Markov‐Weibull Model of Monthly Streamflow

Abstract

A conceptually simple, month‐to‐month lag‐1 Markov streamflow simulation model is developed and tested extensively. Streamflow in any month is represented probabilistically by a family of three‐parameter Weibull distributions conditioned on flow in the preceding month. The marginal distributions of the monthly and annual series of the simulated and historical data compare well statistically, and goodness‐of‐fit is excellent. Month‐to‐month correlation coefficients, auto‐correlation coefficients of the annual series, and the Hurst coefficient of the simulated and historical data compare well, as do comparative statistics of various drought‐high flow characteristics and flows of both sets of data through several hypothetical reservoir systems.