Development of daily, 3-day and 7-day ahead statistical streamflow forecasts

Climate variability plays an important role in effecting water availability, which could threaten reliability of water resources management in various river basins. In particular, frequency and magnitude of extreme events (e.g. hurricanes) have increased in Texas during recent decades, resulting in significant damages due to flooding. In order to improve decision making for flood preparedness and planning as well as managing daily reservoir operations and hydropower generation, forecasting of daily to weekly streamflow could play an important role. Thus, the primary objective of this study is to develop statistical forecasts for 1, 3 and 7 days’ lead time for multiple sites across different watersheds in Texas. To address this objective, two different statistical models (ARIMA and semi-parametric) were developed for three dam sites. In order to forecast daily to weekly streamflow during 1978 - 2018, only the antecedent streamflow was used as a predictor in the ARIMA modeling. In the semi-parametric method, besides antecedent streamflow, the forecasted precipitation was also used as the predictor to forecasts streamflow from 2011 – 2018. Our results indicated that ARIMA model outperformed semi-parametric approach in forecasting daily to weekly streamflow for all the three sites chosen in this study. The forecasting skill for 7-day and 3-day ahead forecasts were better than 1-day ahead forecasts for all the sites. Such simple statistical models could be used for decision making on 3-day to weekly time scales to better manage hydropower generation and minimize effects of multiday flooding event