Statistical models for identification of potential level shifts in environmental time series
When environmental time series are affected by a change of chemical analysis method or by a change of conducting laboratory statistical methods can be used to determine if the present series shows signs of a level shift. Such shift estimates are considerably more effective when there is a time period available where both chemical methods are used or when both laboratories conduct the analysis. If there are no such overlapping time periods the estimate’s variance will be big and, if there additionally is a trend in the series, can also be strongly biased. A reliable statistical analysis of trend in such series is then hardly possible.
The Article Statistical models for evaluating suspected artefacts in long-term environmental monitoring data by Authors Claudia von Brömssen, Jens Fölster, Martyn Futter och Kerstin McEwan can be read at SpringerLink.