Earth Atmospheric Land Surface Temperature and Station Quality in the Contiguous United States

Abstract

A survey organized by A. Watts has thrown doubt on the usefulness of historic thermometer data in analyzing the record of global warming. That survey found that 70% of the USHCN temperature stations had potential temperature biases from 2°C to 5°C, large compared to the estimated global warming (1956 to 2005) of 0.64 ± 0.13°C. In the current paper we study this issue with two approaches. The first is a simple histogram study of temperature trends in groupings of stations based on Watt's survey of station quality. This approach suffers from uneven sampling of the United States; its main value is in illustrating aspects of the data that are counter-intuitive and surprising. The second approach is more statistically rigorous, and consists of a more detailed temperature reconstruction performed using the Berkeley Earth analysis method indicates that the difference in temperature change rate between Poor (quality groups 4, 5) and OK (quality groups 1, 2, 3) stations is not statistically significant at the 95% confidence level. The absence of a statistically significant difference indicates that these networks of stations can reliably discern temperature trends even when individual stations have nominally poor quality rankings. This result suggests that the estimates of systematic uncertainty were overly conservative and that changes in temperature can be deduced even with poorly rated sites.