Heavy Precipitation Object Tracking and Forecast Tools

Flash flooding events are a leading contributor to damages to property and threats to life by weather phenomenon. Heavy precipitation events that cause flash flooding are difficult to forecast due to a number of factors that models struggle to capture consistently. The National Weather Service’s Weather Prediction Center (WPC) forecasters generate many products to communicate the potential of flash flooding. To create these forecasts, the WPC uses a large amount of raw and post-processed model data. In an effort to assist in this process and to communicate model uncertainty efficiently, forecast tools have been constructed to track precipitation objects within the model and observation data. Three tools have been developed to analyze uncertainty and trend within convective allowing models (CAMs) using an ensemble-based approach. They employ NCAR’s Model Evaluation Tools (MET) Model Object Diagnostic Evaluation Time-Domain (MTD) to identify and track precipitation objects throughout the model forecast. These tools evaluate forecasted precipitation on the hourly and sub-hourly time scales. Uncertainty of the location of the precipitation objects is communicated through the use of object data from historically relevant events. Similar techniques are also applied to areas of forecasted snow, displaying the spatial variability of snowbands within an ensemble of numerical models. Verification of these products is conducted to assess the iterative process to improve the tools’ reliability and skill. We also take a look at how machine learning (ML) techniques can be employed to enhance the use of retrospective observations within the computation of the AnaPrO.