A disaggregation model for storm hyetographs

An event based rainfall generation model was developed by combining a new disaggregation technique with an appropriate rainfall model. The model disaggregates the total depth of an event with known duration into incremental depths with arbitrary time step less than the duration. The disaggregation technique is charactersed by simplicity and parsimony of parameters; it assumes a random shape of the hyetograph and it is compatible with various rainfall models. The technique is based on the assumption that incremental depths are gamma distributed. With this assumption it was found that a simple disaggregation method consisting of two steps, where the first step is a sequential model and the second is an adjustment procedure, can give good approximations of the important statistics of interest. Furthermore, under some ideal conditions the disaggregation method was shown to be exact in a strict sense, i.e. it preserves the complete distribution of the variables. The disaggregation technique was combined with three alternative rainfall models (a scaling model, a Markovian in continuous time and a Markovian in discrete time). The results of the model application and testing at these three cases indicated good approximation of the important statistics of incremental rainfall depths (first, second and third order marginal moments, marginal distributions, and joint second order statistics).