MODEL OF STOCHASTIC STRUCTURE OF DAILY PRECIPITATION OVER AN AREA.

Daily precipitation over an area is studied by modeling the stochastic structure of the time-area daily precipitation process. Daily precipitation at a point is envisioned as a continuous random variable that has been truncated at zero. The zero daily precipitation amounts are considered negative amounts of unknown quantity. The square roots of daily precipitation at a point for a selected study region approximate a sample from a truncated univariate normal distribution. The means and standard deviations of the normalized precipitation are periodic within the year. A method was developed for obtaining maximum likelihood estimates of daily means and standard deviations from the truncated samples. The periodic components of the means and standard deviations are described with Fourier series. The Fourier coefficients are related to position within the study region. Precipitation sequences were generated for two areas in the study region using the truncated multivariate normal distribution model. Parameters of the model were defined using the latitude and longitude of each station.