Economic and climate time series exhibit many commonalities. Both are subject to non-stationarities in the form of evolving stochastic trends and sudden distributional shifts, with incomplete knowledge of the processes generating the data (DGP). Consequently, the well-developed machinery for modelling economic time series can be fruitfully applied to climate time series. We discuss the model selection methodology for locating an unknown DGP nested within a large set of possible explanations, including dynamics, outliers, shifts, and non-linearities, using Autometrics, a variant of machine learning capable of implementing indicator saturation estimators. After a brief excursion into climate science, we illustrate the approach by investigating the causal role of CO2 in Ice Ages and the UK’s highly non-stationary annual CO2 emissions over the last 150 years, and draw some policy implications facing a claimed net zero target by 2050 in the absence of any clear strategy for achieving it.