National-scale changes in crop diversity through the Anthropocene

Abstract

Expansion of crops beyond their centres of domestication is a defining feature of the Anthropocene Epoch. This process has fundamentally altered the diversity of croplands, with likely consequences for the ecological functioning and socio-economic stability of agriculture under environmental change. While changes in crop diversity through the Anthropocene have been quantified at large spatial scales, the patterns, drivers, and consequences of change in crop diversity and biogeography at national-scales remains less explored. We use production data on 339 crops, grown in over 150 countries from 1961 to 2017, to quantify changes in country-level crop richness and evenness. Virtually all countries globally have experienced significant increases in crop richness since 1961, with the early 1980s marking a clear onset of a ~ 9-year period of increase in crop richness in countries worldwide. While these changes have increased the similarity of diversity of croplands among countries, only half of countries experienced increases in crop evenness through time. Ubiquitous increases in crop richness within nearly all countries between 1980 and 2000 are a unique biogeographical feature of the Anthropocene. At the same time, we detected opposing changes in crop evenness, and only modest signatures of increased homogenization of croplands among countries. Therefore context-dependent and, at least, national-scale assessments are needed to understand and predict how changes in crop diversity influence agricultural resistance and resilience to environmental change.

Figure 1

Schematic representation of three indicators of change in crop commodity group diversity, derived from piecewise models predicting crop commodity group richness as a function of year. Detailed explanations of Indicator 1–3 are presented in the Methods section. Data shown here as the example is from Canada, with black dots representing the number of commodities reported by the Food and Agricultural Organization, for a given year. Black trendline represents the piecewise model fit, gray bands represent the 95% confidence limits surrounding the model, and red lines represent model parameters and indicators derived from the model. Note: the figure presented here demonstrates changes in crop commodity group richness (S), though this framework was also employed for assessing change in crop group evenness (J′).

Figure 2

Maps and histograms of three indicators of crop commodity group richness (S) change across 165 countries. Values for all three indicators for each country were derived from piecewise linear models predicting S as a function of year (see Fig. 1 for example). Countries coloured gray in the maps were those where either data was not available or the piecewise models failed to converge (denoted in Table S1). Histograms and associated descriptive statistics for each indicator are also presented, with means (± s.d.) or medians (± m.a.d.) denoted visually by the points and error bars below the histograms. All piecewise model parameters for each country are presented in Table S1. Maps were generated using the mapCountryData function in the rworldmap R package.

Figure 3

Latitudinal patterns in crop group richness across 164 countries in 1961 and 2017. Only countries with data from both years are included in this analysis. Trend lines represent results from linear models (including a 2nd-order polynomial term) that predict crop group richness as a function of latitude (and latitude²) in 1961 (dashed line, filled circles) and 2017 (solid line, open circles). Complete diagnostics for both models are presented in Table 1.

Figure 4

Maps and histograms of three indicators of crop commodity group evenness (Pielou’s evenness index (J′)) across 185 countries. Values for all three indicators for each country were derived from piecewise linear models predicting J′ as a function of year, where harvested area (in ha) was used to approximate group abundance. Countries coloured gray in the maps were those where either data was not available or the piecewise models failed to converge (denoted in Table S2). Histograms and associated descriptive statistics for each indicator are also presented, with means (± s.d.) or medians (± m.a.d.) denoted visually by the points and error bars below the histograms. All piecewise model parameters for each country are presented in Table S2. Maps were generated using the mapCountryData function in the rworldmap R package.