Greenhouse statistics-time series analysis

Abstract

The relationship global mean temperature — atmospheric concentration of carbon dioxide is modelled by means of time series analysis as it is used in a non-experimental statistical context. The goal is to test the hypothesis that the global mean surface air temperature rises due to the rising atmospheric concentrations of greenhouse gases. The common climatological approach to confirm this hypothesis has not yet succeeded because of the overly ambitious model design and the statistically less efficient manner of information processing in interpretating the output of general circulation models. Earlier statistical attempts to detect the greenhouse signal in the temperature record failed partly because of inefficient modelling. Starting with some naive time series models we show that the enhanced greenhouse effect is plausible. Taking the long-term natural variability of the climate into account casts doubt on this claim but properly quantifying the size of the variability restores the significance of the greenhouse parameter. Extending the model to explain part of the shorter term variability by including the influence of the sun, volcanoes and El Nino the hypothesis is again but stronger confirmed. A battery of tests reveals that this model describes the observed temperature record (statistically) well. We also show that the outcomes are robust, i.e. insensitive to changes in the model. Although statistics cannot constitute a proof of the hypothesis, the results of this paper are strong enough to conclude that at least part of the recent high temperatures is, with high probability, caused by the increase in the atmospheric concentration of carbon dioxide.