Probabilistic Streamflow Forecasts Based on Hydrologic Persistence in Central Texas

In many cases, streamflow persistence (month-to-month or season-to-season correlation) can be used in place of climate forecasts to provide useful forecast information to water managers. In this study, an ordinal polytomous logistic regression model is proposed to generate tercile probability stream flow forecasts (i.e., probability of low, medium, and high categories) based on persistence for the Lower Colorado River system in central Texas. Forecast performance is evaluated by cross-validation using the Brier skill score (BSS) and the Ranked probability skill score (RPSS). The results show that stream flow persistence can provide significant forecast skill during the winter and spring seasons, when water allocation decisions are being made for the coming summer growing season.