2)  Decadal-scale variance and persistence patterns exist.  That is, there appear to be cycles of high and low streamflow variability and cycles of high, low, and anti-persistence in annual streamflows.  These patterns are most prevalent in the Sprague streamflow data because the unique geology and groundwater influence in the Williamson smoothes much of the seasonal snowmelt signal as well as introduces lags in streamflow response to climate forcing.  These cycles of variability and persistence have significant ramifications in water management.

3)  There has been a forward shift in the timing of peak seasonal streamflow for both the Sprague and Williamson rivers since streamflow records began in the 1930s (Figure 2).  There has been a pronounced shift from April clearly being the month of peak streamflow volume to March becoming as significant or more significant than April.  This finding suggests that the March streamflow volume should be included in the forecast period, with water management decisions being shifted forward accordingly, and that the 1 March forecast is now equally as important as the 1 April forecast.

4)  The TNI was identified as being strongly associated with seasonal streamflow and 1 April SWE within and outside of the Upper Klamath Basin (Figure 3).  This association with Upper Klamath Basin hydrology is strongest during the current warm phase of the Pacific Decadal Oscillation (PDO) (1978 – present) but does persist to a lesser degree into the previous PDO phase (1945 – 1977).  This signal is not only present in the Upper Klamath Basin but also extends to other regions in the western US.

5)  Incorporating fall and early winter values of the TNI into Upper Klamath Basin seasonal streamflow prediction models reduces the uncertainty of early season streamflow volume forecasts (Figure 4).  Results suggest that the TNI reduces the early season forecast uncertainty observed in the 1 December, 1 January, and 1 February forecasts for both the Sprague and Williamson rivers. An early more accurate streamflow prediction model should provide the USBR with better tools to make earlier decisions regarding water resource management in the basin.