A Method to Adjust Long-Term Temperature Extreme Series for Nonclimatic Inhomogeneities

Abstract

A method to homogenize nonclimatic discontinuities in temperature extreme exceedence series is developed and evaluated. The method is based on a set of complementary tests with the application of an individual test depending on the availability of an adequate network of nearby homogeneous reference stations and the presence of significant trends in the resulting difference or original exceedence series. Given a suitable set of neighboring reference stations, a comparison of the differences in exceedences between the inhomogeneous station and neighboring sites is made for the periods before and after the documented discontinuity. In the absence of one or more reference stations, the exceedences at the inhomogeneous station are compared before and after the nonclimatic change. A method by which nonstationary series are detrended and subsequently evaluated is also presented. When tested using homogenized data series into which an artificial discontinuity of known magnitude was introduced, as many as 80% of the ±1°F discontinuities were detected by the difference series approach. The performance of the single-station exceedence series test was less accurate. Although in a few cases, less than 40% of the ±1°F discontinuities were detected, between 60% and 76% of the ±2°F discontinuities were identified. Using both tests, the probability of falsely detecting a discontinuity (i.e., identifying an inhomogeneity when none existed) was 5%. Provided both methods deemed a documented inhomogeneity significant, the magnitude of the adjustments imposed by both methods was similar.