Effect of missing data on estimates of near-surface temperature change since 1900

It has been suggested that the apparent warming of the earth's surface during the twentieth century may be biased by large changes in the coverage of surface temperature measurements since 1900. This issue is investigated using climate model simulations. By imposing observed coverage changes on simulated surface temperatures. Estimates are obtained of twentieth-century temperature change for both full global coverage and for actual historical coverage. In 10 out of 16 simulations including human climate perturbations, the temperature change from the globally complete model output is significantly larger than that derived from the historically masked model output. The remaining six simulations show no significant difference between complete and masked model output. Thus, these results do not support the hypothesis that the increase in the earth's surface temperature has been overestimated because of incomplete observational data. Rather, if the simulations analyzed are realistic, the true temperature increase over the last century is slightly larger than that estimated from available observations. Eight simulations of natural internal climate variability, which omit human climate perturbations, were analyzed. In none of these simulations does the temperature change during 100 yr - whether obtained from globally complete or masked model output - come close to the observed twentieth-century temperature increase.