Abstract
Monthly rainfall variability in Japan is quantitatively assessed using Shannon's informational entropy (SIE) theory. Different two aspects of rainfall variability, temporal variability associated with the over-a-year rainfall apportionment and intensity variability, are investigated in terms of information content or entropy contained in probabilistic variations of the rainfall events. The apportionment entropy (AE) and the intensity entropy (IE) are thus defined for measuring the respective variabilities by use of monthly rainfall series that can be obtained from daily rainfall data observed at a network of 1,038 raingauges. Based on AE- and IE-values obtained at the raingauges, their spatial distributions are delineated on a nationwide map of Japan. Comparisons with the other maps available indicate that IE is distributed over the whole country as reproducing the nationwide distribution of annual precipitation, and the spatial distribution of AE has a clear match with that of CV (the coefficient of variation) even though these two are negatively correlated. It is thus concluded that monthly rainfall variability could satisfactorily be measured in terms of IE and AE, and therefore coupled IE and AE can become a comprehensive measure of the uncertainty of rainfall events or of the throughout-the-year potential availability of water resources on a monthly base.
