Global mean surface temperature change is a key number which scientists and policy makers use to track the impact of human activity on the climate system. However the concept of 'mean' is more complicated is often realized when some of the data are missing. The most obvious method of calculating the mean gives poor results in such cases. We look at this statistical issues behind this problem and how it has been solved in the past. We also propose a simple solution which is usable with surface temperature observations from recent decades.