Process-based evaluation of Eastern Mediterranean heatwave development in the CMIP6 models

Understanding how well climate models represent the physical mechanisms leading to heatwaves is critical for improving future projections and early warning capabilities. Here, we present a process-based evaluation of 11 CMIP6 models for simulating the dynamic and thermodynamic precursors of summer heatwaves over the Eastern Mediterranean and the Middle East, using ERA5 reanalysis as a benchmark. We combine composite anomaly diagnostics, a feature-based spatial evaluation framework (structure–amplitude–location; SAL), and a physically grounded decomposition of the temperature-tendency budget to assess model realism across multiple atmospheric levels. While CMIP6 models broadly reproduce the spatial structure of heatwaves, they systematically underestimate the amplitude and delay the onset of pre-event warming, upper-level ridging, and dynamical heating. Horizontal temperature advection, particularly the advection of climatological temperature gradients by anomalous winds, emerges as the dominant driver of heat accumulation in ERA5. Yet, its buildup is consistently delayed in the models. Inter-model differences in heatwave intensity are strongly linked to biases in lower-tropospheric geopotential height over Türkiye, which serves as an integrated diagnostic of synoptic ridge strength and horizontal advection efficiency. In addition, CMIP6 models fail to capture the observed transition from subsidence to ascent over India in the days preceding heatwave onset, indicating a systematic deficiency in representing tropical-extratropical teleconnections associated with the South Asian monsoon. Together, these results highlight the value of process-oriented model evaluation for diagnosing the origins of heatwave biases and for improving sub-seasonal predictability and future climate projections in this highly vulnerable region, while also offering a transferable framework for evaluating heatwave dynamics and predictability in other regions worldwide.