An improved SST reconstruction for the 1854–1997 period is developed. Compared to the version 1 analysis, the new analysis resolves more variance in the western tropical Pacific, tropical Atlantic, and Indian Oceans. It also uses sea ice concentrations to improve high-latitude SST analysis and a modified historical bias correction for the 1939–41 period. Additionally, the new analysis includes improved error estimates. Analysis uncertainty is largest in the nineteenth century and during the two world wars due to sparse sampling. The near-global average SST in the new analysis is consistent with version 1. The 95% confidence uncertainty for the near-global average is 0.4°C or more in the nineteenth century, near 0.2°C for the first half of the twentieth century, and 0.1°C or less after 1950. The ERSST.v1 analysis is produced from the latest version of the Comprehensive Ocean Atmosphere Data Set (COADS). It uses monthly and 2° spatial superobservations, which are defined as individual observations averaged onto a 2° grid. The analysis separates low- and high-frequency components, which are then added together to form the total SST anomaly. The low-frequency analysis is computed by smoothing and filtering the QC superobservations. The high-frequency analysis uses a 2° version of the 1982–2000 REA analysis to define anomaly increment modes. The ERSST.v1 analysis has some weaknesses, including weaker variance in the western tropical Pacific, Atlantic, and Indian Oceans. The ERSST.v2 analysis improves upon ERSST.v1 by using a better high-frequency analysis, incorporating sea ice information, adjusting the historical bias correction, and improving error analysis. The ERSST.v2 analysis uses anomaly modes instead of increment modes, which allows for better resolution of tropical variations. It also includes sea ice concentration information to improve high-latitude SST analysis. The ERSST.v2 analysis also includes improved error statistics and a spatial gridded field of the monthly error. The ERSST.v2 analysis is more accurate than ERSST.v1, especially in the western tropical Pacific, Atlantic, and Indian Oceans. The ERSST.v2 analysis also has a lower standard deviation in the tropical regions compared to ERSST.v1. The ERSST.v2 analysis is compared to other analyses, including HadISST. The ERSST.v2 analysis shows similar results to HadISST in most regions, but differs in some areas. The ERSST.v2 analysis has a lower standard deviation in the tropical regions compared to ERSST.v1. The ERSST.v2 analysis also has a lower standard deviation in the Southern Hemisphere compared to ERSST.v1. The ERSST.v2 analysis is more accurate than ERSST.v1 in the western tropical Pacific, Atlantic,An improved SST reconstruction for the 1854–1997 period is developed. Compared to the version 1 analysis, the new analysis resolves more variance in the western tropical Pacific, tropical Atlantic, and Indian Oceans. It also uses sea ice concentrations to improve high-latitude SST analysis and a modified historical bias correction for the 1939–41 period. Additionally, the new analysis includes improved error estimates. Analysis uncertainty is largest in the nineteenth century and during the two world wars due to sparse sampling. The near-global average SST in the new analysis is consistent with version 1. The 95% confidence uncertainty for the near-global average is 0.4°C or more in the nineteenth century, near 0.2°C for the first half of the twentieth century, and 0.1°C or less after 1950. The ERSST.v1 analysis is produced from the latest version of the Comprehensive Ocean Atmosphere Data Set (COADS). It uses monthly and 2° spatial superobservations, which are defined as individual observations averaged onto a 2° grid. The analysis separates low- and high-frequency components, which are then added together to form the total SST anomaly. The low-frequency analysis is computed by smoothing and filtering the QC superobservations. The high-frequency analysis uses a 2° version of the 1982–2000 REA analysis to define anomaly increment modes. The ERSST.v1 analysis has some weaknesses, including weaker variance in the western tropical Pacific, Atlantic, and Indian Oceans. The ERSST.v2 analysis improves upon ERSST.v1 by using a better high-frequency analysis, incorporating sea ice information, adjusting the historical bias correction, and improving error analysis. The ERSST.v2 analysis uses anomaly modes instead of increment modes, which allows for better resolution of tropical variations. It also includes sea ice concentration information to improve high-latitude SST analysis. The ERSST.v2 analysis also includes improved error statistics and a spatial gridded field of the monthly error. The ERSST.v2 analysis is more accurate than ERSST.v1, especially in the western tropical Pacific, Atlantic, and Indian Oceans. The ERSST.v2 analysis also has a lower standard deviation in the tropical regions compared to ERSST.v1. The ERSST.v2 analysis is compared to other analyses, including HadISST. The ERSST.v2 analysis shows similar results to HadISST in most regions, but differs in some areas. The ERSST.v2 analysis has a lower standard deviation in the tropical regions compared to ERSST.v1. The ERSST.v2 analysis also has a lower standard deviation in the Southern Hemisphere compared to ERSST.v1. The ERSST.v2 analysis is more accurate than ERSST.v1 in the western tropical Pacific, Atlantic,