Several meteorological, climatological, regional and human perception uncertainties apply to constrain the application of warning strategies. These include the uncertainty in cyclone forecasts, regional and seasonal variations in cyclone characteristics and forecast difficulty, and language and communication ambiguities that lead to misinterpretation of warning content.
Because the atmosphere is non-linear, some level of uncertainty, or error is present in all forecasts and must be accounted for in warning strategies. This uncertainty can be quantified by mean forecast errors (Chapter 3) that generally increase in 100 km increments for every 12 h of the forecast. Pike and Neumann (1987, see Chapter 1) have shown that the degree of uncertainty, as defined by a forecast difficulty index using the CLIPER technique varies considerably between ocean basins. Marked seasonal and latitudinal variation also is experienced.
In many situations the forecast errors will prove to be less than indicated by the forecasting difficulty level because the forecaster will use a number of techniques, allied with professional judgement and experience, and there may be much better than normal observations available. In other cases the warning strategy must account for the worst case scenario, for example, the high degree of uncertainty that can be present when a tropical cyclone is in a potential recurvature situation.
The warning strategy, therefore, is to account for regional characteristics and the current situation and for the forecast period and then to provide a prudent margin for error. At the alert or watch stage, a large section of coast will be notified, and this will be reduced and focussed as the cyclone landfall approaches and the forecast confidence level increases.
Care needs to be taken to maintain realistic and conservative levels of forecast uncertainty in the warnings. Counter-disaster managers tend to regard the occasional instances of excellent forecasts as the basis for routine expectancy of exceptional performance which, if not achieved on subsequent occasions, quickly inspires criticism of apparent monitoring negligence. Forecasters have a natural tendency to overestimate their real predictive skills by making overconfident forecasts of the probable location of landfall in warning announcements several days in advance. This subsequently leads to substantial amendments, confusion and potential tragedy. Forecast improvements over the past two decades have been quite modest, averaging only approximately 1% per year (Neumann, 1981). When considered in terms of the disproportionate increases in coastal populations, intensive land-usage and vulnerable investments, such improvements barely make improved response measures effective in human and economic terms.
An indication of the best possible forecast skill is provided by that obtained by the United States National Hurricane Center at Miami, staffed by specialist hurricane forecasters who have the best available initial location accuracy and forecast track guidance. The mean errors for the decade 1975-84 were (Neumann, 1985):
|Forecast Period (h):||12||24||48||72|
|Mean Error (km):||104||205||414||633|
|Accuracy Limit (km):||110||520||360|
Whilst current research and model development (Chapter 3) indicates that a substantial improvement in forecast accuracy is occurring, the best forecast performance that can be reasonably expected will result in errors greater than half those listed above. This should be accounted for in planning for future warning strategies.
One method of quantifying the uncertainties inherent in a particular situation is by the issue of strike probability forecasts (Chapter 3). However, many forecasters are unaware of the practical significance of these statistics, since forecasts are judged correct, partially correct or incorrect according to other criteria. They thus may not appropriately relate hit or miss probabilities to suitable preparedness measures by people living in the threatened areas.
For the purpose of determining hurricane strike probabilities in the U.S.A. a hit is adjudged a successful forecast for appropriate preparedness decision-making if actual landfall occurs within 100 kilometres to left or right of predicted landfall. This is the approximate length of coastline that may be considered liable to the severe destructive power of hurricane-force winds and storm surge encompassed by the central core of an intense tropical cyclone (Carter, 1983). A stricter interpretation would incorporate the left-right asymmetry due to storm motion. The probabilities of successful hit predictions, given the aforementioned mean errors for specific forecast validity periods can be shown to be:
|Forecast Period (h):||72||48||36||24||12|
|Maximum Probability (%):||10||13-18||20-25||35-45||60-70|
|Miss to Hit Ratio:||9:1||7:1||5:1||2:1||2:3|
Thus the odds of a specific location forecast issued more than 36 h in advance of estimated landfall being assessed as a successful hit prediction are rather slim. An even probability of success is not reached until nearing 12 h before landfall.
As a caveat to this discussion it should be noted that the narrow assessment used in the U.S.A. for hit and miss criteria, based on the premise that the destructive power of a mature cyclone will be limited to a spread of destruction of about 200 km from the coastal crossing of its centre, may be expanded for the Asian and other regions. Asian cyclones tend to be larger than those in the North Atlantic (Merrill, 1985). In developing countries the impact of winds within 300 km or so of the centre of a mature storm may well cause significant damage to modest houses built of traditional materials. Similarly, gale to storm force winds may develop very rough seas sufficient to endanger the huge number of small craft engaged in fishing off most tropical coasts.
As has been emphasised in Chapter 3, the view of a "normal" cyclone track extending along a straight or neatly parabolic arc into the subtropics whilst the cyclone develops along the standard Dvorak climatological curve is actually the exception. The majority of cyclones are associated with some sort of abnormal behaviour. Examples include:
These "unusual" characteristics must be properly accounted for in warning strategies. In particular, communities need to understand that cyclones are capable of erratic behaviour and that the uncertainty in warnings is designed to account for such behaviour.
The content of a tropical cyclone warning message is of prime importance in an effective warning system. This information has to be converted into convincing and credible images of the approaching threat which will create a climate of expectancy and responsive action over a very wide spectrum of people with differing preparedness interests and responsibilities. In countries which possess a communications and media infrastructure capable of presenting a current visual view aided by expert commentary the reliance on the technical formulation of a warning message is not high. In most countries affected by cyclones, however, one prime warning message for the general public often carries the responsibility of catering for a vast diversity of user needs. At best a limited number of messages semi-tailored to the needs of major socio-economic groups must suffice. In some cases, the dissemination relies on a series of relays, often by telephone and single-side-band radio, during which the message may be abbreviated or distorted and lose some of its original intent.
The format, content and terminology of warning messages should be determined by the status of the warning phase, category of user, level of public understanding, vulnerability of the threatened areas, particular medium used for dissemination, and regional agreements on standardised procedures. A major difficulty is that the messages are prepared by technical experts yet even the most obvious technical terms can cause considerable confusion amongst many members of the community. For this reason, careful use of key words, backed up by public education programs is essential.
Inspection of recent warnings issued by many different warning services indicates a wide variety in either pragmatic or narrative style, and a seemingly reluctance to introduce new styles of persuasiveness and format to focus on the most critical information. Examples of various styles of messages may be found in WMO (1983b, 1990) and in Elsberry et al. (1987).
The optimum design for tropical cyclone warning advices, including audio-visual signals, is discussed in detail in WMO (1990). This discussion takes into account the diverse requirements of users and the available communication modes. Norton (1975) suggests that it is feasible to separate the main facts, other important facts, and then details and explanations in descending order of priority in the compilation of messages. Special consideration should be given to the limited attention span of most people and their ability to retain the important elements of a message. The essential information may be determined by surveys of disaster specialists, the media and communities. Use of information and communications specialists is advisable.
Contents Chapter 7.5