Home header left
| Login Help



Thermal Soaring Forecast Methodology

In this tutorial I will provide an overview of the process and the steps I use in preparing thermal soaring forecasts. Since I live and fly in SE Pennsylvania, I do not need to deal with terrain effects, and generally work with surface elevations of between 500 - and 1,000 ft msl. However the methodology works equally well in areas with much higher mean elevation. The presence of significant terrain however injects major new variables into the mix and I have not addressed these.

Elsewhere I have provided a step-by step guide for preparing thermal soaring forecasts. That guide includes all of the web links needed to get the data needed to make the forecast. I appreciate of course that with the advent of BLIPMAPS, many may be tempted point and click and then head out to the airport on orange and red, or to the office on blue and green. I will have more to say about BLIPMAPS but for the time being I note that on occasion the day does not turn out in quite the orderly fashion implied by the beautiful presentation. BLIPMAPS should be used as part of the process of producing a thermal soaring forecast, not as a substitute for that process.

Synoptic Charts and Satellite Images

With skill and experience it is possible to make a pretty good forecast by reference to the synoptic charts alone so it would be foolish not to start here. It's worth noting too that of all the products issued by the NWS, the synoptic charts are least likely to be in error and that alone is reason to take a good look.

Surface Analysis
At a glance this chart identifies fronts, highs, lows and allows a quick estimate surface winds. It's a pretty good bet that whenever a front is either in the vicinity of the task area, or likely to move there during the day, uncertainties in the forecast will be greater.

Visible Satellite Images and Loops
These are very useful alone, doubly so if superimposed on a surface analysis chart. The loops of course also have prognosticative value since frequently a simple extrapolation of cloud motion provides a reasonable forecast of future cloud cover. With skill it is sometimes possible to infer the existence of disturbances not picked up by the models, but I usually leave that to the professionals.

Infrared Satellite Images and Loops
Since forecasts often need to be issued early in the morning when visible satellite data is unavailable I generally also look at the IR data. These provide much the same sort of detail and information as do the visible images, with the added advantage that cloud height is depicted.

850 hPa Analysis
These charts are particularly useful in assessing the wind direction and strength at about 5,000 ft msl. Since they also include temperatures it is possible to get an overview of the temperature field. A temperature difference between the forecast surface value and the 850 hPa value of at least 15°C is generally a minimum requirement for good thermal soaring.

500 hPa Analysis
The winds at 500 hPa are known as "steering winds" for good reason and it's a pretty good bet that any changes in the weather will be moving in the direction of these winds. These charts also depict vorticity vorticity (strictly speaking absolute vorticity) and where this is a mximum so is divergence aloft and when there is divergene cloud tends to form.

Surface Dewpoint Field
Surface dewpoints and temperatures, and the depth of the convective boundary layer control cumulus cloudbase and whether cumulus clouds form or not. Although surface temperatures always change significantly (at least on any soarable day), surface dewpoints generally do not so it is very useful to know the values early in the day and useful to compare actual to forecast values. Obviously knowledge of the upwind surface dewpoints is particularly useful.

Forecast Discussions
Every NWS regional office issues at least twice-daily detailed discussions of the current and near-term situation. In active weather systems, updates are also issued. These discussions are written by professional meteorologists with experience of local weather patterns. Although they are aimed at other professionals and are quite technical, they are accessible and essential reading to anyone trying to produce a thermal soaring forecast. I generally read them before and after looking at the synoptic charts and satellite images.

When, as is often the case, models are not in full agreement the better performing model is identified. That information is obviously important when it comes to choosing which of several model soundings to use. Perhaps the most useful aspect of the forecast discussion is the disclosure of the known uncertainties that quite frequently underlie the official forecasts.

Prognostic Charts

For each synoptic chart a corresponding prognostic charts exists. Comments made about the former apply to the latter and a comparison of the two, particularly after having read the discussion for the day, is useful.

Other Forecast Products

Most of the numerical models provide MOS (Model Output Statistics) forecasts and these are of great interest to soaring pilots and forecasters. The valid times are at three hourly intervals, so at least one and usually two of the times will cover the most important part of the soarable day. They forecast surface temperature, dewpoint and wind and also cloud cover. The cloud cover data is less than we might like, but it is nonetheless useful to know that cloud will be absent, scattered, broken, or overcast.

