Detecting human influence on climate using neural networks based Granger causality

In this note we observe that a problem of linear approach to Granger causality testing between CO2 and global temperature is that such tests can have low power. The probability to reject the null hypothesis of non-causality when it is false is low. Regarding non-linear Granger causality, based on multi-layer feed-forward neural network, the analysis provides evidence of significant unidirectional Granger causality from CO2 to global temperature.