Forecasting is a critical service that helps organizations make informed, analytical decisions regarding water resources. Forecasting provides actionable information for flood warning, hydropower operations, water allocation, agriculture operations, and more.
To help us understand the complexity and importance of flow forecasting systems and services, Michael Thiemann, Senior Water Resources Engineer and Data Systems Architect for RTI's Center of Water Resources, had a conversation with Ken Nowak from the Bureau of Reclamation's Research and Development Office to discuss factors impacting forecasts and how one can determine how useful a forecast truly is.
Q: How does the uncertainty of streamflow forecasting impact your use of forecasts?
A: Contending with forecast uncertainty is part of reservoir operations. Forecasts with high or unknown uncertainty may result in more conservative operations as compared with a more confident forecast. Forecasts should strive to expose, quantify, and, to the extent possible, minimize uncertainties. In the case of an ensemble forecast, if robust characterization of uncertainty results in a wider forecast range, that is preferable to a tighter forecast distribution that is overconfident.
Q: How do you determine whether streamflow forecasts are useful? (How they are used in decision making)
A: The usefulness of a forecast is informed by decision-making needs and how well currently available products meet those needs. Aspects of a forecast include outlook horizon, temporal resolution, frequency of update, historical performance, etc. These dimensions are juxtaposed with specifics of the facility, basin, and situation. Of note, the skill/performance aspect of usefulness can be informed by analyzing hind-casts, but operational forecast performance and track record often have significant weight within the reservoir operations community.
Q: How good do forecasts need to be and how is “good” determined? (For instance, is any forecast better than nothing?, Good enough to determine categories (imminent flood/no-flood; seasonal drought/no drought)?)
A: Building on the response to the previous question, the skill of a forecast is important, but not the only factor in considering a forecast’s usability. With appropriate context, useful information can be gleaned from different [skill] forecast products. At minimum, forecasts should reliably point operations in the “right direction” (e.g. are you building sufficient storage space in your reservoir before a flood).
Q: What aspects of forecasts do you see as the most important to “get right”?
A: Reservoir operations often balance providing flood control and water supply, among other objectives. During the crux of that operational balance, such as spring runoff and major precipitation events, it is important for forecasts to capture the “hydrology signal” to have operational value. A valuable complement is an accurate representation of forecast uncertainty; this allows the forecast user to more effectively manage risk given the circumstance.
Q: How do you gain confidence in the forecasts?
A: Hind-casts are important for gaining confidence in operational forecasts. These efforts should be as representative of the operational forecasting process as possible. Further, for maximum utility, hind-casts should cover a period that is sufficiently long and hydrologically varied to offer a robust assessment of the forecast’s performance. This, coupled with experience using the forecast product operationally, are the primary avenues for building forecast confidence. Additionally, transparency in forecast development and performance metrics from the issuing agency can be assistive in building confidence.
Q: How will your knowledge of historical or recent forecast performance impact your confidence in the forecasts?
A: In considering a forecast product, historical performance plays a significant role in assessing the forecast’s value proposition. However, as consideration of a forecast shifts from overall performance to an individual event, the role of a forecast’s recent performance becomes increasingly important. Often operators will have multiple forecast products at their disposal. The forecast with the best acute performance may be given increased weight as operations are projected into the future. Furthermore, to the extent that a forecast is consistent with or corroborated by hydro-meteorological observations, confidence is likely to be increased.
Q: What makes measuring or judging the quality of forecasts challenging?
A: There are a variety of factors that contribute to challenges associated with judging a forecast’s quality. Ultimately, familiarity with a basin’s hydrology, history, and operations are indispensable for forecast evaluation. This knowledge is developed over time with experience and generally bears the greatest influence on how a forecast’s quality should be judged.
As Nowak makes clear, flow forecasting provides critical information for managing vital water resources and assessing the performance of the forecast products provides the insights needed to utilize the forecast information appropriately.
At the RTI Center for Water Resources, we provide services that can help make critical water resources decisions based on real-time conditions, as well as conditions that are hours, days, weeks, or seasons into the future. To see what type of forecast offering may be best suited for your organization, take our short river forecasting quiz.