Climate smart agriculture and global food-crop production

represent about 41% of the global harvested area and approximately 64% of the estimated 2–3 Gt CO2e per year emitted by crop production globally [27]. The effects of CSA practices on production, food security, and GHG emissions were assessed via comparison with the out- comes of a business-as-usual (BAU) scenario in which farmers retain the current practices during the period 2010–2050. All scenarios, BAU and alternatives, were created using the IMPACT system of models [26,28] which links crop and climate models to the core economic model in which agricultural production is represented across 320 sub-national regions called “food production units” (FPUs, see Supplementary Discussion S1 in S1 File). Because of the linkages among these mod- els, IMPACT’s outputs reflect the interactions between biophysical, economic and population trends, the combination of which represents the functioning of the global food market. A styl- ized representation of the modeling steps and of the information flow is provided in Fig 1. The process represented in the figure is used to generate each one of the scenarios described in the sections below. 2.1 Simulation of production, prices and food security—the core of the modeling system To simulate the effects of CSA adoption on yields, production, harvested areas, world com- modity prices and indicators of food security, we linked the spatially-disaggregated data of three models: the Spatial Production Allocation Model (SPAM) [29], the Decision Support System for Agrotechnology Transfer (DSSAT) [30], and the International Model for Policy

Analysis of Agricultural Commodities and Trade (IMPACT v3.3) [28] (Supplementary Discus- sion S1 and Supplementary Fig 1 in S1 File). The SPAM model was used to identify the location of crop production on a global grid in which each grid-cell is of size 30 arc minutes (a square of approximately 56 km by 56 km at the equator). For each grid-cell, SPAM provides a database cataloging the dominant crops, repre- sentative cultivars, management practices, and the inputs used. Data about climate, irrigation type, and soil properties were geo-linked to each SPAM grid-cell and were essential to run crop model simulations at the grid-cell level. The DSSAT crop model was used to simulate crop yields, with current and or with CSA practices, as a function of the interaction between biophysical elements of the crop systems (e.g. soil, weather, and crop) and management prac- tices (e.g., tillage, nutrient application, and water availability). After the completion of a cali- bration step (Supplementary Discussion S2 with Supplementary Table 2 and Supplementary Figs 2 and 3 in S1 File), the yield responses were aggregated to evaluate agriculture production across IMPACT’s FPUs. IMPACT utilizes the yield responses to agricultural practices as shift- ers for the crop-specific supply curves and for the yield’s growth rates [1,9,28,31]. Starting from these changes in growth rates, the systems of equations at the core of the model endoge- nously determine crop areas, agricultural production, commodity prices and food availability.

In turn, taking into consideration population and income growth, changes in food availability translate into changes in the regional availability of kilocalories [28]. In IMPACT, the share of people at risk of hunger (i.e. suffering from undernourishment) is calculated based on an empirical correlation between the share of undernourished people within the population and the relative availability of food. The calculation is adapted from Fischer et al. [28,32]. The share of undernourished children under the age of five is based on the calculation of the average cal- orie availability per capita per day, women’s access to secondary education, the ratio of female to male life expectancy at birth, and health and sanitation conditions [28,33]. It is an estimate of undernourishment in terms of weight for age. PLOS ONE Climate smart agriculture and global food-crop production PLOS ONE | https://doi.org/10.1371/journal.pone.0231764 April 29, 2020 3 / 19