Stochastic Modeling of Irrigation Requirements

Abstract

The objective of this paper is to study the stochastic structure of weekly irrigation requirement of a crop. The irrigation requirement time series is assumed to be represented by an additive model with trend, periodic and stochastic as its components. Each component is identified and, if found, removed from the original series. The turning point test and Kendall's rank correlation test are applied for detecting the trend. In the analysis of series, the correlogram technique is used to detect the periodicity, which is then analyzed by Fourier series method. Harmonic analysis is done for identifying the number of significant harmonics. The series is then tested for stationarity and the dependent part of the stochastic component is found to be well expressed by the second order autoregressive model. Therefore, as a result, the developed model superposes a periodic‐deterministic process and a stochastic component. The adequacy of fit is judged by the insignificant correlation and normal distribution of the obtained residuals. The developed periodic‐stochastic model may be used for representing the time based structure of the irrigation requirement time series of a paddy crop.