The “Value” of forecast information and improved skill has different qualitative and quantitative interpretations:
Direct measures of value include hydropower generation or storage volumes, but value can come in a range of social, environmental, and economic forms. Value includes improved trust between operators and end-users; value is fish habitat sustainability; value is farm production and economic gains; value is reduced risk and dam safety; value is sustainable water for rituals and ceremonies. Assessing the value from improvements in forecast skill is a multi-perspective and not necessarily quantitative approach.
Conservative bias in reservoir management reduces the potential value from improved forecast information:
Historical poor performance of reservoir metrics from errors in forecasts has resulted in issues with downstream stakeholders, including irrigators, flood managers, and recreational groups – those directly affected by reservoir management decisions. To limit the risk of over-promising, operators act conservatively to maintain an, ‘under-promise and over-perform’ approach to operations. This conservatism helps limit complaints from end-users but is sub-optimal and reduces the potential value achievable from the system. Efforts to improve forecasts and the resulting utility of the system may not be fully realized due to operator conservatism. To realize greater benefits, both improved forecast skill in addition to end-user confidence in best utilizing this information is required.
Legal or contextual parameters may override benefits from improved forecast skill:
The western United States follows a prior appropriation doctrine of water allocation; thus, irrigation and water supply deliveries are prescribed by law, rather than what may ‘optimize’ production. Increasing forecast skill can support improved decision-making, but laws, agreements, or other constraints may limit the potential benefits from the improved forecast skill.
Reduction in ensemble streamflow bias provides significantly better outcome metrics over ensemble dispersion when evaluated in an optimization framework:
Using an optimization framework based on the Stochastic Sampling Dynamic Programming algorithm, which explicitly uses all members of a forecast ensemble, can result in much better outcomes when bias is reduced compared to reductions in dispersion of the ensemble. Therefore, efforts to improve forecast skill that reduce bias should provide much greater value to the end-user community.
Optimization tools nearly match historical performance only when using perfect foresight, indicating real-time assimilation of data beyond ensemble members:
Optimization models provide an unbiased means of directly comparing performance metrics using a range of variable ensemble forecast quality but can only match historical performance levels when using perfect foresight – information not available to operators in real-time. This is an indication that even though operators are using the same probabilistic information to inform decisions, they are supplementing with additional data beyond the ensemble streamflow forecasts. These data include either qualitative or quantitative assimilation of weather forecasts, opinions of Nation Weather Service operators, real-time ground based hydrometeorological observations, or operator experience and domain knowledge – information not easily ingested into an optimization framework.
Water supply forecast errors are associated with reduced farm yields and economic returns for irrigator:
An analysis of annual irrigation and production data for farms receiving water from Dolores Water Conservancy District (via McPhee Reservoir on the Dolores River) indicates that seasonal (April-July) water supply forecasts provide tangible economic value to producers. Alfalfa and other hay producers must make early season planning decisions before total water supplies and allocations for the year are known with certainty. As a result, they must rely at least in part on projections based on streamflow forecasts. Using an ex-post analysis that allows us to control for actual full-season water supply levels, we find that larger forecast errors are associated with lower annual production levels. Therefore, improved forecasts with lower error can measurably increase farm revenues and income.
Higher confidence in the timing and size of excess reservoir inflows in wet years would result in higher economic value for recreational rafting:
Because the Dolores project’s main priority is to supply water for agricultural users, planned spills from the reservoir for whitewater rafting are only contemplated in years when there is high confidence of excess water supplies – i.e., when inflows are expected to be greater than reservoir storage capacity. Even then, uncertainty regarding the timing and size of these excess flows often means that dam operators are unwilling to make spill announcements with the advance notice that boaters prefer for planning multiweek trips. When, for example, a 10-day spill on the Dolores River is fully utilized for rafting, we estimate that it can provide recreational benefits of $1M to $2M. Therefore, in wetter years improved forecasts can result in meaningful benefits for rafters; however, these benefits are primarily associated with improved shorter-term streamflow forecasts based on weather conditions rather than on seasonal forecasts based mainly on snowpack estimates.
NASA impacts and benefits assessment from improved streamflow forecasts