The authors acknowledge that bit-for-bit reproducibility is generally not feasible due to hardware and software differences across HPC systems. While their replicability test helps identify statistical consistency, it may not catch spatially varying biases that could still affect results.
The model's Southern Ocean warm bias and underestimated Antarctic sea ice are persistent issues that could significantly impact simulations of global climate, ocean circulation, and sea level rise.
While EC-Earth3 shows an improved QBO compared to its predecessor, the model still lacks a stratospheric semiannual oscillation above 5 hPa, indicating incomplete representation of stratospheric dynamics.
The model's overestimation of Arctic sea ice volume and thickness, alongside regional discrepancies in sea ice concentration, highlight limitations in simulating Arctic sea ice processes and their feedbacks.
Simplified Lake Treatment
The lack of an explicit lake model and reliance on ocean model extrapolation can lead to inaccurate representation of lake temperatures and ice cover, potentially impacting regional climate simulations.
Limited Carbon Cycle Analysis
The carbon cycle component analysis is limited to a single historical experiment member and does not include climate feedback, hindering a full assessment of the model's coupled carbon cycle performance.