Evaluation of Data Display Methods in a Flash-Flood Prediction Tool

Flash flooding can be difficult to predict using traditional, rainfall threshold-based approaches. New initiatives like the Flooded Locations and Simulated Hydrographs (FLASH) project provide real-time information using rainfall observations to force distributed hydrologic models to predict flash flooding events. However, in order to address the goal of creating a weather-ready nation, system designers must not only possess tools that relay useful information, but such tools must also be able to communicate environmental threats to stakeholders in a clear and easy-to-use interface. Where previous research has addressed the performance of forecasting models, the present study uses a human factors approach to enhance FLASH's ability to present information to decision-makers (i.e., forecasters).

The present work discusses potential directions for future FLASH development with the goal of decreasing users' cognitive workloads and increasing forecasting accuracy and trust in the tool. The 2013 Flash Flooding and Intense Rainfall (FFaIR) experiment, conducted by the Hydrometeorological Testbed at the Weather Prediction Center (WPC), involved a daily evaluation of the previous day's flash flood forecasts using FLASH as a diagnostic tool. Observations of user performance during FLASH evaluations in the FFaIR experiment revealed dissimilarities between the goals of national-level and local-level forecasters, which were particularly evident in discussions about the display of algorithm results at a national level. Based on observations from the FFaIR experiment, a user testing study was developed to compare the effectiveness of two display methods—average versus maximum grid cell value. When combined, the FFaIR observations and preliminary findings from the user testing study provide valuable information to the future of FLASH as a tool to promote communication of flash flooding risk to forecasters and the general public.