Estimation of interaction between climatic processes: Effect of sparse sample of analyzed data series

It is shown that the approach based on the Granger causality may lead to erroneous conclusions about bidirectional coupling (BC) in the case of unidirectional coupling (UC) and quite a sparse sample of the data series analyzed. This effect was revealed in an analysis of coupling between variations in the solar irradiance and global surface temperature. We present a statistical test to confirm or reject speculations about the character (unidirectional or bidirectional) of coupling. The corresponding analysis of coupling between phenomena of El Niño and the Indian Monsoon confirmed the earlier conclusions about their mutual influence.