Modelling cropping periods of grain crops at the global scale

Crop models require information on both weather and agronomic decisions to simulate crop productivity and to design adaptation strategies. Due to the lack of observational data, previous studies used different approaches to determine sowing dates and cultivar parameters. However, the timing of harvest has not yet been sufficiently analyzed. Here we propose an algorithm to determine location-specific maturity (or harvest) dates for applications in global modelling studies. Given a sowing date and the climatic conditions, the algorithm returns a suitable maturity date, based on crop physiological parameters and agronomic principles. We test the method on a global land area with a spatial resolution of 0.5 degrees against global reported datasets for major grain crops: winter-wheat, spring-wheat, rice, maize, sorghum and soybean. A single set of rules is able to largely reproduce the observed harvest dates of the six grain crops globally, with a mean absolute error of 19 (maize) to 45 (rice) days. In temperate regions, the temperature seasonality is the major driver of cropping calendars. In sub-tropical regions, crops are grown to match water availability. In the case of limiting growing seasons, the crop cycle is shortened or extended to avoid stressful periods. In the case of long-lasting favorable conditions the crop cycle is shorter than what the growing season would allow. We find that cropping periods can be largely defined by climate and crop physiological traits. The timing of the reproductive phase is shown to be a general criterion for selecting grain crops cultivars. This work will allow for dynamically representing adaptation to climate change by adjusting cultivars and represents a first step towards improved crop phenology simulations by global-scale crop models.