Weather Service Warnings Use Case

Weather Service Warnings; May session of the ESIP Discovery Cluster

May 18, 2023

Meeting Introduction

Weather service warnings are issued and distributed to the masses to deliver accurate and actionable information to save lives and property in response to weather events. The discovery cluster has invited Nelson Vaz, a warnings coordinator for the National Weather Service, to share the repeatable processes that the weather service uses to forecast and issue warnings. What tools do NWS forecasters use to generate warnings, and how is this information disseminated to end users to provide locality specific, timely, and actionable information.

The Discovery Cluster has developed a Usage Base Discovery (UBD) paradigm for dataset discovery.  UBD could help document how expert users like meteorologist derive actionable information from data, like weather models and satellite products.  Dataset usages are cataloged and documented in a knowledge graph to help make explicit how expert users apply Earth Science information to inform specific derived products and analyses.  The outcome of this meeting with Nelson Vaz may be captured as submissions to the UBD knowledge graph.  However, the weather service warnings use case provides more than this, as the weather service is an example of effective information delivery in response to specific events.

The Discovery Cluster has previously tried to understand the specific distillation of very complex Earth Science information into clear and actionable end user information, like in the case of flooding information (See UBD Use Case). Are there other uses for the information that weather service staff are accessing to develop warnings, and how could the general public discover this additional information in a timely fashion?  What improvements do we need to make to the UBD tool to address this and similar event based use cases (eg. wildfire response), and what can we glean from how the Weather Service disseminates this information for weather warnings?

Meeting Notes

Nelson Vaz, Warnings Coordination Meteorologist for the Weather Service New York Office serving forecast and warnings for the tri-state area: Southern Connecticut, Lower Hudson Valley, New York City, Northeastern New Jersey, and Long Island. A new mission for the NY Office is weather support for public safety officials. They serve ~19 Million people, or 7% of the US population. NY Office collects weather observations, generates 7-day forecasts, provides warnings, and supports several specialized forecast products. Partners include public officials and private partners like the Port Authority hospital systems and utilities. Media is a partner that amplifies the information. The diversity of weather hazards, and of the population in the Tri-State present a lot of different partner needs and a lot of vulnerability that can be a challenge to the New York Office.

Discussion of the use of technology, such as calibrating model output to observations and other modeling techniques. There are more manual processes on the end user side to apply the model output to weather forecasting.

Weather Service has 12 National Centers, the experts on severe weather producing a lot of research and innovation in forecasting technology.

Advanced Weather Interactive Processing System (AWIPS) used to visualize and interrogate satellite, observational, radar, modeling output. They do a lot, including observations, forecast warnings and the decision support. Weather models are key, Days 3-7, use big picture, synoptic, mesoscale, down to city level as you get closer to an event - heavy snow banding, flash flooding, etc. Course resolution and high resolution models including GFS, NAM, European, UK Met. Weather ensembles are huge, or ensembles of an individual models where you change how the model runs to give a spread of solutions what could happen in the future. Take those observations, and see if the model is initializing things correctly, and trade off the strengths and weaknesses of some of these models to come up with the most likely scenario, the best case, and the worst case. Also have automated probabilistic forecasts to compare with the human experience forecasting. Can calibrate them based on model history and performance, bias correction to fine tune them a bit more.

Produce a gridded forecast database, take the forecast, can use certain models and model blends, putting all this analysis into a final gridded forecast. From this we produce text products.