Reading between the lines
What kinds of forecasts are available out there? How solid are their claims of accuracy? What do you do when forecasts from different sources disagree? As consumers of weather information, it works to our advantage to ponder the forecasts we get and learn how to use them wisely. To paraphrase the Roman maxim caveat emptor, let the weather customer beware.
Forecasts for every time and place
The one- to two-day prediction that's so familiar to us as “the forecast” is only a sliver in a spectrum of outlooks that extend from a couple of hours to a year or more ahead. Here are a few of the most common types:
- Nowcast The briefest kind of official forecast, it's a quick take on what might happen in the next two hours or so. For instance, a 2pm nowcast might mention that storms with high wind are approaching your area and will arrive by 3pm. Nowcasts are a growing part of weather prediction in many countries, but unless they pertain to severe weather, you'll have a hard time finding them on radio orTV, since their short half-life doesn't mesh well with regularly slotted programming. Your best bet is to check the Internet site of your local forecast provider. Or you can subscribe to one of the weather-alert services now available in many locations, especially in US high-tech corridors.The services pass on bulletins at almost any desired interval via cellphone, wireless modem, pager, fax, or proprietary device. Some countries provide forecasts, nowcasts and other information via a network of local radio stations accessible by dedicated, off-the-shelf receivers; in the US, for instance, there's NOAA Weather Radio.
- Severe-weatheralerts Warnings for severe thunderstorms, flash floods or other short-fuse events operate on the same time scale as nowcasts, but they're much more likely to be aired through local media. In the US, life-and-death warnings are the one forecast area still considered the sacrosanct domain of government meteorologists. Competing interpretations of whether, say, a tornado is approaching could make the public more hesitant to seek safety (and might leave a private firm vulnerable to lawsuits). Every nation and every media outlet has its own opinion on which official warnings are most critical to put on air. Flash-flood warnings, for instance, aren't even issued by many European governments, and the/re often ignored by commercial radio and TV elsewhere, even though flooding is the single deadliest storm hazard in many countries.
- Short-rangefo recast This is the 12- to 72-hour outlook that's been at the heart of weather prediction for many decades. The standard elements include high and low temperatures, cloud cover, the risk of rain or snow (either in words or in probabilities) and – especially for the first day or so of the forecast period – wind direction and speed. The last of these may be provided in kilometres or miles per hour, knots or as a numerical rating of force from the Beaufort wind scale, used in the venerable British shipping forecasts. Some new features have been added to the standard forecast over the past decade or two. Alr-quallty monitoring has made it possible to give current and projected levels of pollutants, as well as pollen and other allergens. In the 1990s, ultraviolet (UV) Indices were added to forecasts in more than thirty nations. Using a simple numbered scale these popular outlooks take into account the time of year and the level of cloudiness to give people a sense of the peak intensity of sunburn-inducing ultraviolet light
- Medium-rangeforecast Covering from three to as long as ten days (even longer in some areas), this is the extended outlook that's an increasingly standard feature of weather reporting in newspapers and on TV. As of early 2006, official extended outlooks in Canada and Britain cover five days, including high and low temperatures and a word or two of broad-brush description (eg “mainly sunny”). Australia issues seven-day outlooks, although the final three days provide only verbal description of temperatures (eg “warm to hot”) rather than numbers. And the US National Weather Service (NWS) provides full seven-day outlooks, including high and low forecasts. Not to be outdone, many local TV stations in the US now offer ten-day temperature and precipitation outlooks. All this is a byproduct of the increasing foresight of computer models that now track large-scale weather features up to fifteen days. Simply because a model covers that extensive time-frame doesn't automatically mean it's trustworthy, of course. Even the best local forecasts of more than a week or so in advance are only a little better than a climatological estimate, if that. Nevertheless, the persistence of a locked-in pattern, or the emergence of a change in that pattern, may come to light In the models more than seven days in advance. If a big transition appears consistently over several days'worth of extended outlooks, it's a good sign that the change is likely to happen, even if the timing and intensity aren't yet certain.
- Long-rangeforecast Some governments and private firms issue general outlooks that extend more than a year in advance, relying on the influence of slowly evolving climate elements such as regions of warmer- or cooler-than-normal ocean temperatures. El Nlfto and La Nifta are the premier examples of such forecasting tools. Once it's known an El Niho is unfolding, millions of people across the world can be given notice that floods or drought are more likely than usual in the months ahead. Perhaps because their output can be difficult to explain, long-range forecasts are seldom spotlighted in the media; they're easier to find on government Internet sites. Maps typically show the departures from average, by month or by season, expected in the forecast area for both temperatures and precipitation.The NWS uses a complex triad of probabilities to create such outlooks for overlapping three-month periods ranging up to fifteen months. While they're not much help for planning a wedding next year, such outlooks can make a huge difference to industry. For a utility company that generates and trades power each day, for instance, a degree or two of change in average temperature over a season can save (or burn up) millions of dollars. Seasonal forecasting has helped create a market for weather derivatives – a form of investment added to the Chicago Mercantile Exchange in 1999. Utilities, agricultural businesses and other firms purchase weather derivatives as insurance against a sharp temperature trend that might threaten their profits. If a season's climate deviates far from the expected range, the derivative pays off and the buyer may be rescued from financial ruin.
