Crop production and global food security in relation to climate variation: an empirical analysis

The challenge of meeting increasing global food demand is amplified by climate change. Crop yield is vulnerable to extreme conditions, including heatwaves, droughts and downpours, leading to widespread concern about negative effects of climate change on food security. This thesis describes a novel empirical analysis of total production, yields and harvested area data for three major crops (wheat, maize and soybean), using a unique, global, gridded agricultural time-series data set. Trend analysis is applied to changes in production, yield and harvested area of these three crops. Machine learning is used to quantify their responses to climate. A new methodology is introduced to identify “shocks”. Results show a more complex dynamics of agricultural production than is suggested by current liter- ature. Large changes in regional production, driven by harvested area rather than yield, have been driven by policy shifts. A large “killing degree-day” sum depresses yields for some regions and crops, but enhances them in others. Heat deficits can be as deleterious as heatwaves. Shocks can be negative or positive. Production variability has increased, but major negative shocks have been few, and have not become more frequent. Production shocks have been caused as often by changes in harvested area as in yield. These findings do not support a universal negative effect of climate change on crop production. More- over, stable global food supplies will not be assured by maximizing yields. It is equally important that farmers in different countries and environments grow a variety of crops. Climate-related risk is currently concentrated in the most productive baskets, exposing the global food supply to avoidably high risk. Increasing frequencies of climate extremes in the main producing areas only make such shocks more likely. Various measures that are not directly related to climate would help to make global food supplies more resilient.Open Acces