Stochastic Disaggregation Model For Multi-scales Rainfall Simulation

Participation is an importantcomponentofthe hydrologic cycle. The multi-scalesgeneration of rainfall series have great potential use in providing the increasing resolutiondata support for various hydrological research and design and management of water resourcesystem.Adopting the simple disaggregationmethodology provided by Koutsoyannis, multipletime scales rainfall disaggregation models are formed by coupling different stochastic rainfallmodels set up in annual, monthly and daily scales separately using accurate adjustingprocedures. Coupling AR(2) model for annual rainfall total with PAR(1) model for monthlyrainfall total results in an multiscale preservation of stochastic structure of monthly rainfallprocess. Based on Markov chain for daily rainfall occurrence and gamma distribution fordaily rainfall depth, daily rainfall model coupled with monthly rainfall distribution modelresults in the daily rainfall process preserving statistics in both scales. The methods applied to20sites in Wei River basin, the main conclusions are drawn as follows:(1) The annual and monthly scales coupleddisaggregation model canpreservestheannualrainfall statistics including the average, the variance and the coefficient of skewness, as wellas the autocorrelation coefficient, in addition to reproducing the monthly main statistics ofeach month. Although the discrepancy appears among the skewness of several individualmonths, the summer months with high rainfall depth establish a well agreement.(2) The monthly and daily scales coupleddisaggregation model generated daily rainfallseries reproduces the statistics of interest in both monthly and daily scales acceptably.(3) The adjusting procedures applying to rainfall variable with the marginal gammadistribution allocate the departure error of the sum of lower-level variables within a periodfrom the corresponding higher-level variable. They indicate thatthe model canpreservecertainstatistics and distribution of lower-level variables.(4) Simple disaggregation method avoid the drawbacks of the excessivenumber ofparameters, variance and covariance structures and difficulty in parameter estimationprocedures with traditional disaggregation model. There is considerableflexibility in itsmethodology, which will easily be applied to rainfall simulation of gaugesacross the WeiRiver basin. The performance is generalsatisfactory.