HadCRUT5 has been processed on Met Office computing facilities, using HadSST4 and CRUTEM5 inputs.

CRUTEM5 station data are gridded onto a regular latitude-longitude grid using the HadCRUT4 ensemble method. Resulting land air temperature anomaly grids are merged with sea-surface temperature anomalies from the HadSST4 dataset though weighted averaging based on grid cell areal land fractions to produce the non-infilled HadCRUT5 grids.

HadCRUT5 analysis grids are computed using a Gaussian process based statistical method to provide improved estimates of surface temperature anomaly fields, extending the data coverage into regions for which the available observations are informative. Analysis uncertainties are presented through an ensemble method. The Gaussian process based method is applied separately for land and marine temperature grids. Global grids are produced using a land-sea area fraction based weighting, in which sea ice regions (as defined by HadISST2 sea-ice concentrations) are treated as land.

Regional time series are derived as grid box area weighted averages, with accompanying uncertainties. Global time series are defined as the average of northern and southern hemisphere series. Remaining uncertainty from unrepresented regions is estimated from sub-sampling experiments using the ERA5 reanalysis.