Societal Aspects of the 2008 Super

Tuesday Tornado Outbreak

WAS* ISWAS* ISweather & society * integrated studies

www.sip.ucar.edu/wasis/

CULTURE CHANGE

Sponsored by the NCAR Societal Impacts Program

WAS* ISWAS* ISweather & society * integrated studies

www.sip.ucar.edu/wasis/

CULTURE CHANGE

Sponsored by the NCAR Societal Impacts Program

Kevin Barjenbruch* and Julie Demuth**

*NWS Salt Lake City WFO**NCAR Societal Impacts Program

IWT Workshop: Using the WAS*IS ApproachJanuary 22, 2009

• Conducted to evaluate NWS performance during significant (high-impact) events

• Usually convened just once or twice a year• Team composition: experts from both inside

and outside the National Weather Service

NWS service assessments…then

NWS service assessments…then• THEN: Inward focus on procedures, actions,

equipment– Internal review of operations in National Centers,

WFOs, CWSUs, RFCs– Informal external review of information with partners– Quantitative assessment

• Damage, verification, fatalities, injuries, etc.

NWS service assessments…now• NOW: Inclusion of external

focus via societal impacts analyses– Super Tuesday, Mother’s Day,

Midwest floods– Qualitative assessment

• Information sources, perceptions, decision-making

Desired outcomes of NWS SAs• Use findings, recommendations, best practices

to:– Improve delivery of hazard information (format,

content, media) to our customers and partners – Improve clarity of hazard information– Focus research and training– Allocate resources

Provide better and more understandable weather

information so that people will take action to protect life and property!

Impacts of the tornado outbreak

• 87 tornadoes– 5 EF-4 tornadoes– 1 tornado had a

122-mile long path

• 57 fatalities– most since May

31, 1985– 13th overall

• 350+ injuries• $520M damage

February 5-6, 2008, tornado outbreak

• National Weather Service (NWS) predictions– Excellent long lead-time info:• First outlook issued 6 days prior• Day before, outlook mentioned

“potentially strong and long-track tornadoes”

– Mixed quality short lead-time info:• Mean official tornado warning

lead time of 18 minutes• Some problems with timely

downstream warnings• Uncertainty wording for

confirmed tornadoes

18 min1 min

T

Ubiquitous questions• This was a well-warned

event, with good information…– … why did so many

people die?– … why don’t people do

what they’re “supposed” to do … to make the “right” decision?

We get frustrated when we put “good” weather information out there and people

don’t make the “right” decisions!

The “right” decision … in a tornado

• Why might someone not take shelter from a tornado?

.

.

.• What is the “right” decision? • Is there a “right” decision?• How and why do decisions get made?

Bottom line … if you want to study human beings, you’d better have a high tolerance for ambiguity!

Societal impacts scope• The task – To try to understand why so

many people died and the details of those fatalities

.

.

.Can learn so much by having people walk you through their knowledge, thoughts, actions … by letting them tell you their

stories!

• An opportunity – To gather empirical information about people’s actual warning response behaviors

Research objectives• For the fatalities, wanted to gather info

about:– Age, gender, warning received, warning source,

warning heeded, shelter sought, structure where they died, availability of safer shelter

• For the survivors, wanted to assess:– what info people had, how they interpreted it

(knowledge)– how people perceived the situation (perceptions)– what decisions people made (decision-making) – what information they had about the fatalitiesThis is a highly interdependent,

iterative process

Methodology• Semi-structured interviews with the public• Targeted, convenience, and snowball

sampling• 41 interviews by 3 sub-teams in the 6 WFOs

visited– We did 17 public interviews over 4 days in the

field, another day on the phone• Audio recorded with consent

Some of the questions• When did you first realize there was a threat of a

tornado? – How did you learn about the threat? (sources,

environmental cues) – What were you thinking after you received that information? (trust? confusion? uncertainty? barriers to

action?) – What did you do next? (confirmation?)

• Have you ever been in a similar type of extreme-weather situation in the past? (experience, false alarms)– Did anything from that experience influence what you did during this most recent event?– Have you ever been warned about an extreme weather event in the past that did not occur?

