The contribution of spike photosynthesis to wheat yield needs to be considered in process-based crop models

•Contribution of spikes photosynthesis to grain yield of wheat was 20%.•No current models explicitly consider contribution of spike photosynthesis to grain.•Simulation results of excluding & including spike contributions are investigated.•APSIM overestimated leaf contribution to compensate neglection of spike contribution.•Most current models may simulate wheat yield correctly partly for the wrong reasons. Process-based crop models are increasingly used to simulate Genotype by Environment by Management interactions (GxExM), and to evaluate importance of physiological traits to assist in breeding selections. This requires the model to correctly simulate the key physiological processes that determine grain yield. Although spike photosynthesis contributes to grain yield of wheat, it has not been explicitly simulated in most crop models. Here we present experimental data from China and Mexico, derived from estimates of spike contributions to grain yield of wheat, and compared them with previous results. We then used the APSIM model to investigate the consequences of excluding and including spike contributions on simulation results, against the data from China consisting of 4 irrigation and plant density treatments. Our results show that the contribution of spikes photosynthesis to grain yield ranged from 9.8% to 39.0% with an average of 20.1%, consistent with results from previous studies. However, despite the omission of the contribution from spike photosynthesis to grain yield, APSIM captured the dynamics of LAI and biomass and grain yield with acceptable accuracy. Across the treatments, the APSIM (version 7.9) explained >78% of the variation in yield (RMSE = 737 kg ha-1) and biomass (RMSE = 1438 kg ha-1) compared with the experimental data. This highlight the fact that APSIM performed well in simulating yield partly for the wrong reasons. Adding spike contribution to APSIM improved the simulations of biomass and yield, particularly under high yield level. More detailed analysis revealed that APSIM overestimated leaf contribution to compensate neglection of spike contribution, explained by the overestimation of leave biomass and underestimation of stem biomass. While APSIM captured the trend of changes in pre-anthesis remobilization contribution to grain yield, it also underestimated this contribution across treatments. Future improvements should include the inclusion of spike photosynthesis contribution with correct spike light interception and RUE, and imp