Stochastic Rainfall Synthesis Using Regionalized Model Parameters

Stochastic point precipitation models are typically used to generate rainfall at locations where observations are available to estimate the model parameters. A method for the synthesis of short time increment precipitation at ungauged points is presented using a simple stochastic model with spatially interpolated parameters. The motivation for this investigation came from a special practical problem in urban hydrology—the planning and rehabilitation of storm-water overflow tanks in municipal drainage systems. For this task, the application of urban rainfall-runoff models using long-term short-time increment precipitation data is becoming more and more important. The precipitation model, developed in an “alternating renewal model” framework, has nine parameters describing the external process structure. For the parameter regionalization, conventional and geostatistical methods are applied using easily available topographic and hydrometeorologic information. Cross validations for the various rain statistics and comparisons of rainfall-runoff simulations using observed and generated precipitation data, respectively, show the applicability of the method.