What have we learned? A quick glance at emergency managers’ use of Probabilistic Hazard Information (PHI)

The Emergency Manager (EM) experiment has hosted over 40 emergency managers since its inception in 2016, including 19 EMs in 2019!  Before coming to the NOAA HWT housed in the NOAA National Severe Storms Laboratory in Norman, OK to participate in the Experiment, EMs completed a pretest questionnaire.  In particular, they answered questions about their demographics, weather product use, and decision-making styles.  

As a whole, our EMs have been predominantly white (91%), male (71%), college graduates (75%), with degrees ranging from Associate degrees to Doctoral degrees, who work primarily for government agencies (86%).  The majority of participants indicated working as EMs for more than 10 years.  They represent a variety of jurisdictions and services—cities, counties, state governments, hospitals, utilities, and federal agencies—from across the US.

EMs completed various questions assessing their decision-making styles and preferences.  One measure, the Subjective Numeracy Scale (Fagerlin et al. 2007), assessed EMs’ self-reported abilities and preferences for working with numbers over words.  Sample questions included, “How good are you at working with percentages?” and “When you hear a weather forecast, do you prefer predictions using percentages (e.g., ‘there will be a 0% chance of rain today’) or predictions using only words (e.g., ‘there is a small chance of rain today’)?”  Thirty-nine EMs answered these questions.  As a whole, EMs self-reported high numeracy, Mean = 4.48 (out of 6), Standard Deviation (s) = 0.98, which seemed to be driven by a preference for receiving numerical over verbal information, M = 4.66, s = 0.99 more than self-reported ability for working with numbers, M = 4.30, s = 1.36.

For a more detailed description of the Experiment, please see our previous posts here and here.  Let’s dive right into the good stuff!

Was PHI useful?  EMs overwhelmingly reported that PHI was useful for their decisions, M = 6.84 (out of 7), s = 0.99, n = 19.  Specifically, it delivered information quickly, M = 6.78 (out of 7), s = 0.43, n = 18; made EMs more confident in their decisions, M = 6.72 (out of 7), s = 0.58, n = 18; delivered pertinent information, M = 6.60 (out of 7), s = 0.70, n = 10; and was easy to use, M = 6.28 (out of 7), s = 0.75, n = 18.  Through discussions, EMs elaborated that one of the nice aspects of PHI was the ability to pinpoint where the most damaging impacts could be expected and give people more time to prepare.  Their warnings and communications could be more nuanced in that they were better able to indicate which areas should take immediate shelter, which areas should stay on high alert and be ready to shelter soon, and which areas had a bit of time before they could see potential impacts.

What is it about PHI that is most important/useful?  Does that depend on the hazard to which they were responding?  For both tornadoes and severe thunderstorms (high wind/hail in particular), similar elements of the PHI and the Enhanced Data Display (EDD; i.e., the tool used to deliver PHI to EMs; see below)

This is a picture of what the Enhanced Data Display (EDD) tool looks like from the user’s perspective

emerged as highly important: the forecaster discussion box, time of arrival estimate, hazard probability of occurrence, having one-hour lead time, and the PHI plume coloring scheme.  Of these elements, two were significantly different in their importance across hazards.  First, the importance of hazard probability of occurrence for understanding and communicating about tornadoes (M = .90) was significantly higher than for severe thunderstorms (M = .86), t(39) = 2.48, p = .02.  Second, the importance of the PHI plume coloring scheme for understanding and communicating about tornadoes (M = .83) was significantly higher than for severe thunderstorms (M = .79), t(39) = 2.70, p = .01.  The graphic below depicts these elements in the EDD and the data tables depict the differences in these PHI and EDD elements.

This is another view of the EDD tool with clear labels of the featured elements (e.g., the Forecaster Discussion box and Time of Arrival estimate).
This graph depicts participants’ ratings of the importance of each PHI element to their understanding and communication of Tornado hazards
This graph depicts participants’ ratings of the importance of each PHI element to their understanding and communication of Severe Thunderstorm hazards.

