The difference between land surface (LST) and near surface air temperature (Tair) provides information about Earth’s surface energy balance and can give insight into the status of vegetation (e.g., if the plants are stressed). Therefore, LST and Tair are essential components of Earth System Models (ESMs), and inaccuracies in LST-Tair can skew other aspects of model projections or be indicative of existing errors. For example, inaccurate simulation of LST-Tair can suggest errors in the model representation of vegetation response to climate variability (e.g., droughts), thus skewing their associated feedbacks on climate and weather patterns.

In this study we explore the representation of LST-Tair across CMIP6 ESMs using measurements from satellite and observationally based reanalysis data products as benchmarks. Using an emergent constraint, we discover a systematic bias in the difference between these temperature datasets across the ensemble of CMIP6 ESMs and investigate its causes. We examine the implications for biogeochemical cycling in the ESMs due to the misrepresentation of LST-Tair, and examine potential sources of ESM improvement.