DCENT-I: A Globally Infilled Extension of the Dynamically Consistent ENsemble of Temperature Dataset

A spatially infilled Dynamically Consistent Ensemble of surface Temperature (DCENT-I) has been created by infilling land-air and sea-surface temperatures from DCENT using ordinary kriging with anisotropic and heterogeneous kernels. By incorporating air-temperature anomalies over sea-ice areas, DCENT-I provides spatially complete monthly temperature fields at 5° resolution from 1850 to the present (currently the end of 2024) as a 200-member ensemble. Uncertainty estimates that account for the need to infill for missing observations are made using a Multivariate Gaussian Process, and these are consistent with estimates derived from masked climate model simulations. The use of anisotropic and heterogeneous kernels leads to a reconstruction of El Ni\~no variability whose spatial pattern and temporal variance is generally consistent throughout the record. As compared with taking the unfilled average, infilling increases the global mean surface temperature (GMST) warming estimate for 2005--2024 using a 1850--1900 baseline by 0.08 [0.05, 0.11]ØC (95% confidence interval), largely because of infilling in rapidly warming Arctic regions. Compared with HadCRUT5, GISTEMP v4, NOAA Global Temp v6, and Berkeley Earth, DCENT-I shows a steadier and slightly faster GMST warming trend, reflecting the bias-adjustments inherited from DCENT.