• Think back over the entire tornado event, from the time you learned there was a tornado threat through when the tornado actually occurred.– Do you feel that any of the information you received was unclear?– Is there any other information you would have liked to have had?

Data analysis• Analyzed iteratively, cooperatively by 2

coders• Coded with Excel

– Pre-determined categories – Categories created inductively during analysis

• Caveats and considerations!! – Balance between scientific rigor and rapid

operational needs– First step, hopefully leading to more related work

in the future (more in-depth studies, various weather contexts, etc.)

Findings − knowledge • People get information from multiple sources

– Majority via television– Also commonly from other people (family, friends,

neighbors, co-workers)• People get information multiple times• NOAA Weather Radio was used, but not

common• Tornado sirens are useful, but…

– Misconceptions about sirens as a warning device• Local EMs confirmed this is a problem; one is actively

trying to correct this via newspaper and radio– Misconceptions about what sirens mean

Findings − perception• Integration of seasonality, weather salience,

situational awareness about the event– Majority of people associate tornado outbreaks as

occurring in March or later…– … so many minimized threat because they

perceived it as being outside “traditional” tornado season

– … BUT, for many people, situational factors (e.g., unusually warm temps) heightened their awareness

Findings − perception (cont)

• Personalization of the threat– People often seek confirmation of the threat; a

single source of info will not necessary spur protective action• E.g., Atkins, AR, woman and couple

– Many people recognize a risk exists, but believe that their personal risk is less or that they aren’t at risk at all (optimism bias)• E.g., Hardin County, TN, family; Arkansas family

Findings − decision making• NOT a singular event• Happens numerous times throughout the

warning process• Implicit part of people’s gathering and

interpreting weather information to evaluate their risk

Findings − sheltering• Sheltering definitions– Safer = safer relative to one’s current location

(e.g., frame home is “safer” than mobile home)– Safest = a basement, storm cellar, or safe room

• Decision to shelter– Vast majority of people who received warning

heeded it and sought shelter in best location available to them …

– … but less than half of people had a “safest” shelter available

– Very few people who received a warning opted not to shelter

Findings − 57 fatalities• Collected as much good data as we could• Nearly 2/3 of victims were in mobile homes

– 15 in houses, 4 in warehouse, 1 in vehicle, 1 unknown

• Warning and sheltering info for victims– Of the 18 people we had warning info about

• 11 people heeded the warning, 3 did not, unknown for 4• 10 people sought shelter, 5 did not, unknown for 3

– … but most did not have an safest sheltering option• 8 people did, 34 did not, unknown for 15

– … and many did not have a safer sheltering option• 17 did, 21 did not, unknown for 19

Opportunities – National Scope

• Establish a ‘pool’ of societal impacts experts– Communication, sociologists, anthropologists,

economists, GIS specialists, etc.• Develop a common set of societal impacts

survey questions, tailored as appropriate

NWS should consider adding a Societal Impacts Program to operational branches of NWSH and Regions, to organize and focus

these efforts

Opportunities – Regional/WFO Scale• Gather impacts/socio-demographic data for

local events • Utilize academic community for research

initiatives– Resource for survey methods, questions,

analyses– Utilize COMET grants, NSTEP process

• Build partnerships!This workshop is a great first-step!

Broader lessons learned• Value of having some exposure to

quantitative and qualitative research• Emergency managers are invaluable• Partnerships among social scientists,

research meteorologists, operational meteorologists, broadcasters, emergency managers and other practitioners, policy makers, etc.– Building this community!– Interest and willingness to work together, to

listen, learn, exchange ideas!

Discussion questions• Would more definitive wording (call to action

statements) in warnings & statements may prompt better protective response?

• Should we continue with the watch / warning / advisory methodology?

• Should mandatory protective action be taken at longer lead times (e.g., evacuating mobile home parks, dismissing classes, large-venue considerations)?

• Should local, state, and federal governments partner (legislate?) to build local shelter facilities?

Acknowledgements• Super Tuesday service assessment team

members– Especially Mike Vescio, Daniel Nietfeld

• NWS• NCAR Societal Impacts Program and WAS*IS• Contact

– Kevin Barjenbruch (kevin.barjenbruch@noaa.gov)– Julie Demuth (jdemuth@ucar.edu)

www.sip.ucar.edu