- Climateprojectlons A few research centres around the world are attempting to project global climate a hundred years or more into the future. These aren't considered forecasts so much as scenarios – portraits of how climate might evolve if society grows and industrializes at a particular pace. No single projection can be taken as gospel, but these remain the best tools for gauging our possible impact on tomorrow's climate.
Before television weather came of age in the 1960s and 1970s, the public had to live with the forecast – the only one in existence, issued by the government. Now you can find numerous predictions through TV, radio, newspapers and the Internet (as well as pagers and other portable media). In many areas you can find differing outlooks among the four sources above, with each forecast prepared by a different meteorologist in the public or private sector. Factor in the multiple outlets in bigger cities for each form of media, and you could be facing literally dozens of forecasts valid for the same period.
Nearly all forecasts are at least partially derived from a common base of government models, so the differences between competing outlooks are often fairly minor, especially when the weather is tranquil. But when major changes are on the way, big discrepancies may be exposed, not only in the substance of competing forecasts but also in their timing. Professionals often go with a consensus of several different models in crafting their forecasts. This isn't such a bad course to follow, assuming the mixed forecasts consulted are all reasonable interpretations of a complex scenario. It's important to pay attention to when the forecasts are issued, however. All else being equal, the most recently issued forecasts should be weighed most heavily.
Consensus isn't necessarily the ideal option. Some forecasts are simply better than others. There are variations in procedures and skill levels even among seemingly uniform entities, such as the larger meteorological services. Only by paying attention to forecasts over time can you assess, if only in a gut-level way, whose predictions you trust the most. If you're really motivated, try scoring your local forecast sources for a few days on some easy-to-calculate variables like high and low temperature.
Don't be dazzled by mere technology. In some places you can now get hourly forecasts of temperature or precipitation through a subscription service via cellphone, email, pager or the Web. While such services may be convenient, and they may in fact give you the most accurate forecasts for your area, their quality depends solely on the models and other data on which they're based. No matter how finely a forecast is sliced and diced, increased precision alone doesn't improve accuracy.
What can possibly go wrong?
The official forecast in New England on the morning of September 21, 1938, was for cloudy, windy weather. Cloudy and windy it was – but nobody knew a severe hurricane would be tearing its way through the region by late afternoon. The storm was projected to move eastward and out to sea from its location off the North Carolina coast the previous night. But the forecasters of that time lacked data on the jet stream winds that steer hurricanes from above. In retrospect, we can now deduce that the jet was positioned in an unusual south-to-north orientation. It pulled the hurricane into Long Island and New England at the stunning rate of 113kph/70mph, one of the fastest motions on record for a tropical cyclone. Winds gusted to 299kph/186mph at Blue Hill Observatory near Boston, and catastrophic flooding resulted in Providence, Rhode Island. In all, some seven hundred people died.
Forecasts don't have to be this wildly off the mark to cause problems. Participants in Australia's 54th annual Sydney-to-Hobart Yacht Race were warned about gales as they embarked on December 26,1998. The tailwinds got the race off to a fast start, but an unexpectedly strong storm – too compact to be well forecast by the models at first – cranked up just north of Tasmania. An hour into the race, the Bureau of Meteorology upgraded its outlook to a storm warning after a high-resolution model had finally revealed what was to come. Yet conditions were far worse than the yacht teams expected. As it pulled in westerly winds at near hurricane force, the fast-growing storm sent huge waves crashing almost directly into the existing swell. The result was not just a typical high sea – the average waves ran 9m/30ft – but a confused one, with occasional rogue waves that exceeded 15m/50ft. The nation's largest-ever marine rescue saved 55 people, but only a third of the 110 participating craft reached Hobart. Five boats sank, and six people drowned.
A failed prediction is every forecaster's worst nightmare. There's no way to tell in advance which forecasts are problematic, but there are clues that can help a layperson sniff out uncertainty hidden in the seeming certainty of a particular outlook. Here are a few of the most common forecast predicaments and how meteorologists deal with them:
- Flip-flops There is far more weather data available for some parts of the world than for others. When a system is developing in a data-sparse region, such as the Atlantic or (especially) the Pacific, computer models have a more difficult task projecting its future. Every six or twelve hours, the models take another stab at it.The projected scenario may shift dramatically from one model run to the next until the true character of the weather system becomes clear. This is one source of what forecasters call flip-flops. The model guidance for next weekend may be screaming “cold'on Tuesday, then shift to “mild” by Tuesday night, and then “bone-chilling cold” on Wednesday. What should a forecaster do in such circumstances? The most common approach Is to gradually shift the forecast in a plausible direction by looking at how several different models treat the situation. At all costs, forecasters want to avoid putting the public through the back and forth gyrations that the models inflict on the forecasters themselves. Thus, the weekend forecast might run cold on Tuesday and then a slight bit colder on Wednesday, disregarding the models mild flip-flop on Tuesday night as well as the apocalyptic deep freeze projected in Wednesday's model run. If the public forecasts continue to get colder as time goes by, you can bet that Saturday might be still more bitter than they're acknowledging. Even when the models insist on an unusual event, it may take a forecaster some time to adjust the forecast so it reflects that event This is partly to avoid the dreaded flip-flops and partly a recognition that model error goes down as the prediction period gets closer.
