A globally complete, 200-member ensemble of monthly surface temperature (SST) and land surface air temperature (LSAT) anomalies at 5°x5° resolution, spanning the period from 1850 to present. It is derived from version 2.0 of the unfilled Dynamically Consistent ENsemble of Temperature at the Earth surface (DCENT) dataset and represents 2-metre air temperature over land and sea surface temperature across the marine domain. In addition to the ensemble realisations, summary statistics including the ensemble mean, standard deviation, and the 5th and 95th percentiles, are provided. All temperature values are expressed as anomalies relative to the 1982-2014 climatology, which is also supplied to enable reconstruction of monthly absolute temperatures. A diagnostics file is provided, which contains the land and sea weights and masks, as well as α which quantifies the local kriging variance as a proportion of the prior variance to indicate data reliability. The files contain SST and LSAT separately and combined in the tas variable. DCENT-I combines DCENT’s state‐of‐the‐art bias/homogenisation corrections with a spatially adaptive interpolation method to produce globally complete monthly temperature fields with estimates of uncertainty. The dataset was generated using GloMarGridding v1.0.0-rc.2 (https://github.com/NOCSurfaceProcesses/GloMarGridding) by scientists at the University of Southampton, National Oceanography Centre (NOC), Woods Hole Oceanographic Institution and Harvard University. NOC was supported by the NERC-funded National Capability Programme ‘Atlantic Climate and Environment Strategic Science’ (AtlantiS), grant reference NE/Y005589/1.