Evaluation of Rainfall Characteristics Using Entropy

Abstract

Precipitation as a key factor for rainwater utilization has been conventionally analyzed paying attention to its quantity. Even though the total amount of annual precipitation is the same, one district might have the concentrated rainfall during a short period, and another district might have rainfall constantly throughout a year. For successful use of rainwater, therefore, it is an essential need to investigate time-varying precipitation from both quantity and uncertainty aspects of rainfall events. Use of the concept of entropy in the realm of information theory is made to evaluate rainfall patterns in terms of rainfall uncertainty. An illustrative entropy evaluation is made to know spatial distributions of the rainfall entropy (with a trade-off relation to the uncertainty) over the whole country of Japan, using the 22-year daily precipitation data recorded at 1,105 observatories of the AMeDAS. The results show that the entropy indeed provides a means of comparatively evaluating local rainfall characteristics on a probability basis, possibly being an alternative measure of regional water resources potential.