- Elther/or Warm and cold fronts are better managed by the models than ever before, but not every parry and thrust of a front is predictable. Cold fronts (the more dramatic of the two) can be especially troublesome. The sheer density of the air behind a shallow, but intense, cold front can help It to roll across the countryside more quickly than a model may predict. Back-door cold fronts – those that ooze westward, against the upper-level flow – can also be tricky. Once a mass of cold air is lodged in place, predicting its erosion is a challenge of another sort Wintertime cold air may be dammed against mountains like the US Rockies or the German Alps for days on end, eventually to be dislodged by a burst of warmer air from above. The models might see the warm air coming, but be unable to tell whether It'll have the force to punch through the cold. In a perfect world, forecasters would telegraph uncertainties like these in a forecast phrased something like this: “Tomorrow – a 60 percent chance of continued warm, muggy weather and a 40 percent chance of sharply colder conditions''. But it's hard enough to deal with one forecast, let alone two or more for the same period. Until the public is ready to accept multiple-choice forecasts, the meteorological world is in no hurry to provide them. Typically, a forecaster will either go for broke – focusing on just one of the two or more options at hand – or else hedge his/her bets and issue a forecast somewhere between the alternatives. Instead of the muggy 30°C /86°F, or the chilly 10°C/50°F, the predicted high might be a lukewarm 20°C /68°F. From the standpoint of numerical accuracy, this is the safer forecast, though in a sense it's the less honest one.
There's no easy way for the consumer to spot an either/or forecast. However, any sharp contrast in air masses will eventually make its way into the forecast It's wise to bear in mind that this changeover could happen a little sooner or later than predicted, and the transition could be sharper than indicated, especially when a cold snap is on the way.
- Rain, snow and all that If you consider how difficult it Is to analyse how precipitation forms within a cloud, it's amazing that rain and snow forecasts are as good as they are. It's now rare that one will encounter precipitation in real life without some hint of its onset in the forecast. Probabilistic forecasting of rain and snow allows people to decide for themselves whether a given chance of moisture is enough for a picnic to be cancelled or a hike to be postponed.
Forecasters approach model output on precipitation with a healthy dose of scepticism. While the physical equations within most models are quite skilful at moving air around, they're less able to make the fine distinctions between, say, air that's not quite saturated (90 percent relative humidity) and air that's fully saturated (100 percent RH).This helps lead to a depiction of rain or snow in the model that can be way off in location or intensity. As a rule, gentle, steady rains or snows are more reliably simulated than are the big events, such as intense winter storms, hurricanes and thunderstorms. The last of these depend on convective effects – where the air is unstable and buoyant, rather than merely pushed upward – and many models are notoriously poor at handling convection. Indeed, models traditionally “parameterize” thunderstorms, portraying their physical consequences in broad-brush strokes without even trying to simulate the storms directly.
Over time, forecasters have learned to look at the large-scale features in a model and infer from those the most likely patterns of rainfall or snowfall. In the past decade, mesoscale models have provided a narrower focus on features from 10 to 100km/6-60 miles across. In some cases, these models can depict thunder and snowstorms in a far more realistic fashion than ever before. A few experimental models can telescope their resolution down to as fine as 1 km/0.6 miles in targeted areas. However, the heavy demand on computer time needed to produce such detail means that it may take years for a mesoscale model to move from the world of research into day-to-day forecasting.
- Snow remains a particular challenge to forecast. This is especially true at the margins, when temperatures are just cold enough to support snow, or just warm enough not to. When you see a forecast of “rain mixed with snow” it's fair to assume that the situation could just as easily be all wet or all white. Location is another factor to consider. Snowfall – and rainfall, for that matter – can vary by more than 50 percent in as little as 1.6km/1 mile, particularly near mountains or large bodies of water where local effects come into play. Even if a forecast calls for a uniform amount of snow across your area, don't be surprised to find large variations across short distances.
- When will It stop? Predicting the end of a rainy or snowy period is one of the more clear-cut tasks in forecasting – except when a certain kind of storm has taken hold. In Los Angeles, weathercaster George Fischbeck popularized the phrase “cut-off low, weatherman's woe” and it hits the mark nicely. Especially at mid-latitudes, upper-level centres of low pressure can get pinched off from the jet stream that once carried them along. Think of a river's wide meander which, over many years, becomes a small lake as the river cuts a straighter course nearby. In the atmosphere, a cut-off low can form in a day or two, then spin virtually in place for a week or more. The result is a prolonged bout of unsettled weather in the low's vicinity and just to its east Because there may be no clear “kicker” – an incoming system that pushes the low away – models and forecasters can't say for sure when the dreariness will end. If a cut-off low Is giving you woe, it may well stick around for a day or two longer than forecasts indicate.