Impact of climate change on maize yields in North-East China

Abstract

We assess climate change impacts on maize yields across China’s Northeast Farming Region (NFR) for two future periods (2020–2049 and 2050–2079) using a province-scale statistical crop model driven by projections from 41 CMIP6 global climate models (SSP5-8.5 scenario). This model represents climate influences on maize yield using a Gaussian function of June-August temperature and precipitation, and has previously shown skilful retrospective yield predictions for the region (1981–2016). We compare projected yields with results from 11 process-based crop models from the Global Gridded Crop Model Intercomparison (GGCMI) project which typically use heat accumulation to estimate the crop physiological stage and incorporate CO2 concentration effects in different ways. By 2020–2049, the statistical model projects reductions in mean yield, relative to 1985–2014, of 22% (12%–43%) for Liaoning; 16% (6%–34%) for Jilin and 6% (1%–23%) for Heilongjiang, with 5th–95th percentile ranges reflecting CMIP6 uncertainty. Because these projections exclude CO2 fertilisation, they generally indicate larger yield reductions than the GGCMI models. However, including a plausible CO2 fertilisation effect based on the GGCMI models (+0% to +12%) produces broad agreement with the process-based estimates. This gives a firmer indication of how current maize varieties might respond under future climates, and highlights the importance of assumptions about CO2 fertilisation. The relative simplicity of the statistical model enables us to explore sources of uncertainty in projected yields. We find that the choice of CMIP6 model contributes the greatest uncertainty, typically followed by CO2 fertilisation assumptions, while model parameter uncertainty is comparatively small. Given the importance of maize production in the NFR, this analysis highlights the need for climate adaptation activities, including crop breeding and water resource management. Comparative modelling approaches can contribute to such adaptation by enabling process-based and data-driven assessments of climate change impacts, and the improved assessment of uncertainties.