Methods used to quantify temperature changes of the earth must be asse
ssed relative to an appropriate definition of global mean temperature.
In this paper, global mean temperature is defined and the adequacy of
using weighted average anomalies to estimate a global mean anomaly is
investigated, with special emphasis on the effects of gridding and da
ta reuse, Distance weighting is shown to be effective for estimating g
ridpoint temperature anomalies but it is not optimal for estimating re
gional or global means. Gridding is shown to be more effective than us
ing irregularly spaced individual temperature anomalies in the estimat
ion of regional or global means. The reuse of station data in the calc
ulation of averages for neighboring grids is not recommended.