Modeling monthly hydrologic persistence

Abstract

This paper empirically studies the autocorrelations and Hurst parameter properties for monthly streamflow for a set of U.S. Geological Survey gaging stations in the eastern United States. The analysis is motivated by a desire to develop operational criteria to select persistence measures and generating models. The work is part of a larger effort to evaluate the trade‐offs between complex and simple data generators. Markov generators are simple in comparison to the more complex self‐similar methods. Other papers in this effort are those by Chi et al. (1973) and Jettmar and Young (1975). Monthly streamflow data are found to have Hurst statistics similar to annual data. Least squares is presented as a possible method to estimate persistence parameters and select generating models. Empiric autocorrelations are treated as data in a fit of theoretic covariance functions by minimum squared error. The Hurst and least squares estimators are nearly the same. The lag 1 correlation is much smaller than least squares Markov estimators. A step‐by‐step procedure is suggested for implementation of least squares persistence analysis.