AbstractThe East Asian summer monsoon (EASM) is a dominant driver of East Asian climate, with variations in its strength potentially impacting the livelihoods of millions of people. Understanding, predicting, and assessing uncertainties in these variations are therefore important area of research. Here, we present a study of the projected twenty-first century changes in the EASM using a ‘perturbed parameter ensemble’ (PPE) of HadGEM3-GC3.05 coupled climate models, which samples uncertainties arising from differences in model parameter values. We show that the performance of PPE members for leading order EASM metrics is comparable to CMIP5 and CMIP6 models in many respects. But the PPE also exposes model biases which exist for almost all parameter combinations. These ‘structural’ biases are found mainly to affect metrics for the low-level circulation. We also show that future changes in regional circulation and precipitation are projected consistently across the PPE members. A more detailed moisture budget analysis of the precipitation changes in a region covering the Yangtze River valley shows that the spread of these changes is mainly due to spread in dynamic responses. We also perform parameter sensitivity analyses and find that a parameter controlling the amplitude of deep-level entrainment is the main driver of spread in the PPE’s representation of the EASM circulation. Finally, we discuss how the information provided by the PPE may be used in practice, considering the plausibility of the models, and giving examples of ways to sub-select ensemble members to capture the diversity in the moisture budget changes.