Multiple global temperature datasets are used in the assessment. Each of the datasets: 

• Provides global or near-global coverage in the modern period using statistical methods to fill gaps or using data from a reanalysis; 
• Has been adjusted to minimize the effect of known systematic errors associated with changes in how measurements were made over time or is otherwise free from known systematic errors; 
• Is described by a peer-reviewed paper; 
• Is publicly available and routinely updated. 

Nine datasets met the above criteria and were used in the calculation of global temperature. These include seven traditional datasets: Berkeley Earth, CMA-GMST, CMST v3, DCENT-I, GISTEMP v4, HadCRUT5, NOAAGlobalTemp v6, and and two reanalyses: ERA5 and JRA-3Q.

The method for calculating global mean near-surface temperature anomalies relative to an 1850–1900 baseline is based on the assessment of long-term change and its uncertainty in IPCC AR6 WG I. In 2021, IPCC AR6 WG I assessed change from 1850–1900 to other periods based on an average of four datasets – HadCRUT5, Berkeley Earth, NOAA Interim and Kadow et al. (2020) – which start in 1850 and are globally or near-globally complete in the modern period. 

To include shorter datasets, which can help to better understand recent temperature changes, in the present report, the estimate made by IPCC for the temperature change between 1850–1900 and 1981–2010 is combined with estimated changes between 1981–2010 and the current year from nine datasets – HadCRUT5, Berkeley Earth, NOAAGlobalTemp v6, GISTEMP v4, DCENT-I, CMST v3, CMA-GMST, ERA5 and JRA-3Q (for details of these see below) – to calculate anomalies for 2025 relative to 1850–1900. There is good agreement among the datasets on changes from 1981–2010 to the present, as this is a period with good observational coverage. The additional uncertainty from the spread of the datasets is combined with that of the IPCC’s estimate of the uncertainty in the change from 1850–1900 to 1981–2010. 

Global mean temperature anomalies were calculated relative to an 1850–1900 baseline using the following steps starting from the time series of global monthly mean temperatures for each dataset: 

• For each dataset, anomalies were calculated relative to the 1981–2010 average by subtracting the mean for the period 1981–2010 for each month separately.  
• An annual mean anomaly was calculated from the monthly mean anomalies.  
• The amount of 0.69 °C was added to each series, based on the estimated difference between 1850–1900 and 1981–2010, calculated using the method from IPCC AR6 WG I (see the caption for Figure 1.12 in that report).  
• The mean and standard deviation of the estimates were calculated.  
• The uncertainty in the IPCC estimate was combined with the standard deviation, assuming the two are independent and assuming the IPCC uncertainty range (0.54 °C–0.79 °C) is representative of a 90% confidence range (1.645 standard deviations).  

The number quoted in the key message for the most recent year was calculated in this way, with the value being the mean of the nine estimates. The estimated uncertainty is quoted as a 90% uncertainty range.