In response to drought and increasing demand for water, water managers across the West have expressed an urgent need for more information about future snowpack and water supply conditions. Because water managers make decisions within specific watersheds and regions, their need for water supply predictions is at the local and regional level. Further, there are a range of water management decisions – including reservoir operations, annual water supply decisions, and longer-term infrastructure and water rights acquisition decisions. From an annual water supply perspective, understanding how likely it is that drought conditions will prevail over the months-to-years time frame is critical information. Other types of water management decisions may require predictions on different time horizons.

System storage turns out to be critical to climate information needs, in part because the amount of storage capacity in individual watersheds can significantly buffer the effects of climate variability. For example, the reservoirs on the Colorado River system, including Lakes Mead and Powell, have so much storage capacity (four times the average annual flow) that the system can deliver at a “normal” level even under severe drought conditions for extended periods of up to a few years. Meanwhile, the reservoir capacity in California is significantly lower relative to the average annual flow of its rivers, in the range of half to a full year. This means that drought vulnerability within the California state reservoir system is higher than it is on the Colorado River system.

In the western US, precipitation and temperature variations have been linked to variations in the ocean circulation associated with ENSO, PDO, and other indices (Pierce, 2005; Cayan et al., 1999; Quan et al., 2006, among others). There is a need to combine instrumental, paleo, and model data to investigate the dynamics behind PVD and AMV. Once we have a better understanding of the mechanisms that cause the variability, our capacity to predict future conditions based on analysis of current conditions and trends will improve dramatically. Dynamical ocean and ocean-atmosphere models and the methods used to employ the observations that initialize them still require substantial improvement before their predictions of ocean evolution years into the future can be used with confidence. Of particular importance is the need to predict “phase shifts” within decadal oscillations.

The current mismatch between our ability to predict climate and the demand for long lead climate predictions is a source of frustration for water managers. It may be another 5-10 years before the dynamical models can predict natural decadal climate variability, yet that does not mean that no information is available. Two obvious sources of information can be mined now. The first source of information lies in the observed historical record based on proxies (such as tree rings or lake levels). The second, a likely result of humankind's industrial activity, is the slowly unfolding reality of human-caused climate change. Thus, given no additional information on the specific evolution of natural decadal variability in the coming 10 years, one could specify the range of historical variability, while recognizing that it rides atop trends associated with human-caused climate change.

Though the skill associated with decadal climate prediction is relatively low, it clearly has value if estimates of uncertainty are available to decisionmakers. Developing the capacity to communicate the sources of the uncertainty is often challenging since most water managers are not experts in atmospheric physics and other relevant fields. Critical to moving forward in this arena is identifying the right questions and close collaborations between water managers and climate scientists to ensure that outcomes are at the right scale and are clearly communicated.

A recent workshop focused on identifying the decadal prediction needs for water management identified the following key research questions:

• What is the role of the land surface, particularly in the persistence of drought conditions?

• Is persistence of climate conditions recognizable and is it predictable?

• What are the mechanisms of persistence?

• Water managers are most concerned about our ability to predict transitions from one climate condition to another. Can we identify which physical mechanisms drive the phase shifts?

• What are the key sources of predictability of impacts that key users care about (water managers, fisheries, fire managers, and so forth)? This should inform the observational (monitoring) priorities for supporting a decadal prediction system.

• Can we separate the anthropogenic signal from natural variability? This is important for identifying future trends in water supply and in predictive capacity.

There are strong economic and social reasons why working on decadal climate prediction should be a high priority research investment. It is clear that we can reduce risks related to water supply availability with better information about the likelihood of wetter vs. drier years, especially given our understanding of the seasonality of precipitation. The high priority questions that need further research are emerging from a series of conversations between researchers and water managers. The critical question now is whether funding can be found to support this important work.