Evaluating Surface Heat Feedbacks: Insights from CMIP Models

Surface feedbacks are critical to understanding global temperature patterns, as they directly regulate the exchange of energy at the ocean-atmosphere boundary. While surface feedbacks have been studied using methods like the Gregory regression, Climate Feedback-Response Analysis Method (CFRAM) and Partial Radiative Perturbation (PRP) approaches, their role in shaping climate responses remains less explored through analytical frameworks. Here, we assess surface feedbacks across 48 models from the Coupled Model Intercomparison Project Phases 5 and 6 (CMIP5/6) using a novel analytical formulation. This approach focuses on net surface heat flux feedbacks (latent and longwave), with observational constraints to improve robustness.
Our results indicate that the globally averaged surface feedback is dominated by the latent heat component for both the multi-model mean and the diversity. This is because of strong compensation between upward and downward longwave feedbacks. Observational constraints further reveal that CMIP models tend to overestimate the negative latent heat feedback in the tropics, attributed to an exaggerated air-sea temperature gradient . This bias may contribute to a ~10% underestimation of the global warming signal in these models.
By leveraging this novel analytical approach, we provide insights into surface feedback processes that are obscured in traditional TOA-based analyses. While our findings for net longwave feedbacks align closely with the standard regression approach, the analytical method reveals systematically higher negative latent feedback values. This divergence points to the need of further investigating the roles of short-term feedback processes and dynamical contributions, paving the way for reconciling analytical and regression-based methodologies.