A New Application of Wavelet Analysis: Estimating Event-Timing Uncertainty in Streamflow Predictions

Common verification metrics used to evaluate streamflow simulations do not distinguish between errors in magnitude and errors in timing. Nevertheless, diagnosing timing errors explicitly has potential benefit for both practical forecast guidance as well as for model diagnostics. In high resolution prediction, a fundamental challenge to evaluating timing errors is objectively identifying what constitutes as an event for which the timing error should be calculated. Another related challenge is that timing errors are both time and timescale dependent, meaning that their evaluation requires a localized approach in both these dimensions.

Wavelet-based approaches offer a powerful verification tool for weather and climate applications across a range of scales. NCAR scientists have been working on a novel approach that uses wavelet analysis to estimate timing errors for events in hydrologic predictions. This provides a systematic methodology that both integrates advances in statistical significance to identify events and is appropriate for benchmarking timing errors in high-resolution prediction.

This research is funded by the NOAA Office of Water Prediction and the Joint Technology Transfer Initiative grant 2018-0303-1556911. This material is based upon work supported by the National Center for Atmospheric Research (NCAR), which is a major facility sponsored by the National Science Foundation (NSF) under Cooperative Agreement No. 1852977.

Wavelet analysis can be used to compare timing errors across hydrologic model versions, such as for three versions of the National Water Model, shown here.
Figure: Wavelet analysis can be used to compare timing errors across hydrologic model versions, such as for three versions of the National Water Model, shown here.