Both Acuweather and Weather.com issue hourly forecasts of surface temperature, dewpoint and wind and I generally rely upon these. The forecast values represent the best estimate by a professional forecaster based on a range of inputs including MOS.

TAF's are a useful additional source, and do include the height, as well as the coverage of any clouds.

Tracking the Forecast

At least for surface winds, temperatures, and dewpoints, it's quite easy and often revealing to compare measured and forecast values. Any significant deviation should be grounds for caution. For contest forecasting (where launch is seldom before noon) there is plenty of time to create a detailed record of forecast and measured values.


Anyone following all of the steps outlined above will already have a pretty good idea of what the thermal soaring possibilities will be for the day, but for a quantitative forecast of lift height and strength, of cumulus cloudbase and of the chances of spreadout or vertical overdevelopment, it is essential to evaluate soundings. I rely almost entirely upon model forecast soundings which are not observed but calculated. Observed data is available from any one of the many RAOB sites which cover the country. I have discussed the problems of relying only upon RAOB data elsewhere.

Which Model?
The first step is to decide which of the several model soundings available is preferred. For a forecast made on the morning of the day the RUC is almost invariably the model of choice. This model is run far more frequently than others and for that reason alone it is likely to be better. It also uses a relatively fine grid and makes some allowance for terrain. It underlies one of the two BLIPMAP models. For longer range forecasts the RUC (limited to 12 hours) is not useful. The 12 km NAM runs out to 48 hours, the 40 km NAM to 84 hours, the GFS to 84 hours, and the GFSX to 180 hours. If 48 hours is long enough, the 12 km NAM is usually preferred. If it is not, the discussion will usually note as much, and provide guidance as to which model is.

All of the sites providing sounding data offer the SkewT diagram as a format. Although a pretty good job can be done using these diagrams on a computer screen, and a considerably better job done if they are printed, I find that the FSL site with it's interactive Java based chart to be particularly helpful. I also use RAOB 5.6, an application which allows me to create an interactive chart for any sounding. I find this to be very very useful. It performs parcel lifting (as does the FSL site) and includes a range of useful tools. Because the plots it makes have the same format regardless of the particualr sounding origin it is easy to assess small changes.

Degree and Depth of Convection
The first and most basic piece of information which can be extracted from the forecast sounding is the degree and depth of convection. The degree is the separation between the surface adiabat and the environmental temperature lapse rate, the depth is the height at which that adiabat intersects the temperature lapse rate.

A major advantage of the PM forecast sounding is that the degree and depth of convection are not particularly sensitive to errors in the forecast surface temperature. This is because the model computes the environmental temperature lapse rate based on its estimate of the surface temperature. If the surface temperature were different, the lapse rate would simply move correspondingly to the right (higher surface temperature) or to the left (lower surface temperature). It follows that it is not sensible to make adjustments to the model surface temperature although there seems to be an irresistable temptation to do so.

Cumulus Cloud
The sounding also predicts whether or not cumulus cloud will form and if so at what height. All that is necessary is to construct the constant mixing ratio line passing though the surface dewpoint and extend it until it intersects the temperature lapse rate line. If the intersection is below the top of the mixed layer, cumulus cloud will form at that level, if above, no cumulus will form.

Although the degree and depth of convection are not sensitive to errors in the model forecast surface temperature, this is not the case for the height of cumulus cloudbase or even for determining whether or not cumulus will form. To make matters worse, errors in the surface dewpoint forecast are equally consequential.

As noted above, I pay particular attention to the observed surface dewpoints. If the preferred model sounding seems likely to have the wrong value for the surface dewpoint I use another approach to determine cloudbase (and of course whether or not cumulus cloud will form). The simplest variation is to take the best estimate for surface dewpoint (generally the Acuweather or Weather.com number) and simply plug that into the diagram. However, since cloudbase is uniquely determined by the difference between surface temperature and dewpoint through the relation

Cloudbase = ((T - DP)/4.4)) x 1,000 ft

where T and DP are both in degrees F, a more accurate estimate may be made by using the best estimate of T and DP from Acuweather or Weather.com. If the formula predicts cloudbase above the intersection of the surface adiabat with the temperature lapse rate it will be blue. In following this approach I am depending upon the model to get the depth of the convective boundary layer right, and upon other and better estimates of surface T and DP to predict cloudbase. This often yields better cumulus predictions than simply relying upon the sounding values of T and DP.

Lift Height and Strength
It is simple to estimate lift height: If cumulus form the top of the useable lift is cloudbase. If not, it is generally close to the intersection of the surface adiabat and the temperature lapse rate.

It is not so simple to predict lift strength. The BLIPMAP computes lift strength ab initio and this is a major strength of the model. Almost all other approaches, including mine, are empirical. Although less elegant, empirical methods often yield good results, and can be adjusted to experience.

The RAOB 5.6 program uses three empirical approaches to predicting both lift height and strength (Appendix A). I have found these to work quite well.

Over Development
Two kinds of over development are of concern: Vertical and horizontal. Sometimes both are present. Vertical OD is a potential problem if it extends to above the freezing level since it often means rain. Massive vertical OD giving rise to thunderstorms creates further problems. Horizontal OD, which I prefer to call spreadout, is a problem because it shuts down convection.

Spreadout is indicated whenever the dewpoint and temperature aloft converge. They necessarily do to some extent (because moisture is convectively transported from the surface and trapped by the inversion), but when the separation is less than 5°C spreadout is a possibility and becomes more likely as the separation decreases. Vertical OD is not immediately obvious from the skewT, but once a surface parcel is lifted it becomes clear whether or not it will be a problem. RAOB 5.6 and the FSL Java site will do the lifting for you. Absent either it is not too difficult to make a good estimate by sketching first the dry adiabat, and then the saturated one from cloudbase.

An important consequence of vertical OD, even in the absence of rain, is the possibility that much stronger winds aloft may be mixed down to the surface. This should always be checked.


In this brief review I have summarised the approach I follow in making predictions for thermal soaring. The step-by-step procedure I follow, and the links I use are presnted in "Forecasting for glider Pilots".


RAwinsonde OBservation Program Conventions

Thermal Index:The TI is determined by subtracting the environmental profile temperature from the temperature of the dry-adiabat profile for each level of interest. Most soaring literature indicates that a TI value of -3 reflects a good chance of sailplanes reaching the altitude of this temperature difference. TI values of -8 to -10 generally indicate very good conditions. TI values of 0 or greater are generally unfavorable for soaring. The pressure level at which TI=0 is referred to as Hyd. The RAOB software provides two options for deriving this altitude.

Mpi: for Mario Piccagli, whose equations were developed for use over U.S. mid-Atlantic states.

ALT (ft agl) = 1580 + [0.57 * Hyd(ft agl)]
LIFT(tpm) = 50.0 + [0.049 * Hyd(ft agl)]

Rpe: for Russell Pearson, whose equations equations were developed for use over the southwestern U.S.

ALT(ft agl) = 133.72 + [1.03 * Hyd(ft agl)]
LIFT(fpm) = 41.49 + [0.07 * Hyd(ft agl)]

The above equations apply only to dry thermals, which are conditions where no clouds exist. The effects of moisture on thermal development and cloud formation are discussed in WMO's Handbook of Meteorological Forecasting for Soaring Flight. Although not employed in RAOB, this handbook presents several manual techniques for analyzing soaring potential with respect to atmospheric moisture. One good indicator of low level moisture is the Convective Condensation Level (CCL), which typically identifies the height of cumuliform cloud bases, which are normally produced from surface heating and associated thermal activity. Mario Piccagli (e.g. Mpi) developed a lift equation using the height of the CCL (CCLht), from which RAOB also displays resulting lift strength in fpm. This equation is:

LIFT(fpm) = -10.0 + [0.078 * CCLht(ft agl)]

Trigger Temperature: This is the surface temperature required to produce a dry-adiabatic lapse rate which will intersect the sounding at the altitude specified.

Soaring Index (SI): Researchers have recently developed the Soaring Index which is designed to incorporate the vertical temperature gradient between the trigger temperature and the maximum altitude of thermal activity (Armstrong and Hill, 1976). Like the above Pearson and Piccagli lift strength estimates, the Soaring Index also produces an estimated lift strength in feet-per-minute (fpm).

Soaring Index = [3 * (Z/100)] + [10 * t]

where: Z = maximum thermal altitude (ft., agl)
t = (T : trigger temp.) - (T : maxthermal) degrees C
T : trigger temperature = sounding temperature at trigger altitude
T : maxthermal = sounding temperature at Z (max thermal altitude)

Posted: 1/18/2007 By: Richard Kellerman

Member Blogs 

Search Posts

Recent Posts

Legal Notice

The SSA policy on member posting is located here