A new method (termed Principal Filter Analysis (PFA)) for analysing large time series of luminescence excitation-emission matrices (EEMs) is proposed, based on the idea of identifying 'filters' that detect time periods where interesting variations in the EEMs occur. A mathematical exposition of the technique is supplied, followed by a discusion of how it may be implemented in practice. The method is applied to EEMs taken from a stalagmite in Crag Cave, W. Ireland resulting in three distinct time periods of luminescence properties being identified.