Forecasting is still all about chance

Study thoroughly and then play the probability game.

Tropical storm and hurricane forecasts are typically displayed as a cone of probability that expands with time. The most likely track of the storm's eye is indicated by a dark path line, but the storm might actually track anywhere within the white shaded region, and large storms may affect areas outside the cone.

Furthermore, such probabilities give you little guidance as to how much rain you might get. That would be a different set of probabilities altogether. The 70% chance of rain only means that there's a 70% chance for some amount of rain will fall.

On top of a 70% chance of any rain, there may be a 30% chance it will be more than 1 inch. Because they can get very complicated very quickly, a lot of these probabilities are hidden from the end user. Unfortunately, we often base our risk decisions on such probability.

However, there are proponents and detractors for such information. Having probability information would give one a better basis for making a risk decision, but it might also lull one into a false sense of security.

For example, if you are given a forecast of light icing along your route, you are faced with a choice. You may decide to go be­cause your deice boots are capable of handling light icing, or you might avoid the area because of the icing potential.

How about if that forecast were to indicate a 60% chance of icing? Where it could become problematic is if you also received the probability information that, if icing occurred, there was a 30% chance of trace, a 40% chance of light, 20% chance of moderate, and a 10% chance of heavy ice.

Although you calculate that you actually have just a 24% chance of there actually being light icing somewhere in the forecast area (0.6 x 0.4), you also know that this could be a dangerous trap since, if icing does occur, there is also a 30% chance it will be moderate or greater. For reasons such as this, the folks who issue the aviation weather forecasts have elected to stick with forecasts that simply display the most likely scenarios.

Aviation forecasts

Although probability is a key component in making an aviation forecast, if you look at a forecast product, it often has little direct probability information associated with it. Look at a terminal aero­drome forecast (TAF). TAFs are point forecasts covering conditions at an airport (out to a 5-sm radius) valid for up to a day ahead of time.

What they provide is a report of the most likely conditions during the forecast period. In highly dynamic situations, where conditions may be difficult to forecast correctly, the actual probability behind a TAF forecast may be somewhat low.

For example, the ceiling 12 hrs from now may be anywhere between 200 and 2000 ft, but there is a 60% chance of a 1000-ft ceiling, so a 1000-ft ceiling is forecast. TAFs do, however, occasionally provide limited probability information.

If the probability for a set of potentially significant conditions (such as visibility, ceiling, winds, storms) is between 30 and 50%, you will sometimes see the notation PROB with a percentage number (eg, PROB50), followed by a 4-digit start-and-end hour, and then the conditions to which that probability applies.

These probabilities are not given for the initial 6-hr period, and it's as­sumed that conditions with less than a 30% chance are not included in a TAF either. TAFs are updated frequently (every 6 hrs), which can help refine the forecast as the relevant time period approaches. A guide to reading TAFs can be found at aviation­weath­er.gov/static/ help/ taf-decode.php.

Equally common to aviation are area forecasts. Area forecasts (FAs) are issued 2–4 times a day (depending on the region), and provide similar information to TAFs. However, FAs cover many thousands of square miles, and are limited to describing conditions affecting VFR flight, such as clouds, turbulence and winds. If IFR conditions are forecast, that information will be provided in an associated Airmet Sierra (IFR or mountain obscuration). All FAs contain a 12-hr forecast, followed by a categorical outlook (VFR, MVFR, IFR) from 6–18 hrs.

The FA usually begins with a synopsis section that provides a summary of fronts and circulation patterns at work over the area during the initial 18-hr period. When conditions will vary across the FA region, such as if a front is moving through, forecasters will break the report into parts, based sometimes on states or provinces, sometimes on geographic features such as a mountain range or coastline, or relative to the weather feature creating the difference in conditions. The spatial breakdown depends largely on the variability of the forecast weather over the area.

However, like the TAF, an area fore­cast steers away from probabilities, focusing more on the most likely scenarios. While this approach does make interpretation easier for most pilots, it also affects their ability to assess the level of risk to the flight relative to the chance some weather conditions might occur.

But some general probability guidance is included in area forecasts. Often a forecast for thunderstorms over a particular zone will be preceded by a qualifier such as possible (POSS), chance (CHC), isolated (ISOL), widely scattered (WDLY SCT), occasional (OCNL), frequent (FRQ) or widespread (WDSPRD).

While such wording won't tell you directly the degree of likelihood that the storms will actually materialize, you might reasonably assume that a forecast of FRQ TRW implies a greater chance of encountering a storm than does a forecast for ISOL TRW. If the forecaster leaves off the qualifier, that is also a good indicator that the probability of occurrence is pretty high.

Airmets and Sigmets are also types of area forecasts. Unlike TAFs or FAs, Airmets and Sigmets are only issued when weather conditions are forecast that are likely to affect flight safety adversely within a specific area of at least 3000 sq miles.

The caveat of Airmets and Sig­mets is that the forecast conditions must be wide­spread—that is, affect more than 50% of the identified region—although, except for IFR conditions (ceiling below 1000 ft and/or visibility less than 3 sm), only a small part of the area may be affected at any given time if the Airmet region is much larger than 3000 sq miles.

As with the other forecast products, Sigmets and Airmets don't explicitly give probability. Rather, when they are issued it is either because those conditions already exist, or are forecast to be more than 50% likely.

Determining confidence

Looking over the TAFs, FAs, Air­mets and Sigmets will give you a pretty good idea of the weather conditions you are most likely to encounter, but the degree of likelihood is still somewhat elusive in these products.

If you want to get a better handle on how confident the forecasters are that the conditions they are forecasting will actually occur, you need to turn to the aviation (or TAF) forecast dis­cussions. Forecast discussions are a weather product issued by many weather agencies around the world that forecast the weather. In the US, each National Weather Service forecast office issues them for their area. (You can find them at aviationweather.gov/products/afd/)

Although less narrative, an experimental product that can be of great use as a supplementary planning product is the aviation consistency table. Going along with the forecast discussion, these tables are produced for TAF airports and give an hour-by-hour overview of the ceiling and visibility categories being predicted by the numerical weather models and, perhaps more importantly, what proportion of the model runs are agreeing.

Although there is always a chance that all of the model runs could be wrong, generally the closer the agreement is to 100% the more confidence you can have that the forecast conditions will occur.

Overall, though, it is important to remember that all forecasts are an exercise in probability. A low probability of thunderstorm activity doesn't mean a thunderstorm won't rise up to ruin your day. One way in which pilots can help, however, is to continue to filing Pireps. Each Pirep represents an additional data point that can be used to evaluate the forecast models and help meteorologists reduce that uncertainty.

Karsten Shein is a climatologist with the National Climatic Data Center in Asheville NC. He formerly served as an assistant professor at Shippensburg University. Shein holds a commercial license with instrument rating.


1 | 2|