Changes in the circumstances behind in situ temperature measurements often lead to shifts in individual station records that can lead to over or under-estimates of the local and regional temperature trends. Since these shifts are comparable in magnitude to climate change signals, homogeneity "corrections" are necessary to make the records suitable for climate analysis. To quantify the effectiveness of surface temperature homogenization in the United States, a randomized perturbed ensemble of the pairwise homogenization algorithm was run against a suite of benchmark analogs to real monthly temperature data from the United States Cooperative Observer Program, which includes the subset of stations known as the United States Historical Climatology Network (USHCN). Results indicate that all randomized versions of the algorithm consistently produce homogenized data closer to the true climate signal in the presence of widespread systematic shifts in the data. When applied to the real-world observations, the randomized ensemble reinforces previous understanding that the two dominant sources of shifts in the U.S. temperature records are caused by changes to time of observation (spurious cooling in minimum and maximum) and conversion to electronic resistance thermometers (spurious cooling in maximum and warming in minimum). Trend bounds defined by the ensemble output indicate that maximum temperature trends are positive for the past 30, 50 and 100 years, and that these maximums contain pervasive negative shifts that cause the unhomogenized (raw) trends to fall below the lowest of the ensemble of homogenized trends. Moreover, because the residual impact of undetected/uncorrected shifts in the homogenized analogs is one-tailed when the imposed shifts have a positive or negative sign preference, it is likely that maximum temperature trends have been underestimated in the real-world homogenized temperature data from the USHCN. Trends for minimum temperature are also positive over the three periods, but the trend bounds produced by the 100 algorithm runs encompass trends from the unhomogenized data.

This content is only available via PDF.
You do not currently have access to this content.