Comparing Feature-Based vs Ensemble-Based Tools to Estimate and Communicate Weather Forecast Risk and Uncertainty

SD Mines is developing two parallel means of assessing forecast risk and uncertainty or “Confidence Index.” One method leverages the spread of forecast ensembles to meteorological error, which requires a number of forecast members to evolve a number of scenarios of developing weather. The second is a feature-detection method in which key weather features that impart confidence or risk are identified and quantified with respect to past resulting forecast error. This technique is also applicable to other Hazard vs. Fragility systems.

Here we present both methods as candidate guidance products in the forecast decision-making process. This is done by creating a Risk vs Fragility (or Hazard vs Error) state space which produces a “wedge of uncertainty” with low values of forecast error at the low-risk end with increasing spreads of both low and high error as the risk or hazard increases. We articulate existing and recent risk by plotting current (pre-validation) and past forecast risk scores in this space to show the evolving patterns of complex weather over a given region.