Biases in Historical SSTs Propagate to Key Metrics of Radiative Balance and Global Change

Recent work has shown that many observational products of sea surface temperature (SST) contain substantial biases in the early 20th century. Historical SST data is critical for estimating many key properties of the global climate system through its role in modulating global and regional temperature variability and change. The global and regional responses of the atmospheric and land surface are quantified using atmosphere-only GCMs (AGCMs) forced with historical SSTs. Here, we investigate how SST biases affect  atmospheric variability and trends using  an AGCM ensemble forced with SSTs from the infilledDynamically Consistent ENsemble of Temperature (DCENT-I), a recently published surface temperature product that better accounts for such biases. We compare this ensemble with an identically configured AGCM ensemble forced with ERSSTv5, a SST product with substantial early 20th-century biases. We find that DCENT-I SSTs produce more realistic terrestrial temperature trends. In addition, we explore the consequences of this updated SST dataset for estimates of climate sensitivity and pattern effects. Together, we demonstrate the critical need for accurate estimates of historical SST for understanding both the forced response and internal variability.