New rating system charts a path to improved tornado forecasts

All tornadoes — whether small or large — originate from thunderstorms, but not all thunderstorms are the same. Different environments and situations create forecasting challenges. For instance, nighttime twisters, summer tornadoes and smaller events can be tougher to forecast.

Researchers wanted to quantify how much tougher, and have published a new method of classifying tornado environments according to their forecast difficulty.

In a new paper published online in the Bulletin of the American Meteorological Society, University of Washington scientist Alexandria Anderson-Frey, and Harold Brooks from the NOAA National Severe Storms Laboratory describe a new way to rate and possibly improve tornado warnings.

“With this research, we’re trying to find ways to truly level the field related to the difficulty of the forecast situation,” said Brooks. “This will help us identify areas for research, as well as better understand the long-term historical statistics.”

 The paper presents a new method to rate the skill of a tornado warning based on the difficulty of the environment. It then evaluates thousands of tornadoes and associated warnings over the continental United States between 2003 and 2017.

The NOAA-funded study finds that nighttime tornadoes have a lower probability of detection and a higher false-alarm rate than the environmental conditions would suggest. Summertime tornadoes, occurring in June, July or August, also are more likely to evade warning.

“The forecasting community is not just looking at the big, photogenic situations that will crop up in the Great Plains,” said Anderson-Frey, the lead author. “We’re looking at tornadoes in regions where vulnerability is high, including in regions that don’t normally get tornadoes, where by definition the vulnerability is high.”

The technique could be applied to forecasts of other types of weather as well.

This research began while Anderson-Frey was a postdoctoral researcher at the Cooperative Institute for Mesoscale Meteorological Studies, a partnership between the University of Oklahoma and NOAA.

This story was adapted from a  University of Washington news release.

Share this:

Research continues to improve forecasting tools

Weather forecasters use a suite of sophisticated computer models to help them predict the weather every day. To make better forecasts, you need better models. That’s where researchers play an important role. 

Every spring for more than 20 years, researchers and forecasters have come together in the NOAA Hazardous Weather Testbed to evaluate and improve weather models designed to predict severe storms, with the goal of providing new tools for forecasters. The NOAA Hazardous Weather Testbed is a facility housed in the National Weather Center in Norman, Oklahoma. The physical space allows researchers, forecasters, emergency managers, broadcasters, and behavioral scientists to gather and study future forecasting tools and techniques. 

This year the research continues with one major difference. Instead of gathering side-by-side, participants in the Spring Forecasting Experiment will be working from home in a virtual experiment from April 27 to May 29.

“We want to know how forecasters can use different tools and how we can convey information to the public, all while documenting the performance of different forecasting models,” said Adam Clark, NSSL research scientist.

Clark is one of the experiment’s co-principal investigators, along with Israel Jirak, Science and Operations Officer with the NOAA National Weather Service Storm Prediction Center. In addition to NSSL and SPC, this year’s virtual spring forecasting experiment includes NOAA participants from the Global Systems Laboratory, Geophysical Fluid Dynamics Laboratory, Environmental Modeling Center, Weather Prediction Center and Aviation Weather Center, as well as a variety of governmental and academic partners including the University of Oklahoma, Iowa State University and the National Centers for Atmospheric Research, and international partners from as far as Australia, Brazil, and the UK Met Office. 

Screenshot of experiment participants displayed in Google Hangouts on the left side of the screen with an experimental forecasting product on the right hand side of the screen.
The Spring Forecasting Experiment went virtual this year as participants gathered from around the world to test and review experimental forecasting tools and modeling systems, as shown here. (Photo provided)

In addition to the analysis of regional high-resolution forecast models during the experiment, some participants will explore the Warn-on-Forecast system. The WoFS is a short-term forecast model that could be used to fill the gap in the watch-to-warning time scale. A watch is issued several hours in advance of potential storms, alerting the public of possible hazardous weather. A warning may be issued immediately before a storm and alerts the public they need to seek shelter immediately.

“Until a warning is issued, there is not much middle ground between a watch and warning,” said Clark. “There are several scientific and behavioral science questions as part of the experiment because the systems we’re testing are different, with different ways to visualize threats.”

Your not so typical day

Experiment participants — located throughout the world — begin each day with evaluations of the output of a variety of forecast models, followed by virtual small group discussions and analysis.

