Spatiotemporal co-optimization of agricultural management practices towards climate-smart crop production

Co-optimization of multiple management practices may facilitate climate-smart agriculture, but is challenged by complex climate-crop-soil management interconnections across space and over time. Here we develop a hybrid approach combining agricultural system modelling, machine learning and life cycle assessment to spatiotemporally co-optimize fertilizer application, irrigation and residue management to achieve yield potential of wheat and maize and minimize greenhouse gas emissions in the North China Plain. We found that the optimal fertilizer application rate and irrigation for the historical period (1995-2014) are lower than local farmers' practices as well as trial-derived recommendations. With the optimized practices, the projected annual requirement of fertilizer, irrigation water and residue inputs across the North China Plain in the period 2051-2070 is reduced by 16% (14-21%) (mean with 95% confidence interval), 19% (7-32%) and 20% (16-26%), respectively, compared with the current supposed optimal management in the historical reference period, with substantial greenhouse gas emission reductions. We demonstrate the potential of spatiotemporal co-optimization of multiple management practices and present digital mapping of management practices as a benchmark for site-specific management across the region.