The purpose of this paper is to examine the causality between DUST, CO2 and temperature for the Vostok ice core data series [Vostok Data Series], dating from 420 000 years ago, and the EPICA C Dome data going back 800 000 years. In addition, the time-varying volatility and coefficient of variation in the CO2, dust and temperature is examined, as well as their dynamic correlations and interactions. We find a clear link between atmospheric C02 levels, dust and temperature, together with a bi-directional causality effects when applying both Granger Causality Tests (1969) and multi-directional Non-Linear analogues, i.e. Generalized Correlation. We apply both parametric and non-parametric statistical measures and testing. Linear interpolation with 100 years and 1000 years is applied to the three variables, in order to solve the problem of data points mismatch among them. The visualizations and descriptive statistics of the interpolated variables (using the two periods) show robustness in the results. The data analysis points out that variables are volatile, but their respective rolling mean and standard deviation remain stable. Additionally, 1000 years interpolated data suggests positive correlation between temperature and CO2, while dust is negatively correlated with both temperature and CO2.
The application of the non-parametric Generalized Measure of Correlation to our data sets, in a pairwise fashion suggested that CO2 better explains temperature than temperature does CO2, that temperature better explains dust than dust does temperature, and finally that CO2 better explains dust than vice -versa. The latter two pairs of relationships are negative.
The summary of the paper presents some avenues for further research, as well as some policy relevant suggestions.