 

 

 

 

 

 

 

 

 

Can EMs successfully manage all of the different types of probabilistic information that were presented to them?  This research aim was unique for the 2019 Experiments because EMs were presented with multiple products along the continuum of severe weather events, each conveying different probabilistic information.  On a scale of 1 (extremely easy) to 5 (extremely difficult), EMs found it very easy to interpret all of the different kinds of probabilities they received for each case, M = 1.36, s = 0.51, n = 11.  In discussions, EMs would note that even though the probabilities differed in the ways in which they were derived (e.g., climatology based vs. storm based) or what they were communicating—among other dimensions—, they “all made sense together” and “told a complete story.”  EMs said having the complete picture helped, and having all of the probabilistic information helped them be more confident in making some decisions sooner while holding off on some other decisions.

So, after immersion in our probabilistic information world, did EMs walk away preferring probabilistic information?  Well, EMs actually had an overwhelming preference (95% of those asked; n = 20) for a combination of probability and text information.  Even among those not asked this question in the posttest survey, discussions revealed support for this preference.  In particular, EMs noted that, while they may prefer receiving probabilistic information for their own personal use (to help with interpreting the incoming storms), they would not want to pass probabilistic information on to their constituents (e.g., their bosses, their publics, their peers, etc.).  Therefore, they would also need textual information that is ready to pass on to others for their consumption.

Lastly, a really cool feature of the EDD is the user’s ability to customize the different layers that are shown (e.g., radar, map, PHI plumes) based on their personal preferences.  These layers included the color schemes used for the PHI plumes and the particular hazard plumes that are displayed.  Participants predominantly (70% of those asked) preferred a monochromatic color scheme over a rainbow color scheme; this color scheme separated each hazard by color with varying probability levels displayed as gradients of the assigned color.  For example, tornado PHI objects were red, where higher probabilities were represented by deeper/darker shades of red and lower probabilities were represented by fainter/lighter shades of red.  Severe wind/hail PHI objects were yellow with the same probability coloring concepts.  EMs noted the monochromatic plumes were “less cluttered/confusing,” “easy to discern on the map,” and “[didn’t] blur out the radar.”  They also noted the rainbow-colored PHI plumes were “too busy,” “distracting,” “[easily] confused with radar.”  EMs noted difficulty trying to tell differing hazard plumes apart when they were all rainbow-colored.  Thus, the monochromatic plumes were preferred because it was easier to distinguish between threats and ostensibly easier to discern the probabilistic information.

This picture depicts the monochromatic PHI coloring scheme; the red coloring is used for Tornado PHI objects.
This picture depicts the rainbow PHI coloring scheme; PHI objects are depicted with rainbow colors similar to radar.

 

 

 

 

 

 

 

 

 

The findings discussed here only briefly touch on a few of the concepts that were tested and evaluated over the years of the EM Experiment.  Taken together, our EM sample reported relatively high numeracy and a preference for receiving numerical (over textual) information.  This preference was expressed in their responses to PHI; they found PHI provided important information and improved their decision-making.  Further, they had an overwhelming preference for the PHI visualization that made it easy to distinguish hazards and probabilities of occurrence, i.e., the monochromatic color scheme.  Unfortunately, our sample is small.  We don’t have the variability in our EMs numeracy scores to speak to whether these preferences change as a function of self-reported numeracy, but that is a great long-term goal for future research!  Stay tuned for more updates and information about other aspects of the Experiment!

Tags: None

The 2019 Emergency Manager Experiment Unwrapped! Insight Into This Year’s Experiment and Product Innovations

The Spring 2019 Emergency Manager (EM) Experiment hosted eight EMs across two weeks in May.  The EMs hailed from a variety of jurisdictions and services–city, county, and state governments, as well as utilities and hospital networks–and represented several different states, including New York, Colorado, Kentucky, Florida, Oklahoma, and Ohio.  

In the experiment, EMs worked archived cases with the help of  experimental forecast products under development at NSSL, CIMMS, and SPC.  The products have been generated as part of the Forecasting a Continuum of Environmental Threats (FACETs) program, which seeks to improve the communication of Probabilistic Hazard Information (PHI).  The new products all represented forecast uncertainty in different ways, offering deeper insight into forecaster thinking about storm likelihood, timing, and location.  EMs first received longer-range forecasting products that were issued days before the event, and worked their way to products issued at the warning time scale, covering a fuller “continuum” of forecast information.  The archived cases encompassed a variety of severe weather threats, e.g., severe thunderstorms, QLCS storms, and supercell tornadoes that occurred across the continental US.  

