Identifying anthropogenic causes of the observed twentieth century surface temperature change : frequentist and Bayesian approach
Date
2003
Authors
Lee, Terry Chun Kit
Journal Title
Journal ISSN
Volume Title
Publisher
Abstract
It has been hypothesized that the recent warming trends in the climate system are caused by the increase in concentration of greenhouse gases and anthropogenic sulfate aerosols. The 'optimal fingerprinting' methodology is generally use to quantify the extent to which of forcing agents affect the climate. Recently, interest in applying a Bayesian methodology to the probĀlem has emerged. Compared to the classical 'optimal fingerprinting' approach, the Bayesian approach allows the incorporation of prior beliefs of climate change into the analysi and allows direct probability assessments on the hypotheses of climate change detection and attribution. These probability assessments are more quantitative compared to the classical significance level and p-value. We analyze the output from CGCMl , CGCM2 and HadCM2 climate model with the HadCRUTv observational dataset using both the 'optimal fingerprinting' and Bayesian apĀproaches , to identify which forcing agents are likely to have contributed to the temperature change in the twentieth century. The results from both approaches are very similar. With the Bayesian methodology, we find that the combined effect of greenhouse gases and sulfate aerosols may have contributed to the temperature change in the early twentieth century. Our analysis also suggests that the warming trends in the late twentieth century are highly unlikely to be due to natural climate variations alone. We also conduct robust analysis on the choice of prior used in Bayesian analysis to demonstrate that our results are insensitive to the choice of prior.