Shasta Lake, in northern California, has recently experienced reduced cold water storage, making it difficult to meet downstream temperature objectives for endangered winter-run chinook salmon spawning habitat. This study used a novel form of time series analysis to examine the causes, timing, and predictability of cold water storage in Shasta Lake. This analysis detected two independent modes of variability in Shasta Lake cold water storage. The first mode, representing variability during February--July and describing 64% of the overall variability in cold water storage, was negatively correlated with both the preceding year's late summer hypolimnetic discharges and that spring's air temperatures. A second mode, representing December--January and describing an additional 24% of variability, was negatively correlated with Shasta Lake fall water temperatures and winter air temperatures and positively correlated with winter inflows. These results suggest hypolimnetic discharges, air and water temperatures, and inflows act in concert to determine cold water storage in Shasta Lake. These results also suggest water column mixing should be promoted during the cold midwinter period and thermal stratification should be promoted the remainder of the year to minimize surface warming of the entire water column.