Published: Aug. 17, 2018 By

Bracken, Cameron WÌý1Ìý;ÌýRajagopalan, BalajiÌý2

1ÌýUniversity of Colorado at Boulder
2ÌýUniversity of Colorado at Boulder

Current operational forecasting methods in the Upper Colorado River Basin (UCRB) use antecedent flow and soil moisture conditions, snowpack and climate indices (Brandon, 2005). Experimental methods include large scale climate drivers and obtain similar results (Bracken et al., 2010; Regonda et al., 2006; Grantz et al., 2005). These methods only provide reasonable results in the current season. Longer term information, especially for large reservoirs, is useful to water managers in terms of planning. The annual flow at the outlet of the UCRB, Lees Ferry, is very difficult to model with traditional time series methods (AR, KNN, MC, etc.) due to its low autocorrelation. Hidden markov models (hmm) provide a flexible and attractive alternative to more traditional approaches. We use HMM's to generate forecast distributions for recent years (1980-2010). These methods give insight into the driving patterns behind UCRB flow and show increased skill over climatology and AR models in certain cases.

Bracken, C., B. Rajagopalan, and J. Prairie (2010), A multisite seasonal ensemble streamflow forecasting technique, Water Resour. Res, 46(3), W03,532.

Brandon, D. G. (2004), Using nwsrfs esp for making early outlooks of seasonal runoff volumes into lake powell, Am. Meteorol. Soc., pp. 1–20.

Grantz, K., B. Rajagopalan, M. Clark, and E. Zagona (2005), A technique for incorporating largescale climate information in basin-scale ensemble streamflow forecasts, Water Resour. Res.

Regonda, S. K., B. Rajagopalan, M. Clark, and E. Zagona (2006), A multimodel ensemble forecast framework: Application to spring seasonal flows in the gunnison river basin, Water Resour. Res, 42.