Stochastic Procedure for Generating Seasonal Flows

Author:
Publisher website
Google Scholar
Abstract
The generation of seasonal flows that can reproduce the historic basic statistics at more than one aggregation level is usually desirable for most reservoir simulation studies. The objective of this paper is to propose a stochastic procedure for generating seasonal flows at one or several sites. The proposed procedure combines a singular value decomposition (SVD) with a multivariate autoregressive of order one AR(1) model [SVD/MAR(1)]. The SVD transforms a standardized matrix of normalized flows into its standardized principal components, and the MAR(1) model is applied to the standardized principal components matrix. The SVD/MAR(1) procedure was compared with the existing SVD and MAR(1) approaches using the Great Lakes and Quebec rivers data. It can be concluded that the proposed SVD/MAR(1) procedure is applicable where the available flow data are limited and/or the number of considered sites is large. Furthermore, it can preserve adequately the basic statistics and some drought related properties of historic monthly and annual flows.

This publication has 10 references indexed in Scilit: