Close Window

University of Charleston: "Improving hurricane forecast effectiveness"

Final Report

Section 1: Project Objectives and Accomplishments

The project designed and tested alternative hurricane warning messages for the Charleston office of the National Weather Service (NWS). These revised warning messages examined various ways of communicating the risks and uncertainties associated with hurricanes to the general public with the goal of increasing the public's self-protective behavior. Alternative hurricane warning messages included the use of graphics, the use of inverted strike probability maps, less use of undefined terms in NWS advisories such as hurricane watches and warnings, and the use of graphic and textual landmark references. The project also attempted to measure the public's understanding of probability and science issues surrounding hurricanes.

The project used a survey of opportunity to accomplish the goal stated above. After consultation with Dr. Robert Kennedy of the Department of Biometry and Epidemiology at the Medical University of South Carolina it was decided that a Chi-squared analysis of the count data from the survey would serve no useful purpose since the sample was not randomly collected. Instead, basic descriptive statistics, such as the percentage answering a given question in a certain manner, were used. The fact that we were unable to use a random selection of subjects limits the results obtained, but the results are quite useful as indications for future research into this topic.

Results of the research support many of our hypotheses. Over 57% of our sample did not know the definition of hurricane warning. The NWS could overcome this problem by stating there was a hurricane warning and then proceeding to state that a hurricane warning means the NWS expects hurricane conditions in the forecast area within 24 hours. As McLuckie said in "Warning---A Call to Action", be specific. Write to your audience. If the public doesn't know or understand, the warning must be written so these issues are well explained. This will enable the public to confirm the threat. This idea extends to other terms such as hurricane conditions, storm surge, storm tide, etc.

55% of our sample did not realize the main threat from a hurricane in coastal areas was from storm surge and 53% did not understand that a decrease in the forward speed of a hurricane would increase the time spent by a hurricane over their location, causing an increase in the amount of rainfall experienced. Additionally, over 32% did not understand that errors by the NWS in forecasting the forward speed and direction of hurricanes would affect the time available for evacuation (i.e. increase or decrease). These results support our hypothesis that the public often does not understand basic scientific principles and this lack of understanding could cause them to ignore, misinterpret, or underestimate a threat. We recommend writing so as to make the science very clear.

We found that a large majority of the sample preferred a graphical approach to information dissemination in contrast to the text-based approached used in local NWS advisories. This was especially true in probability issues. Over 62% of the sample answered correctly, from a map of the probabilities, that there was a 20% chance of the center of Hurricane Georges striking within 75 miles of Mobile, Alabama within 72 hours. This result was in striking contrast to a correct answer of 24.75% on a similar question using a copy of a text-based warning issued by the NWS for the probability of Georges striking Panama City, Florida. Also, in a comparison of a text-based wind advisory versus a graphical wind advisory, 74.75% of the sample preferred the graphical advisory as opposed to 11.88% who preferred the text-based product. Besides the expressed preference for the graphical approach, the strike probability results strongly support the hypothesis that the public is better able to interpret graphical products that textual products.

We also found the use of references, whether text or pictorial, aided in the dissemination of warnings. 63.89% preferred advisory B over advisory A in question 12 of the survey (see Appendix A). Advisory B contained specific text informing the public of storm surge heights relative to landmarks such Charleston's city hall. Pictorial references accompanied the text. In contrast to advisory B, advisory A was a copy of an NWS advisory and it contained no reference points, either textual or pictorial. 12.83% preferred advisory A. References to specific, well-known landmarks can help increase the perceived risk by people who reside in close proximity to the referenced location. Although local weather service offices are currently limited to text-based warnings such as the Hurricane Local Statement, a combination of text and pictorial references seems most effective. Perhaps as an interim measure NWS personnel could make pictorial references available on the local NWS web site so that media could cut and paste these references into their weather forecasts. This would personalize the information, increase the perceived risk, and confirm the urgency for taking self-protective action.

A small percentage of our sample expressed interest in Internet based information dissemination. 6.93% felt the use of an Internet based system of information was very important. 8.42% found it moderately important, 33.17% unimportant, and 45.55% did not know. It seems logical that the preferred graphical approach would work well with an Internet approach especially since the use of the Internet is growing rapidly.
The results of the survey did not support the use of an inverted strike probability map (i.e. the probability of a hurricane missing a location versus the probability of a hurricane striking a location). 42.08% did not support this alternative while 37.13% felt it could prove useful. The chief concern expressed by many centered on possible misinterpretation of the graphic since the information has historically been presented as the probability of a hurricane striking a location. Any attempt to use this method would require a large investiture in education to decrease any chance of the public misinterpreting the inverted probabilities.

Section 2: Summary of University Exchanges with the Charleston Office of NWS

The research project was initially conceived by Dr. B. Lee Lindner of the Department of Physics at the University of Charleston, South Carolina and Stephen Brueske, Science and Operations officer at the Charleston Office of the National Weather Service. As a graduate student in Environmental Studies, Mr. Charles Cockcroft was brought into the project in May of 2000. In preparation for the research Mr. Cockcroft met several times with Mr. Brueske at the Charleston NWS office. Mr. Brueske introduced Mr. Cockcroft to the various tools he had at his disposal for the dissemination of severe weather warnings. Mr. Cockcroft was also provided samples of previous hurricanes warnings issued by the Charleston NWS office. Discussions at the time provided Mr. Cockcroft with a feeling for the direction in which Mr. Brueske felt the college should proceed. The discussions addressed alternatives to current text-based hurricane warning packages as well as incorporation of Geographic Information Systems, pictorial reference points, placement of educational materials about hurricanes on the Charleston NWS web site, and inverted probability maps.

Mr. Brueske and Mr. Jerry Harrison, Warning Coordination Meteorologist and NWS liaison for Project Impact, met with Dr. Lindner, Mr. Cockcroft, and four undergraduate interns: Rhonda Crawley, Kristi Owens, Jay Sonner, and Blake Williams, and discussed the overall plan of the project. Mr. Harrison suggested contacting Ms. Joni Rennhack, Charleston County coordinator for Project Impact. This contact led to the addition of a ten-part question, provided by Ms. Rennhack, which measured the public's preference for dissemination of hurricane educational materials. The results of the survey have been made available to Ms. Rennhack and the Charleston County Project Impact Office.

Mr. Brueske and Mr. Harrison provided additional help by putting the students in contact with previous research investigating public response to emergency warnings. This helped with the development of their understanding of the theoretical basis established by previous research.

Section 3: Presentations and Publications

Mr. Cockcroft is finishing a thesis entitled "Increasing Hurricane Warning Effectiveness". A copy of this will be forwarded to the Comet Outreach program upon completion. In addition, Mr. Brueske and Mr. Harrison are writing a paper that summarizes the results of this study. This Summary will be provided to NWS Eastern Region Headquarters for further distribution throughout the NWS.

Section 4: Summary of Benefits and Problems Encountered

4.1 The collaboration between the University of Charleston, South Carolina and the Charleston office of the NWS provided many benefits to both parties. Students received an inside view of the problems, tools and constraints that are an everyday part of the life of NWS personnel. Mr. Cockcroft had an opportunity to participate in some original research and experience the challenges involved in the design of a proper measurement tool, the organization of a group of undergraduate interns, and the collection of meaningful data. The Charleston Office of the NWS has received results that indicate possible alterations to their hurricane warning products.

The university benefited enormously by having five students involved in research related to this grant. Also, this project significantly enhanced the visibility of the growing meteorology program here at the College of Charleston. This project also involved significant community outreach, and established many links to governmental and non-governmental agencies involved in hurricane related matters.

