The communication by forecasters of tropical cyclone (TC) descriptions and forecasts to user communities necessarily involves the transmission of information based in science to different classes of users composed primarily of nonscientists. Inherent in the problem is the necessity of translating or converting the scientific content of the forecast, including its associated uncertainty, which is mathematical and statistical in its native structure, into restructured content comprehensible to populations not generally schooled in those disciplines. The forecast interpretation problem encompasses not only the forms in which the information is presented or communicated (e.g., text versus graphics), but even more so the complexity and transparency of the scientific content contained between those forms. This article investigates the substantive areas of dissonance and disconnect between the scientific content of TC descriptions and forecasts, including the uncertainty, and the ability of end users to accurately comprehend and interpret the information. It centers on the three storm attributes for which there is a forecast, namely intensity, track, and size, within the context of existing research studies, public surveys, and original official documents that specifically provide insights into this subject matter. The results suggest that the TC descriptions and forecasts, once their scientific substance has been processed for the benefit of nonscientists, still require some preexisting scientific knowledge that may or may not be present among the different groups of nonspecialist users.
Warning members of the public about impending tropical cyclone (TC) threats to their communities necessarily involves the communication of TC descriptions and forecasts, and their associated uncertainties, based in science, to audiences composed primarily of nonscientists. Since the data cannot be communicated in raw form outside the discipline, what is actually disseminated is information about what meteorologists know—and what they do not know—about the current and projected state of a specific TC at a given time and the conditions it is likely to produce when it intersects a land area. Although still based in science, and meteorological in its composition, this information is intended for nonscientists to use in formulating responses to the physical dangers posed by the TC; all of its elements, taken together, compose the TC descriptions and forecasts that the public sees.
Successful implementation of the response therefore depends on the accuracy of comprehension and interpretation by nonscientists of this information, no less than on the accuracy of the forecast itself. It does not depend on their understanding it at the depth a scientist would, or even in the way a scientist would, but in a way that enables them to draw the same conclusions about the storm threat that scientists intended them to have when they released it, rather than alternate conclusions, and in a way that is not inconsistent with the knowledge of modern science.
Understanding and accurate interpretation of the informational content is only a first step, but it is a vital one. While accurate interpretations have the potential to elicit, though cannot guarantee, accurate assessments by nonscientist audiences of the levels and types of danger posed to them by any given TC, incorrect interpretation of this information poses a fundamental impediment to a successful public response.
At the national level, the state of public understanding of science-based concepts is a moving target relative to the rest of the industrialized world, but it has been trending mostly downward (Program for International Student Assessment 2004, 2007, 2010). Currently the United States is 17th in scientific literacy and 25th in mathematical literacy, in a study of 34 industrialized countries. Part of literacy as defined in these studies is the ability to make decisions using science-based information; the comprehension of uncertainty and probability, which belong to the statistical disciplines, falls under the category of mathematical literacy.
This article addresses the subject of whether and to what extent the TC descriptions and forecasts that are communicated to the public, information of meteorological origin that has been reformulated for public consumption, nevertheless still requires, in light of the available evidence, some a priori familiarity with scientific concepts and modes of expression to be accurately comprehended. This is particularly important with regard to the official information, which is the source of the public forecast information and thereby defines its structure and organization.
Although the publicly disseminated TC descriptions and forecasts were initially directed primarily toward technically trained user communities—such as emergency managers and members of the media—to help them elicit the desired responses by members of the general public during TC emergencies, the existing body of research suggests that some members of the general public are not just taking the general response instructions at face value but are attempting to interpret the meteorological information provided by forecasters more directly.
Across studies, respondents surveyed from among members of the public in different hurricane-prone regions are consistently ranking meteorological factors, both current and projected storm attributes, as among the primary factors in their decision (e.g., Howell and Bonner 2005; Zhang et al. 2007; Morss and Hayden 2010; Dow and Cutter 1998; Morrow and Gladwin 2006; U.S. Army Corps of Engineers 2010a–d). In some but not all cases, respondents are assigning the meteorological factors a higher importance ranking than the advice of emergency management officials. They are not merely relying on what public officials tell them, but are sometimes interpreting the TC descriptions and forecasts more directly and reaching their own conclusions about the nature and scope of the dangers posed to them by approaching storms. The fact that members of the public are interpreting meteorological information initially designed for technically trained user communities makes it all the more important that forecast products and messages distributed by the meteorological services be vetted by social scientists before they are issued, because members of the public have to be in a position to respond to the information coming to them from meteorologists, regardless of the extent of their own scientific knowledge.
The full value of the tropical cyclone forecast, its associated products, and scientists’ knowledge about a given storm is preserved so long as the knowledge received by the public is the same as what was transmitted, with little to no loss of signal. To the extent that the knowledge acquired by the recipients of the TC descriptions and forecasts is different in its meaning or significance from the knowledge forecasters intended to convey, there is an informational disconnect, and the transmission is compromised, resulting in some or all of the value of the knowledge scientists have about a given TC threat being lost to those members of society. This is a problem of deep concern to meteorologists, because the information on which the disconnect exists is meteorological in origin. Addressing the informational disconnect as it is now, including a deeper understanding of the ways in which it manifests itself through specific types of substantive misunderstandings on critical points (e.g., uncertainty), can help illuminate potential directions in which adjustments to the existing information could be usefully made, with the ultimate objective of curtailing misunderstanding and ensuring that the full value of the TC forecast is realized for the benefit of the society at large.
While forecasters are able to interpret and give meaning to their meteorological observations through the context of the collected knowledge of modern science, and the operational models based on it, to make their forecasts, members of the public do not, as a practical matter, have access to the complete body of knowledge. They have only the tapestry of their existing knowledge to rely on in making sense of their own weather-related observations, which are the TC descriptions and forecasts they see. To the extent that they do not possess the preexisting interpretive contexts required to assimilate this content or that the interpretive contexts they have available are in whatever way inconsistent or incompatible with the collected knowledge of modern science, their interpretation of the TC descriptions and forecasts may not be as intended by forecasters. As a result, their conclusions about the dangers posed to them by a given storm, through no fault of their own, may not be consonant with the objective physical realities.
2. Methodology and data
A spectrum of published academic research studies, public surveys, hurricane readiness questionnaires, and original documents going back through approximately the past decade was collected and reviewed. Data explicitly providing insights into the current state of comprehension of the TC descriptions and forecasts were extracted and analyzed, both separately and together, in light of one another, with particular emphasis on the information pertaining to intensity, track, and size.
Among the studies examined were poststorm and hypothetical storm surveys and polls conducted, sponsored, or commissioned by research meteorologists, social scientists, universities, the National Weather Service (NWS), public service entities, and private corporations. In addition, the series of NWS poststorm service assessments was reviewed in its entirety, and the official observations and findings concerning interpretation difficulties that have occurred and/or were believed to have interfered with the public response during specific storm events were extracted. U.S. Army Corps of Engineers poststorm behavioral studies were reviewed for similar data, and National Hurricane Center (NHC) requests for comments and Tropical Cyclone Reports were consulted. Finally, the events that took place during the approach of Hurricanes Ike (2008) and Charley (2004), as two conspicuous hurricane response failures by the general public during the past decade and pertaining to intensity and track forecast interpretation, respectively, were reviewed in their fine details.
While the data suggest that the TC descriptions and forecasts are partially understood, the study found that they are sometimes subject to interpretations other than those that would be correct for a given instance, and which are often inconsistent with the science that produced them. An accurate understanding of what a TC description or forecast means, and being able to realize its full value, is not just a matter of the presence or absence of factual scientific knowledge on the surface, or in its literal sense. It is just as much a matter of knowing that a certain fragment or piece of general scientific knowledge or context is applicable to the new information for purposes of ascertaining its intent and significance, comprehending its second-level implications and consequences, and relating it to a specific and unique real-world situation. Difficulties in understanding the information as evidenced in the source data were found to be present at multiple levels, starting from the bottom up:
some unfamiliarity with basic physical and mathematical relationships, especially those concerning time, space, change, and motion, and with the broader physical context that frames the TC descriptions and forecasts;
spatial–temporal problems, including interpretations that run contradictory to the natural arrow of time, that employ spatial perspectives and priorities inconsistent with those assumed in the information provided, and that have some lack of clarity with respect to time sequences and distinctions;
some difficulty in sparking access to general knowledge of scientific and mathematical significance that may already be known, drawing upon it, and applying it to the meteorologically derived information presented, to unlock the new knowledge conveyed in the specific TC description or forecast;
problems decoding the data packaging, whether text or graphics, including inaccuracies in the identification of its constituent parts and extending to the whole, which for nonscientists often requires a multistep process;
insufficient context through which to discern the intent and purpose of the forecast products; and
insufficient context through which to ascertain the implications or significance of the new knowledge, and apply it to a unique situation.
This section explores and analyzes results concerning TC descriptions and forecasts of current and projected physical storm attributes and the associated uncertainties, specifically, information that is conveyed about storm intensity, track, and size.
Much of the documented misinterpretation of the intensity-related information, conveyed primarily through the Saffir–Simpson (SS) hurricane wind scale (formerly the SS hurricane scale), reveals a contrast of perspectives and priorities between forecasters and users. A specific focus was found to be present within several classes of users on the forecast for intensity at landfall, a forecast that technically does not exist. In a multiple-choice poll question asking respondents from the general public which type of forecast they believed was least reliable, “storm intensity at landfall” was available as the correct answer (National Hurricane Survival Initiative 2010). Moreover, 75% of respondents did not choose that answer, opting instead for answers pertaining to the track forecast, including storm direction (36%), storm duration (19%), and forward speed (20%), demonstrating a great degree of confidence in this forecast for intensity at landfall as compared to different aspects of the track forecast. The distance from focus to expectation is not great: indications are that an expectation may exist, even among some emergency managers, to be informed of a storm’s intensity at landfall (NWS 1996).
This manner of interpreting the intensity forecast information, specifically with that focus, is a problem of sequential logic. Holding out landfall as the main focal point effectively deemphasizes the normal step progression of the sequential forecast lead times, which begins, as does the forecaster, with the storm’s current condition out over the ocean and looks forward with its projected motion, through space and time, toward land. The landfall-centered interpretation instead jumps outside the present time to look at the projected intensity at a single arbitrary forecast lead time, when the storm is forecast to intersect land. Following naturally from a perspective that is consistent with the intended time sequence, starting in the here and now and following the natural arrow of time, is the potential to note whether the forecast trend is projecting intensification, deintensification, or persistence of the storm’s current intensity, together with the gradual, sequential broadening of possibilities over time, which is the uncertainty.
Interpreting the intensity forecast through the lens of landfall takes the focus away from noting which lead time that particular forecast pertains to, how many hours or days out it is from the storm’s current intensity estimate, and how much uncertainty coincides with that lead time. Since the uncertainty becomes greater with the progression of the sequence, and since it is tied to the specific lead times, the focus on the uncertainty term itself may also be degraded. Furthermore, without it, a forecast near landfall may become subject to being deemed right or wrong, black or white—a hurricane was forecast to make landfall as an SS2 but actually did so as an SS4, and therefore the forecast was wrong—instead of seeing the forecast in a more fluid way, as a changing set of approximations about a changing storm within a changing atmosphere. Meteorological services could work more closely with the media to help guide audiences sequentially through the series of forecast lead times, to focus their attention on the intensity forecast trend instead of intensity at landfall, and to show them how the uncertainty in the intensity forecast becomes greater with distance from the present.
The tendency to focus on landfall, however understandable, reflects the population’s location-centric perspective. Yet the intensity forecast itself, like the track forecast, contains within its spatial–temporal structure and sequence the storm-centric perspective of the forecasters (Fig. 1). The location-centric interpretation of the storm-centric forecast carries deep implications that may be logically inhibitory to the assimilation of uncertainty for both track and intensity forecasts, because of the way it handles time, space, change, and motion.
Difficulty assimilating the passage of time as context for a changing storm may also allow some nonscientist users to conflate the distinction between the storm’s present conditions and its future conditions. For example, “They [emergency management] apparently focused on an earlier description of Marilyn as a small category 1 hurricane and were not prepared for a direct hit from an intensifying storm” (NWS 1996). These users had effectively frozen the hurricane in its current condition and adopted that description as their expectation for the future. While they could see the storm moving forward in space, toward their location, they had not accounted for it also changing through time. No sharp distinction was made between intensity in the present and in the future, with the hurricane’s current description and its uncertain future all rolled up into one.
A second, intersecting problem lies in the interpretation of the SS wind intensity categories as all-purpose danger–decision thresholds, as is suggested across regional studies of general users in addressing actual and hypothetical TC threats (e.g., Whitehead et al. 2001, 2000; Howell and Bonner 2005; Morrow and Gladwin 2006; Zhang et al. 2007; Morss and Hayden 2010), as well as in past storm events. This problem is manifesting itself through the appearance of effective “categorical evacuation thresholds,” which might be used to describe situations in which people believe they should categorically evacuate for a storm of one SS category but not another. It intersects with the timing problem because, in the hypothetical cases, the distinction is almost never made as to whether the category under consideration is the storm’s current category, which is known, or its future category closer to land, which is unknowable at the time of decision, but for which there is a forecast. A tropical cyclone’s current intensity at the time of decision could be and often is different from that later experienced at the location, even dramatically different, because of the possibility of rapid intensification during the interval.
The interpretation of the SS scale as a generalized danger scale cannot logically account for the dangers posed by a unique low-category storm. It also allows the intensity categories to be used by themselves as a basis for numerical comparison of the anticipated effects of different storms with different wind structures on different approaches into different locations, or into the same location. Powell and Reinhold (2007) observed, with respect to SS3 Hurricane Katrina, that “regardless of warnings well in advance, some people did not evacuate because their location was known not to have been flooded by Hurricane Camille, an SS5 storm that devastated the area in 1969.”
The response failure by some members of the general public during Hurricane Ike was largely attributed to its low SS category, an intensity analog to Hurricane Charley and the “skinny black line” (center-track line). During its approach, Hurricane Ike, characterized by an extraordinarily large wind field, was projected to produce catastrophic surge heights in some areas along the Texas Gulf coast, but it carried a low, SS2 designation, which it retained right up through landfall. Dire warnings were posted, culminating in the now-famous “certain death warning” issued directly by the local NWS forecast office at Galveston (NWS 2012) in an only partially successful effort to communicate the extreme level of danger this hurricane posed to the area. While more data are needed, this most recent response failure has caused the light of urgency to shine on the more general misunderstanding about what the SS scale is and what it is not, what it conveys and what it does not, and what its purpose is; its purpose is certainly not to get people in danger to stay home for a tropical cyclone emergency of major proportions.
The episode argues for a means of warning the public more explicitly about the level of danger posed by storm effects, especially that posed by storm surge. At the same time, just as it was difficult for some residents during Charley to reconcile a center-track line pointing at Tampa with a landfall at Punta Gorda (NWS 2006), it could be similarly difficult for others to reconcile a TC bearing the SS2 label with one capable of producing extreme surge heights such as those witnessed during Ike, unless there is clarity about the meaning and significance of the SS2 label.
Stewart (2007) reported that only 32% of his college-level respondents believed they had “much” or a “complete” understanding of the SS scale. Understanding what kind of information the scale conveys is prerequisite to the accurate comprehension of the information itself. There is a central distinction between storm attributes and storm effects, as two separate kinds of information. Storm attributes are those changing characteristics that belong to the TC itself, that are internal to the storm and specific to it (e.g., maximum sustained wind speed); storm effects are physical phenomena produced by the TC at the location (e.g., storm surge). Information about storm attributes embodies the storm-centric perspective of the forecasters, while a focus on the level of danger posed by storm effects reflects the location-centric perspective of the general public.
The generalized danger-scale interpretations of the SS scale, together with the resulting thresholds and implied comparisons of individual TCs, represent a location-centric interpretation of a storm-centric forecast, such that the overall danger posed by a unique TC, which applies to the storm effects, may be read into an intensity forecast (projected SS category) that is actually conveying information about a storm attribute. It is the individual storm’s combined effects, not its internal characteristics, that constitute the immediate source of the danger, that pose the threat to the community, because it is they which directly cause the destructive impacts to life and property. The SS scale does not directly address tropical cyclone impacts, as these are controlled often independently by five TC effects, namely, wind, rainfall flooding, water rise flooding, high surf, and tornadoes.
Finally, danger thresholds based on SS categories were found to vary by region, without any scientific justification. The balance of perceived safety for coastal Louisiana respondents, members of the general public, seemed to break at SS4 (Howell and Bonner 2005), while in the greater Galveston area of coastal Texas the responses to the hypothetical evacuation by category question, also from members of the general public, clearly flipped from “no” at SS2 to “yes” at SS3 (Zhang et al. 2007).1 Hurricane Ike was the next SS2 to strike the area, and the study’s finding effectively predicted the outcome, that an SS2 hurricane would not meet the threshold of danger for a significant portion of these residents. A subsequent study conducted after Ike’s passage has bar graphs showing 30%–40% of respondents in that same area still saying categorically “no” to evacuation from a future SS2 hurricane (Morss and Hayden 2010). While 40%–50% answered “yes” to that question, 10%–20% chose the “don’t know/depends” option that would be indicative of greater understanding.
The scale is, in practice, being widely interpreted as a generalized hurricane danger scale, something the public obviously wants, but which does not now exist. Of the major TC effects, the scale is able to consistently communicate the danger posed by only one of those: wind. The related scientific context for accurate interpretation of the SS categories, then, is that every storm is different, that statistical outcomes do not form physical constraints or “laws” that bind individual storms, that the scale cannot account for all storm effects, and that current descriptions of a tropical cyclone’s intensity must not be confused with the storm’s condition in the future when it actually reaches the location.
Documented misinterpretations of the official track forecast cone (NHC 2004; Broad et al. 2007), or cone of uncertainty, have in common the problem of misidentification of the cone’s core components, and of the product as a whole. Some of the misinterpretations place its sequenced components into an inaccurate temporal context. For example, in the NHC’s 2004 request for comments on its three possible future versions of the cone (NHC 2004), a respondent wrote of the version containing the center-track line (NHC cone option 1): “this shows the past and forecasted path of the storm.” Almost all of the documented misinterpretations involve the misidentification of the shaded area of the cone, a nonrecognition of the cone’s component circles both as representations of uncertainty and as distinctly separate entities.
Some cone misinterpretations, documented not only formally but also known informally in weather forecast offices throughout the southeastern United States, associate the shaded area with something concrete, often a physical storm attribute or effect, rather than with uncertainty radii, which is abstract. These interpretations center around some combination of the size–wind field and danger zone–damage swath interpretations, for example, “the actual area that may be affected,” “the danger zone,” and “EXPECTED IMPACT AREAS,” as a few important examples (NHC 2004, all caps in original). Even if the prerequisite knowledge that a 24-h forecast is more reliable than a 72-h forecast exists among the general public, as is suggested in the data (Morss et al. 2008; Zhang et al. 2007; Morss and Hayden 2010),2 it does not necessarily mean that this knowledge will be applied automatically to cone interpretation. Indeed, it cannot occur directly, because there is an intervening step: recognition of the cone’s constituent circles and the forecast points within them, thereby bringing those 24-h, 72-h, and other forecast lead times into view.
The practice of mapping and contouring abstract mathematical and statistical concepts, among them uncertainty and probability, is common within the scientific disciplines: among scientists, comprehension is automatic and second nature. For those unfamiliar with it, a conscious, two-step process may be required. Despite the cone’s ubiquitous presence in the public arena, available evidence points to interpretations that appear to take the data packaging at face value and that do not undertake the two-step process to “unzip” the data packaging of the cone and reveal its contents.
Doing so enables a user to “see” in a typical 3-day cone not a single, blended object, as in the documented misinterpretations, but a composite image consisting of six distinct objects: one current estimated storm position and five separate forecast objects, each pertaining to an increasingly distant lead time. In addition to the uncertainty, the passage of time itself is a second abstraction in the cone, reflecting the value scientists place on efficiency (compactness, nonrepetition) in the presentation of data. As a multiday forecast, it compresses multiple, sequential time frames in which storm motion is forecast to occur into a single still object. In most instances, the component circles are themselves overlapped to a lesser or greater extent, leading to even greater compression.
In some track forecasts, there may be significant to near-total overlap of the component circles. This is shown in Figs. 2a,b, respectively, which is a representation of the two examples of the NHC’s cone option 3 (NHC 2004). A forecast for a stationary or slow-moving storm or a circular direction of travel will result in the high degree of overlap of the kind portrayed in the second example (Fig. 2b), such that the cone appears on the surface to have lost its familiar cone shape. Broad et al. (2007) noted that one respondent in the NHC request for comments had referred to the (b) examples collectively as “cluttered” and “confusing looking,” not recognizing them as alternate examples of the same three options. The main factor that differentiates the (b) examples visually from their (a) counterparts, however, is precisely the degree of overlap in their component circles.
Separating out the cone components and noting their sequential arrangement may be automatic for the scientist, but nonscientist users do not necessarily have enough information available to ensure the successful uncompression of the data. Indeed, the time sequencing aspect may not be automatic for a population focused primarily on where the TC is forecast to intersect a particular landmass rather than what it is doing out in the ocean, as the track analog of the “intensity at landfall” interpretation. However, as with the intensity forecast and the growth of the uncertainty in the forecasted SS categories over time, accurate time sequencing means starting with the current TC, wherever it may be, and moving with it in a forward direction through time and space. At some level, understanding the track forecast cone almost requires the adoption of the forecasters’ storm-centric perspective for purposes of interpretation, because that perspective is built into the structure and logic of the cone itself.
Seeing the cone as a time series that starts with the current estimated position of the storm, and moves through time and space into an increasingly distant and therefore uncertain future, gives meaning to the ballooning of possibilities across time, to the progressive enlargement of the circles—and therefore to the enlargement of the uncertainty—with increasing forecast lead time. It is for this same reason that the current trend of doing away with the center-track line has had the effect of making the entire cone product more transparent, not less. The uncertainty is an abstract concept represented spatially and temporally in the official cone by circles of increasing radii; the circles themselves are embedded within the shaded area and not directly visible in the cone. Once made visible, as in NHC option 3 (Fig. 2), it becomes possible to determine if there is true understanding: one respondent in the request for comments made reference in this context to “meaningless circles that overlap each other” (emphasis added). Since they are its component parts, the circles are the cone, and therefore the comprehension of the circles is necessary for the comprehension of the cone. According to another respondent, option 3 is “a bit harder to follow because the circles overlap so often and it could get a bit confusing.” The respondent went on to say that option 2, the version in current use, “is more clear about which areas could be affected by hurricane conditions” (NHC 2004). Of course, the cone says nothing about the storm’s wind structure, and here there is no substantive understanding.
It is not only the data packaging, but also some of the context giving rise to the uncertainty that may not be well understood. The constantly changing, shifting, dynamic nature of the steering currents across time is an important part of track forecast uncertainty, and constitutes some of the background context for short-term changes in the track forecast. The steering currents in which TCs are embedded, and the surrounding atmosphere in its more global sense, are changing all the time, with forecast model guidance being continuously updated to accommodate these changes. Emphasizing the existence and continuously changing nature of the steering currents provides an explanation for members of the public as to why some uncertainty in the track forecast is inevitable. Such information can help to prepare them to expect short-term changes in the track forecast as a routine matter, including the incremental repositioning of the track forecast cone, thereby enhancing forecaster credibility.
The absence of background context for the tracks of hurricanes more generally also manifests itself in the nonlinguistic aspects of the misunderstanding surrounding the “100-yr storm” concept (NWS 1996; Myers 2007; Schleifstein 2008). This may not be solely a semantic problem about the misinterpretation of problematic language, though the language itself certainly complicates matters. It suggests to some nonscientist users a predictable long-term periodicity where none exists, and it implicitly conflates TC attributes with effects: it is unclear whether the label 100-yr storm applies to attributes of the storm itself or to the damage caused by its effects at a location. Equally important, however, is the fact that a TC might track in a certain direction and therefore strike one location as opposed to another because it is being steered that way in the moment, not because the location is itself on a predetermined schedule to receive a catastrophic hurricane strike after the passage of a fixed interval of time (100 yr).
Finally, evidence was found by Lindner et al. (2004) of some difficulty in understanding the significance of a decrease in a tropical cyclone’s forward speed; some 53% of their sample, taken from members of the general public, provided an incorrect answer to this question. Forty-seven percent knew that it was “more total rainfall from a slower moving storm.” The next largest percentage, 36%, said that a decrease in forward speed would result in “higher wind speeds and enhanced storm surge,” and the remainder responded with other incorrect answers. Fewer respondents, only 32% of that sample, did not know that errors in forecasting the forward speed would affect the time available for evacuation. The existence of prior knowledge about the relationship between time, speed, and distance does not necessarily mean that it will be applied automatically to the forward speed information, that the intended conclusions will be drawn from it, or even that the forward speed information will be recognized as having practical implications. The authors of that survey concluded that “[t]hese 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.”
The size of a TC has been defined for internal forecasting purposes as either the radius of the gale force winds or the radius of the outermost closed isobar (ROCI); both are estimated by the National Hurricane Center (Knaff and Zehr 2007). Since the gale force winds are the same as the 34-kt or tropical storm–force winds, descriptions and forecasts pertaining to the 34-kt wind radii may be used to convey information about storm size. This information about the storm’s wind structure is contained within the official forecast advisories, given in tabular form in units of nautical miles. It includes the estimated and forecasted extents of both the 34- and 64-kt winds, those of tropical storm and hurricane force, respectively. There is a current estimate, but no forecast, for the ROCI, and it is not included in the advisory, as it is not intended for use by the public. The forecasts for the 34- and 64-kt wind radii, made up through 72 and 36 h, respectively, are subject to large errors and are difficult to verify. A graphical version of the estimated radii, but not the forecast, is available in the form of the tropical cyclone wind field product.
There is a relative sparsity of data about the public’s interpretation of descriptions and forecasts about storm size, about how the public conceives TC size or its significance, or much indication that size is even on the public radar. There is evidence, however, that there is not always a clear demarcation line between TC size and intensity, or knowledge that size and intensity are two different, independent storm attributes that may combine in different ways: “They [the media] also assumed that since Marilyn was a compact storm, much smaller than Hugo or Luis, it would be a fairly minor event” (NWS 1996).
Complicating matters is the fact that size descriptors were found to be present in (nonofficial) public educational information and broadcast media reporting in contexts pertaining to storm intensity.3 Terms such as “large” have been misapplied to hurricanes with high maximum sustained winds, which is intensity; “large” is actually a descriptor denoting a TC with a wind field of great spatial extent, which is size. Similarly, terms such as “growth” have been misapplied to the transition of TCs from one status to another (e.g., tropical depression to tropical storm) or from lower to higher SS categories, which is intensification; “growth” is actually a descriptor denoting the expansion of a TC’s wind field (Merrill 1984), which is size.
It has become apparent in the wake of Hurricane Ike, as well as from Hurricane Katrina, that clarity about the meaning of storm size is important in its own right because of the contribution of size to surge heights, and because of the connection of the 34-kt wind threshold to evacuation timing and deadlines. Preexisting knowledge about the significance of Hurricane Ike’s extraordinarily large size—whether in relation to the potential surge heights or to the early arrival of the storm’s 34-kt winds relative to its center—was not in evidence during the approach of that storm. On the timing issue, some residents in the Galveston area stated after the fact that they had believed they could wait another 12 h, until the following morning, to confirm the threat as the social science research indicates people often do, and delayed their departure. By then it was too late, as water rise from high waves that had raced to shore ahead of the strong winds, surge, and storm center had already overtaken the area and blocked safe exit routes (as seen in The Weather Channel program Storm Stories).
Although the TC descriptions and forecasts are partially translated for the use of nonscientists, the study concludes that the TC descriptions and forecasts distributed for the benefit of nonscientist users still have components that are technical, or that are “using terminology or treating subject matter in a manner peculiar to a particular field.”4 The information still contains content that requires specialized knowledge and close ongoing familiarity with the subject matter and its background context to be understood as intended by forecasters.
The NWS concluded after Hurricane Charley that “[e]ducation on hurricane products needed improving, particularly with regard to the forecast track cone of uncertainty” (NWS 2006). Given the confusion surrounding the cone, future studies on the interpretation of the wind speed probabilities products, which the NHC began issuing in 2006 in Charley’s aftermath, would also be useful. The NWS recommendation with respect to education could certainly be extended beyond the track forecast to include the intensity forecast and the SS categories, and the significance of storm size. However, emergency managers, who are mostly nonscientists, already receive specialized training instructing them in the interpretation of the TC descriptions and forecasts, which includes the technical definition, or identification, of their core components, and in some of the background context, which serves as the inferential reference frame. This context, against which conclusions may be drawn from the information provided by forecasters, is broadly summarized as the interrelationships among, and the making of distinctions between, the elements within four basic groupings: 1) TC attributes (e.g., direction, forward speed, intensity, size, others), 2) framework attributes (e.g., time–change, space–distance, motion–steering, others), 3) land attributes (e.g., bathymetry, topography, orientation–angle, latitude, others), and 4) effects of TC–land interactions (e.g., storm surge, rainfall–flooding, high winds, tornadoes, others). The realization that elements both within and among these basic groupings will combine in different and unique ways every time a TC approaches enables one to conclude that every storm, every location, and therefore every storm event, is different from every other.
Although emergency managers are more highly specialized in this subject matter than are members of the general public, even they have still on occasion been susceptible to the same types of misinterpretations as other nonscientists, as have some members of the media. Rather than attempting to engage in the training or education of large numbers of general users, members of the general public might benefit more from information developed specifically for them: information that comes directly from the hurricane experts to maximize its credibility in their eyes, but that does not require any special explanation, education, or training to understand it. Alongside the ongoing refinement of existing information products, a possibility that might be worth considering is a parallel set designed, from the ground up, for general, untrained audiences.
Either way, while the screening of existing products and communications by social scientists prior to issuance is important, there is a case to be made for increasingly bringing them into the process earlier than that—most usefully into the initial design phase. This would help forecasters to ensure that information products being utilized by the public, by design, by way of their underlying internal structure, would increasingly project an epistemology generalizable to all audiences, and reflect to a correspondingly lesser degree the manner of pursuit and conveyance of knowledge specific to expert communities.
Refinements to the current information aimed at decreasing the need for subsequent education and explanation may be approached by deliberately building some of that education directly into the products themselves. Education as to what the components of an information product are, and equally importantly, what they are not, may be approached through the use of labels directly affixed to each component. This would avoid the use of legends and similar devices, as these techniques are specific to the scientific disciplines and not all users may go through the additional step of referencing them. Broad et al. (2007), in discussing the track forecast cone for a meteorological audience, placed labels on the components identifying each of them; if this was done for those meteorologists specializing in subject matter other than tropical cyclones, then a general audience without any expertise in meteorology could benefit even more.
Along these same lines, the SS categories could be labeled as “wind categories” instead of simply “categories,” to help identify for the public what they are, while simultaneously ruling out what they are not. Logical space would also be cleared out for any storm surge warning categories that might emerge alongside the SS scale in the future; the juxtaposition of the two would help to demarcate the utilitarian boundaries of each. The size products, in particular the tropical cyclone wind field product, could also carry labels and be given a higher profile, which would in itself help to alert the public as to the significance of storm size, together with the hurricane warning, which directly communicates danger to the public, even though its definition still contains the technical construct of “hurricane conditions.”
Ultimately, an information product intended specifically for the general public would be one in which the definitions of what the product shows, and what its core components are, would be nontechnical. A possibility that might be worth exploring is the element the public already seems to be looking for in the current products, which is the danger, in the threat sense of the term. In practice, it could take the form of an early wide-area alert, a tropical cyclone alert issued beginning perhaps in the 72-h time frame. Structured as a direct threat assessment, its purpose would be to place residents of the wide area into a close listening mode, or state of alert. Such an alert could precede the watch and warning phases but would not, because of the longer lead times and the large uncertainties associated with them, be designed to trigger any action.
The underlying danger–threat concept would be operationally legitimate because it contains uncertainty within its definition, but in a way that is nontechnical. A source of danger, such as an approaching tropical cyclone, is reason for the public to fear and even to act because of its potential to cause harm, even though the eventual materialization of that harm is uncertain. The notion is at once storm centric and location centric, potentially qualifying it as a bridging concept for more effectively communicating with the public. Algorithms could be constructed for the purpose of conveying it directly, but ultimately it would rely on the judgment of the forecasters as the TC subject matter experts.
The author would like to thank the professionals in the different parts of the National Weather Service at the local, regional, and national levels. Although too numerous to mention by name, the author thanks all of these individuals for their assistance with data and for their insights. The author also would like to thank Thomas M. Parrish, for his assistance with the print-quality graphical figures appearing in this article, and the three anonymous reviewers, whose comments have significantly improved the manuscript.
In the coastal Louisiana study (Howell and Bonner 2005), in 9 of the 12 parishes, 60% or more of respondents said they felt safe in an SS3 hurricane, while for an SS4, the number was 40% or less in all the parishes in which this question was asked. In the Texas study (Zhang et al. 2007), the bar graphs show 30%–40%, indicating they would leave for an SS2 hurricane, and 50%–60%, indicating they would not; for an SS3, 75%–80% said yes, while 10%–20% said no.
Morss et al. (2008) asked about generic forecasts ranging in lead time from less than 1 day up to 7–14 days, with 90% of respondents reporting decreased levels of confidence with increasing lead time. Zhang et al. (2007) and Morss and Hayden (2010) addressed TC forecasts for Hurricanes Rita and Ike, respectively, with lead times ranging from 1–2 days up to 5 days; both likewise found that confidence levels decreased with increasing lead time. The cautionary element from the Rita study is that confidence in the longer-lead-time forecasts was still too high in an absolute sense: the bar graph shows 20%–30% of respondents reporting high (level 5 on a scale of 5) confidence in the 5-day TC forecasts.
For instance, the following examples are from two hurricane education pages. The first describes Hurricane Felix, which “roared ashore at Category 5 strength,” as “the first Atlantic storm of that size to come ashore since Hurricane Andrew in 1992.” The second discusses the damage potential of “particularly large storms (such as Hurricane Andrew)” within the context of “the force of the wind,” and goes on to describe the SS categories. These pages are available at http://earthobservatory.nasa.gov/NaturalHazards/view.php?id=18981 and http://kids.earth.nasa.gov/archive/hurricane/damage.html, respectively.
The Random House Dictionary of the English Language, unabridged ed., s.v. “technical.”