Regional Homogenization of Surface Temperature Records Using Robust Statistical Methods

Published: September 12, 2013


Adam L. Pintar, Antonio M. Possolo, Nien F. Zhang


An algorithm is described and applied to estimate and remove spurious influences from the surface temperature record at a meteorological station, which may be due to changes in the location of the station or in its environment, or in the method used to make measurements, and which are unrelated to climate change, similar to [1]. The estimate of these influences is based on a comparison of non-parametric decompositions of the target series and series in a neighborhood about the target series. The uncertainty of the estimated spurious artifacts is determined with a non-parametric bootstrap method that accounts for temporal correlation structure beyond what is expected due to seasonal effects. Our computer-intensive parametric bootstrap procedure readily lends itself to parallelization, which makes the algorithm practicable for large collections of stations.
Proceedings Title: Regional Homogenization of Surface Temperature Records
Conference Dates: March 19-23, 2012
Conference Location: Anaheim, CA
Conference Title: 9th International Temperature Symposium
Pub Type: Conferences


Autocorrelation, Change-Point, LOESS, Homogenization, Temperature Series.
Created September 12, 2013, Updated February 19, 2017