Facilitating the Development of Hazard Services Recommenders

Recommenders in Hazard Services are powerful utility tools designed to aid National Weather Service (NWS) forecasters in drafting warning polygons based on radar, model, and observational data. Since the rollout of Hazard Services Version 1 in 2019, forecasters and focal points have had the ability to create new recommenders and customize their behavior. If combined with the Advanced Weather Interactive Processing System (AWIPS) Data Access Framework (DAF), focal points may create recommenders that utilize radar and gridded data for use in warning operations.

Hazard Services development uses Python which has a vast amount of training and documentation available. The DAF allows users to access the Environmental Data Exchange (EDEX) system databases to import data as Python objects. These data can then be modified and augmented to output recommendations for forecasters through the tools section of the Hazard Services Console. Presently, Hazard Services comes with a few recommenders preconfigured; however, development of these tools remains relatively underutilized by NWS offices. There is almost no limit in what can be done with the Python processing behind the recommenders. Utilities can be as narrow as creating a polygon around values in a grid to as ambitious as artificial intelligence models trained to create recommended outputs based on complex data. Since the initial rollout of Hazard Services the Warning Decision Training Division (WDTD) has created and maintained training on how to use the DAF, but have not yet created content on how to create a custom recommender with the DAF. The WDTD is planning on updating the fundamental focal point course over the next year with jobsheets and examples. One such example will include a modified flash flood warning recommender that utilizes the Multi-Radar/Multi-Sensor Coupled Routing and Excess Storage (CREST) model, the framework of which was developed by Robert Megnia of the Boston, MA NWS office.

While they cannot replace the expertise of NWS forecasters, recommenders could provide a first guess approach to hazards, build forecaster confidence, and have the potential to speed up the warning process.