Each day, the participants began with long range SPC Convective Outlooks–Day 4, Day 3 and Day 2.  Then, depending on the issue time of each product, participants saw Day 1 outlooks, Mesoscale Convective Discussions, and Watches.  Interspersed with these products, participants received an experimental Potential for Severe Timing (PST) product, experimental Warn-on-Forecast (WoF) output, and/or experimental hazard timing graphs from SPC.  Periodically throughout the case, participants completed micro-surveys asking about trends they were noticing, details they were keying in on, and decisions/actions they were taking based on the information received.  There were also mini focus groups at each time step to discuss the same topics in more detail.  As the week progressed, participants received more of the experimental products.  On Tuesday, only the PST was given; on Wednesday, participants saw the PST and WoF; and, on Thursday, participants saw the SPC timing products and WoF.  At the end of each case, at the warning timescale, participants received warning-scale PHI.  Then at the end of the day, a wrap-up survey and focus group evaluated how participants viewed the information and forecast evolution in light of what occurred.  

What are these experimental products I just mentioned? 

Sample Potential for Severe Timing (PST) Product

The PST is a product that specifies the 4-hour window(s) for the areas where severe weather is most likely to occur (see graphic to the left).  Ideally, the PST would be issued with the 11:30 Day 1 Outlook and would be valid until the end of the convective day.  This tool is meant to help provide early and specific timing information to users to help facilitate their planning during severe weather days (e.g., should schools be closed, extra staffing brought in, shift scheduled temporarily modified). 

 

Sample Warn-on-Forecast (WoF) Product

The WoF output provided to EMs is a timing product that identifies areas where convection is most likely to develop over the next few hours, and the associated probabilities that it will (see graphic to the left).  Further, the output updates every hour.  The SPC Hazard Timing Graph takes the Day 1 Outlook and breaks it into four-hour windows of time, allowing participants to see when hazards are most likely to occur in their area within a 24-hour period.  Ideally, this graphic would automatically update with updates in forecast guidance. This tool would help users know, for example, when a storm is expected to reach the “moderate risk” threshold and for how long.

The last day of the experiment consisted of extensive debriefing and reflecting.  EMs completed post-week surveys and a focus group interview which asked for their deep evaluations of the tools and products they used.  We wanted to know what they liked/did not like, what worked, what was impossible to figure out or use, and their views on how PHI could be implemented in operations.

 

Ok, so what’s next?

Right now we are in the preliminary stages of analyses.  As a research team, we have met to discuss how to best utilize the wealth of information we gained from the new methodology used this year and rich feedback we received.  Analysis plans have been formed and are underway.  Product development is being informed by observations and early observable trends to continue moving toward operational status.  We are also planning the Fall 2019 Hazard Services PHI experiment for an integrated warning team–forecasters, broadcast meteorologists, and emergency managers working together.  The emergency managers’ portion of the Fall experiment will again feature many of these products, but within a new platform: Hazard Services.

Tags: None

Week 1 is underway!

Week 1 of this year’s Spring EM Experiment kicked off yesterday!  EMs were able to get a first look at some of the new experimental products, and they began to work PHI in the Enhanced Data Display (shown on the screens).  In the coming days we’ll add even more new products.

EM round table discussion
EMs’ first looks at new experimental products.
EMs working PHI in the Enhanced Data Display (EDD).
EMs working PHI in the Enhanced Data Display (EDD).
Tags: None

This year’s experiment is about to kick off!

Spring is buzzing around the building and we are excited to have this year’s EM experiment just around the corner! It’s PHI-nally here!

We have Emergency Managers coming from all over the country–Colorado, Florida, Kentucky, New York, Michigan, Ohio, and Oklahoma–representing a variety of jurisdictions, such as cities, counties, states, hospital networks, utility companies, and more!  We’ll be switching things up a bit this year so stay tuned for pictures and updates of the action.  Week 1 starts next week, May 13, and Week 2 starts right after that on May 20.

Tags: None

What IS the PHI-EM experiment and why is it being run?