A woman sitting in a chair in her home office looking at two computer screens in front of her.
Burkely Gallo facilitating and participating in the Spring Forecasting Experiment from her home office.  The virtual experiment is from April 27 to May 29. (Photo provided)

“We might ask them, ‘did anything stand out to them, what was most interesting’ — we get a lot of value from those organic discussions and perspectives,” said Burkely Gallo, a lead facilitator for the experiment at the University of Oklahoma Cooperative Institute for Mesoscale Meteorological Studies, whose work supports the Storm Prediction Center.  “We didn’t want to lose the ability to have those conversations.”

A subset of participants then delves into issuing experimental severe weather outlooks and forecasts using Warn-on-Forecast output.

“We’ll get a lot of data out of this experiment because we were able to preserve what helps us answer our research questions,” said Gallo.

Even virtually, the experiment will continue to provide different perspectives from the severe weather enterprise blended together, resulting in better forecasts to save lives and property.

models Being used and tested

  • Global System Laboratory’s High-Resolution Rapid Refresh Ensemble
  • FV3
  • Warn-on-Forecast System
An introduction to the NOAA Hazardous Weather Testbed Spring Forecasting Experiment provided to participants.  Even virtually, the experiment will continue to provide different perspectives from the severe weather enterprise blended together, resulting in better forecasts to save lives and property. (Photo provided)
Share this:

Study investigates how future profiling observations improve forecasts

The DOE ARM Southern Great Plains site in Lamont, Okla.

Monthly Weather Review has published the results of an observation system simulation experiment (OSSE) that shows how measurements from various hypothetical remote-sensing networks would impact weather analyses and forecasts.

A team from the Cooperative Institute for Meteorological Satellite Studies and the Department of Atmospheric and Oceanic Sciences at the University of Wisconsin-Madison, lead by Dave Turner who is now at NSSL, performed the experiment.

Currently, the NOAA National Weather Service (NWS) launches weather balloons every 12 hours at 92 stations across the U.S.  As the balloon rises, temperature, pressure, wind and water vapor data are collected and transmitted to be used in weather analyses and forecast models.  However, for a model to correctly predict the strength, timing, and location of precipitation, more dense and frequent observations of the lower atmosphere are needed.

The experiment simulated observations from three different ground-based temperature and water vapor profiling technologies, and a wind profiling system.  Systems evaluated in the study included a Doppler wind lidar, a Raman lidar, a microwave radiometer, and the Atmospheric Emitted Radiance Interferometer.  Pseudo-observations from these systems were used investigate the improvement in the analyses and short-term (0-12 hour) forecast of a cold-season convective event over the central portion of the U.S.

The researchers found that the simulated array of profilers resulted in better analyses of how water vapor is transported. The results also showed that forecasts of accumulated precipitation were most improved when data from the multiple sensors were used.  Researchers also compared the different water vapor profiling technologies and found each yielded approximately the same improvement to the forecast.

This research provides important information to help determine the cost/benefits of these potential upgrades to the NWS observational network.

Share this:

NSSL’s Dave Stensrud teaches course in China

Stensrud 007 (Large)NSSL’s Dave Stensrud, Ph.D., recently gave a Short Course on “Parameterization Schemes for Numerical Weather Prediction Models” at the Institute of Atmospheric Physics in the Chinese Academy of Sciences in Beijing.

The course was a very intensive introduction to the parameterization of physical processes in numerical weather prediction models.

“This is a challenging topic, but one that is becoming more and more important as numerical models are used to study so many aspects of weather and climate,” said Stensrud.

Over 100 graduate students and faculty members attended the two lectures, the first in a series of Short Courses that National Weather Center Scientists will teach at the Institute of Atmospheric Physics during the next couple of years. The lectures are part of a multifaceted effort by the National Weather Center to play a leading role in U.S. interactions with the Chinese Atmospheric Science community.

Stensrud is Chief of the Forecast Research and Development Division of the NOAA National Severe Storms Laboratory and an Adjunct Professor in the OU School of Meteorology. The Short Course was based on Dave’s recently published book on “Parameterization Schemes: Keys to Understanding Numerical Weather Prediction Models” (Cambridge University Press, 2009).

Share this: