WR Forecast Confidence Project

WR Forecast Confidence Project: During the last few years, WR Science Technology Infusion Division (STID) and the WR SOOs have embarked on science project to improve Decision Service Support (DSS) messaging and extend forecast lead time of significant events through improved use of the numerical model uncertainty and confidence information. To achieve this goal, STID implemented three major activities:

1) Development of a second generation Ensemble Situational Awareness Table. The Situational Awareness table compares the 42-member North American Ensemble Forecast System (NAEFS) mean with the running 3-week mean and standard deviation derived from the Climate Forecast System Reanalysis to identify the significance of the event (standardize anomaly). http://talcott.wrh.noaa.gov/satable/ Based on forecaster feedback that the asymmetric nature of standardize anomalies made interpretation difficult , event return intervals are also calculated which takes the anomaly asymmetry into account and produces values that are more meaningful to the forecasters. For example, the expected event is typical of a once in a five year occurrence. Another addition to the table is to compare the GEFS ensemble mean to the GEFS model climate to show how often the model produces a storm of this significance. The table is a very popular tool in WR that helps identify how significant the storm is and what feature of the storm is significant (eg: wind versus precipitation) which help in DSS messaging.

2) Development of Forecast Confidence Tool Kit: This was a 2 year project working with the WR SOOs to identify, assess, provide recommendations, and training about the emerging analytical model tools that the forecaster could use to assess model skill. STID developed a Google site web page with training/documentation that highlights the recommended analytical tools and enables the forecasters to more easily find and use. https://sites.google.com/a/noaa.gov/nws-wr-stid/projects/forecast-confidence. The tools range from assessing short term convective scales to long term synoptic regime changes.

3) Applying “Just in Time” training to help promote how to use new tools in operations: Forecasters learn best when applying new tools and science to the current situation. Toward this goal, STID organized a series of (approx. 15) GOTO conference calls where the scientist who developed the analytical model tool could explain how their application worked. To highlight how to use these analytical tools for real time situations, STID has been generating “just in time” emails prior to an event that illustrates how to use and interpret the tools. These emails employ graphic snap shots in a short, easy to read format that is distributed to all of the WR offices. After the event, STID has been generating post-mortem emails about how the analytical tools performed. Over the last 18 months, over 50 of these “just in time” emails have been generated.

Collectively, these three efforts have improved the use of model uncertainty and confidence information to improve forecast event lead time and DSS messaging, where the Forecast Office messages better focus on the expected significant impacts of the upcoming event.