Normalization of Rainfall Across Different Time Stepsby M. F. Hutchinson, Texas Agricultural Experiment, Station, Temple, United States,
C. W. Richardson, Texas Agricultural Experiment, Station, Temple, United States,
P. T. Dyke, Texas Agricultural Experiment, Station, Temple, United States,
Abstract: A parameter efficient procedure for normalizing the distribution of rainfall totals is examined with the particular aim of assessing its utility for space-time rainfall simulation. It involves fitting, by maximum likelihood, a normal distribution truncated at a small positive observation threshold, to a fractional power of the observed rainfall totals. By allowing the power to vary gradually with geographic location, observed rainfall distributions at monthly and daily time steps can be well matched using just two additional parameters. One of these is determined by the probability of exceeding the observation threshold while the other can be related to the monthly mean rainfall. Both of these parameters can be spatially interpolated from existing rainfall networks. By inserting a first order autoregressive parameter, the daily model offers a simple and accurate alternative to existing statistical point daily rainfall models. Incorporating the model into a space-time framework requires further investigation of an observed disparity between occurrence-based and intensity-based correlations.
Subject Headings: Rainfall | Parameters (statistics) | Autoregressive models | Model accuracy | Power transmission | Probability | Space frames | Regression analysis
Services: Buy this book/Buy this article
Return to search