Item

Abstract

Crop models are used within the GGCMI-AgMIP framework to assess climate impacts on crop yields, but their ability to reproduce observed yield responses to regional climatic variability and extremes remains insufficiently evaluated. To help fill this evaluation gap, we use the North China Plain (NCP), China, as a regional case study and combine prefecture-level winter wheat yield and climate records with GGCMI Phase 2 simulations. The analysis focuses on three growing-season climatic drivers: mean temperature, precipitation, and extreme heat days (EHD), defined as growing-season days with daily maximum temperature exceeding 34 °C. Rather than evaluating yield levels alone, we compare effective yield–climate response types for these climatic drivers, using observed responses as a benchmark for diagnosing model behaviour. We use targeted JULES-crop experiments to test whether a heat-stress phenology modification can reduce the diagnosed bias in simulated extreme-heat responses. Observations show a widespread significant negative linear yield response to EHD in the central NCP, whereas responses to growing-season mean temperature and precipitation are more spatially heterogeneous. Most GGCMI models fail to reproduce this observed response type to extreme heat, suggesting a systematic weakness in the representation of heat-stress impacts. The original JULES-crop configuration fails to capture significant negative linear yield-EHD relationships at the observed heat-sensitive sites. Incorporating heat-stress effects into phenological development and senescence improves this functional response in JULES-crop, increasing the number of sites with a consistent significant negative linear yield-EHD relationship from 0 to 7 out of 16 observed significant sites. Although this targeted modification does not eliminate all model-observation discrepancies, it demonstrates that phenological sensitivity to extreme heat is a plausible process-level source of model bias. This study extends GGCMI Phase 2 evaluation from broad-scale yield-level comparison towards regional, response-based diagnosis, and identifies heat-stress phenology as a practical pathway for improving winter wheat simulations under intensifying heat extremes.