Since 2015, an ever growing and evolving team of researchers at CIMMS/NSSL have been collaborating with academic researchers from around the country to host annual springtime Emergency Manager (EM) experiments, including the Probabilistic Hazard Information (PHI)-EM experiment.  These experiments were designed to elicit feedback directly from EMs about new technologies under development at the lab. The PHI-EM experiments have specifically focused on gathering feedback about a new technology called Probabilistic Hazard Information.

  

The PHI developed at NSSL is meant to help all parties in the communication chain make more informed decisions about impending weather threats.  This new PHI includes the probability of a storm to produce tornadoes, severe thunderstorm hazards (including high wind and hail), and lightning in the next hour.  It also updates rapidly, and is most commonly viewed as a plume of probabilities projected ahead of a storm (see picture below). PHI plumes are created by a forecaster using a suite of tools and algorithms.  The forecaster makes decisions about the probabilistic trend of the storm based on his/her tools (e.g., radar, ensemble model data). After creation, PHI plumes are delivered to EMs and Broadcast Meteorologists through the Enhanced Data Display (EDD).  In addition, the EDD offers other useful information about timing, severity, and the anticipated storm track. EMs and Broadcasters then use the PHI plumes to make decisions for their jurisdictions such as sounding sirens or canceling events (EMs), and whether to run a crawl or cut into on-air programming for live coverage (Broadcasters).  

 

EM feedback is invaluable to the evaluation process.  While PHI may constitute a huge breakthrough on the forecasting end of the chain by conveying richer and very localized forecast information, if the displays are hard to understand or not providing meaningful and needed information on the users’ end of the chain, then more work needs to be done!  Thus, EMs have been brought in as participants to help assess the viability of this new information and product. Specifically, EMs have been asked for their feedback on the types of PHI that have been created thus far as well as the EDD interface to view them. They have been asked to comment on which aspects of PHI are most/least helpful and easy/difficult to use and understand.  Overall, as researchers, we are trying to ascertain the feasibility of the PHI and EDD and move them toward operations!

Tags: None

2019 Spring EM Experiment Recruitment has started!

Hi everyone!  It’s that time of year again!  That’s right, the call just came out and applications are rolling in for the 2019 Spring EM Experiment!  More details can be found on the Recruitment Page for the project!  Take a look through the details when you get a chance.  We hope all of the Emergency Managers out there will consider applying!

Tags: None

We’re Back!

Hello everyone!  My name is Cassandra Shivers-Williams and I am currently a post-doc researcher at CIMMS/NSSL within the Societal Impacts Group (SIG).  I have been working with Emergency Managers (EMs) in the Hazardous Weather Testbed experiments since 2016!  My wonderful colleague Kodi Berry started this blog as a way to talk about all of the awesome EM research happening here!  Well…I’m picking it up and taking it over! (You may have noticed the site facelift =)  )

This blog will be highlighting lots of fun with EMs and the Probabilistic Hazard Information (PHI) Experiment, sprinkled with some other ongoing social science research within the SIG!  Our sister entity, KPHI TV, also conducts social science research, but primarily with Broadcaster Meteorologists.  That work is blasted on the KPHI TV Blog!   I encourage everyone to check out that blog for more fun with PHI!

This website has been updated with some background information about the work we do and the research team, opportunities for EMs to participate in testbed research, presentations/publications of our EM work, and some alumni!  Take a look around and check back in periodically to see the latest and greatest updates from the SIG!

Tags: None

Week 1 Underway

This week emergency managers from hospital, city, county, and state jurisdictions are evaluating how they might use probabilistic hazard information for decision making during severe convective weather. 

Tags: None

2018 EM Selection

We had an unprecedented response to our call for EM applicants for the 2018 HWT PHI EM project. Over 50 EMs applied for 8 spots! So many strong candidates this year! We are actively pursuing funding for next year so we hope those we had to turn away this year will apply next year.

Thank you to our friends in the National Weather Service for helping to spread the word about our project!

Tags: None

Welcome!

This is the information page for emergency manager (EM) participation in NOAA’s Hazardous Weather Testbed (HWT). EMs in the HWT help evaluate probabilistic hazard information (PHI) for severe weather, the core of the future National Weather Service severe weather warning paradigm.  PHI is part of NSSL’s overall FACETS initiative.  Learn more about FACETS here:

https://www.nssl.noaa.gov/projects/facets/

Tags: None