Land-based temperature records need to be screened for inconsistencies that can arise from changes in the measurement site, instruments, instrument shelters, or the way that monthly averages were computed, or due to the growth of cities around the sites (which would lead to a spurious warming trend due to a growing urban heat island). Effects of urbanization can be estimated and removed either by comparison of individual urban stations with nearby rural stations (as in Jones et aL, 1990), or by comparing results when cities with populations over 100,000 are included or excluded from the analysis (as done by Hansen and Lebedef£ 1987).