The main problem that Mr. Cockcroft encountered revolved around the timeframe for the project and the original scope of the research. Mr. Cockcroft finished his spring semester in early May of 2000 and immediately had to design a survey and organize four undergraduate interns. Mr. Cockcroft had the use of these interns from the middle of May until the end of July. The initial goals of the research indicated they would design a survey to test alternative warning packages, randomly sample the coastal South Carolina counties of Berkeley, Charleston and Dorchester, and carry out a statistical analysis of the data. Due to time and personnel limitations it was necessary to abandon a random scheme for the collection of the data. At the suggestion of Dr. Robert Kennedy of the Department of Biometry and Epidemiology at the Medical University of South Carolina, one of Mr. Cockcroft's thesis advisors, the random sampling scheme of data collection was changed to an opportunity sample scheme. The failure to randomly sample has profound impacts on any statistical inferences that can be drawn from the research but this method did allow collection of data from 202 subjects. An analysis of the data using the percentages of subjects answering a question in the same manner has identified areas that are ripe for further research.

In the future, the logistical problem could be avoided by starting earlier in order to make full use of the interns in the random collection of the necessary data. In addition, the original research proposal stated students would randomly collect data from several hundred subjects. In retrospect, problems of time, coordination of intern schedules, and the use of two person teams for safety reasons made this random collection difficult over this brief period. Using statistical methods it is possible to calculate the required sample size, n, to achieve the desired level of confidence in the results. This would be useful in that it would conserve valuable resources (i.e. the time of the interns) and it would allow the random collection of the required sample size. Future research should carefully consider this to avoid the problems I encountered.

4.2 Summary of benefits to the NWS

The research supported by this COMET project has provided invaluable information to the Charleston NWS office, and has initiated changes to hurricane-related statements and warnings issued by the office. In addition, the research provides direction for the further development of new and innovative hurricane-related products that the Charleston NWS office is currently experimenting with. Results of this research will also be summarized in a paper that will be submitted to NWS Eastern Region Headquarters for distribution to appropriate NWS offices throughout the NWS. Finally, the project has enhanced cooperation between the NWS, the University of Charleston, and the Charleston County Project Impact Office and indicated ways in which each of these entities can better disseminate hurricane educational materials.

Results of this study clearly indicate that many residents of Charleston County remain uncertain of the definition of a hurricane watch or warning, despite the fact that 95 % of the county evacuated for Hurricane Floyd in 1999. Therefore, the Charleston NWS office has implemented a policy that includes a clear definition of a hurricane watch and/or warning into products that mention a hurricane watch or warning. For example, every Hurricane Local Statement produced while under hurricane warning conditions will include a statement indicating "a hurricane warning means that hurricane conditions are expected along the coast within 24 hours".
The survey also indicated that a majority of respondents did not realize that the main threat from a hurricane in coastal areas was from storm surge. The Charleston NWS office will clearly indicate this in hurricane local statements buy stating "the most dangerous threat to residents along the immediate coast is the storm surge which can destroy strong structures, flood roadways, cut off escape routes, and prohibit emergency personnel from rescuing those remaining in flooded areas."
The research indicated that referencing specific impacts to well-know landmarks aided understanding of threats. The Charleston NWS office will use specific descriptions of potential water levels over well-known landmarks in the local area to describe potential threats. For example, "storm surge from this hurricane could produce 3 feet of water and large waves across all of Waterfront Park in downtown Charleston, as well as flood roadways within 2 blocks of the waterfront to a depth of 2 feet."

The research clearly indicated that a graphical approach to threat dissemination was preferred and that a graphical approach dramatically improved comprehension of potential threats. The Charleston NWS office is creating graphical representations of the potential water depth at numerous local landmarks for hurricanes of various intensities and will post them on the office's website. In addition, the Charleston NWS office will notify local television meteorologists of these graphics and brief them on the results of this research so that the local media can incorporate the recommendations from this study into their products and procedures.
Finally, results of this study have provided information that allows the NWS, University of Charleston, and the Charleston County Project Impact Office to more efficiently disseminate hurricane educational materials to the general public.

This project has greatly benefited the Charleston NWS office and has provided direction for the further development of new and innovative hurricane-related products and services. In addition, summarized results of this research will be submitted for distribution to all NWS offices. The project has also enhanced cooperation between the NWS, the University of Charleston, and the Charleston County Project Impact Office, and identified ways in which each of these institutions can better serve the public.

Link to Survey on Hurricane Warning Effectiveness