively; Materials and Methods). The results showed that increases or decreases in yield anomaly SD are dependent upon the balance between the rate at which management practices advance and the rate at which the climate change impacts manifest. It was estimated that in 59% (10-to-17) of the areas with decreased SD, management practices adopted after the year 2008 had the effect of decreasing the yield anomaly SD under climate change (Fig. 1B, Fig. 3B, D). However, in 79% (15-to-19) of areas with increased SD, progress in management practices, including implementation of adaptation measures, was considered insufficient in offsetting the negative impacts of climate change on agro-climatic conditions.
Regions and crops lagged behind climate change
The cases under which the improvement of management practices was able to compensate or not the climate change impacts on yield stability were categorized as “kept-pace” and “lagged-behind”, respectively. These two categories were not evenly distributed across the geographic regions and crops. Overall, it was estimated that the lagged-behind category encompassed 60% of the analyzed area (Fig. 4, table S1). Although both areas coexisted within a region or crop, in certain regions, the lagged-behind area was predominant. These regions included South Asia, Southeast Asia, and Oceania, and Western Europe, followed by the Middle East, North Africa, and North America. Conversely, the kept-pace category was more prevalent in Eastern Europe, Central Asia, and Sub-Saharan Africa. In East Asia, Latin America, and the Caribbean, the two trends were mixed almost evenly. For crops, nearly 70% of the areas dedicated to rice and maize fell under the lagged-behind category. For wheat, the area was almost evenly divided between the two categories. Exceptionally, soybean was characterized as being in the kept-pace category.