The net warming effect of clouds on global surface temperature may be weakening or even disappearing

• A pioneering attempt to quantify the effect of clouds on surface temperature. • A phased machine learning scheme combined with physical mechanisms. • Mapping and understanding the spatio-temporal pattern of the cloud effect. • Clouds exhibit a weak net warming effect on global annual surface temperature. Climate change is significantly influenced by both clouds and Earth’s surface temperature (EST). While numerous studies have investigated clouds and EST separately, the extent of clouds’ impact on EST remains unclear. Based on the inspiration and limitation of cloud radiative effect (CRE), this study provides a pioneering attempt to propose a novel indicator, cloud radiative effect on surface temperature (CREST), aiming to quantify how clouds affect EST globally while also analyzing the physical mechanism. Using reanalysis and remotely sensed data, a phased machine learning scheme in combination of surface energy balance theory is proposed to estimate EST under all-sky and hypothetical clear-sky conditions in stages, thereby estimating the newly defined CREST by subtracting the hypothetical clear-sky EST from the all-sky EST. The inter-annual experiments reveal the significant spatial heterogeneity in CREST across land, ocean, and ice/snow regions. As a global offset of the heterogeneity, clouds exhibit a net warming effect on global surface temperature on an annual scale (e.g., 0.26 K in 1981), despite their ability to block sunlight. However, the net warming effect has gradually weakened to nearly zero over the past four decades (e.g., only 0.06 K in 2021), and it’s even possible to transform into a cooling effect, which might be good news for mitigating the global warming.