This dataset contains global mean annual reconstructed surface temperature anomaly time series for merged land, air and sea-surface. This dataset was produced in 2004 for the period: 1880-2009. The reconstruction is separated into separate low and high-frequency components, which are added for the total reconstruction. To create the global mean the Northern and Southern Hemisphere averages are computed separately and then they are then averaged together (to stop the better sampled Northern Hemisphere from dominating the Global average). This dataset improved analysis and error estimation methods to analyze both sea surface temperature and land surface temperature anomalies, and to then merge them to form a global analysis.