This site is best viewed in a modern browser (Chrome 10+, Firefox 3.5+, IE 8+, Safari 4+).
Thermal Soaring Forecasts
I make extensive use of thermodynamic diagrams, the single most important tool used in preparing thermal soaring forecasts. It is certainly possible to produce a qualitatively plausible forecast without reference to these diagrams, particularly with a good deal of experience to rely upon. On the other hand failure to make full use of them will always result in an inferior forecast. I have written a tutorial "Thermodynamic Diagrams" and have also written two other tutorials which will be helpful: "The Convective Boundary Layer and the PM Sounding" and an overview of my methodology. It will be very helpful to read these prior to making use of this guide which is merely an outline and a structured presentation of the resources I use.
Traditionally thermal soaring forecasts have relied upon AM balloon soundings, and on estimates of the surface dewpoint and temperature for the day. In my primer on the convective boundary layer I discussed why this method generally leads to poorer forecasts than reliance upon numerical model forecast soundings. This may seem odd given that balloon soundings are measurements and numerical model soundings only calculated, but extensive experience, and the utility and popularity of BLIPMAPS strongly support the validity of the PM sounding method. I rarely make direct use of ballon soundings, although the data they provide is a critical input to the various model soundings I use.
Getting the Big Picture ("Synoptics")
It's possible of course to just examine the soundings for the day and ignore everything else. Sometimes that approach works well, but more often it leads to an inferior forecast. The forecast sounding is after all only a forecast, and as such is subject to error. How much error, and what kind of errors might be anticipated are questions best answered by a careful assessment of the current observational picture.
My home page on my browser is (Unisys Weather) and this is where I start with the current satellite image and surface map. This provides the best overall picture of the day, and has the added advantage that it can be looped. I next read the local NWS office discussion (click the map with the states outlined and labeled). Although frequently highly technical, and always sprinkled with arcane contractions, these discussions are an absolutely essential part of the data gathering process. They give the professional's assessment of the validity of the various models, and of the synoptic and prognostic picture. They provide the indispensable background against which I evaluate data. They also will generally comment when there is significant divergence in the various models and will advise on which is handling the current situation best.
The Unisys site also has an excellent Satellite/Surface Composite image and I generally look at this as well as the IR Satellite Radar Composite.
I use the NASA GHCC Interactive Global Geostationary Weather Satellite Image Viewer for high resolution visible images - these resolve individual cumulus clouds (at least the larger ones) and give the clearest possible picture of the evolution of clouds of all types through the looping capability. They are generally current to about 15 minutes. Early in the morning the visible satellite images are of course not visible, so I use the IR images. Apart from a slight tendency to make matters look rather worse than they are, these are an excellent substitute. The water vapor images are also very useful, particularly given the pivotal role which water vapor often plays in soaring weather.
Getting the Surface Temperature and Dewpoint Right
The best possible estimates of surface temperatures and dewpoints are essential for forecasting cumulus formation and height.
Every sounding includes of course the model forecast value for the surface temperature and dewpoint and one approach is to accept those values. My experience has been that other sources frequently are more reliable in forecasting the surface temperature and surface dewpoint.
Although errors of a few degrees in the model forecast surface temperature have only a small effect on the thickness of the convective boundary layer, as I have noted in my methodology summary, they have a significant effect on forecast cloudbase, as do errors in the surface dewpoint. Recall that for every 4.4F° spread in surface temperature and dewpoint cloudbase is 1,000 ft higher.
I always look at the MOS ( Model Output Statistics) for three operational models ( NAM, NGM, GFS) then compare these to the reported values of T and DP for the same site. I also use the Accuweather site which allows me to drill down to hourly forecasts of T and DP and it is frequently the case that this is the best data for the day. I always check for any disparity between forecast and observed temperature and DP values - this can be an early warning of trouble to come. It is also frequently revealing to look at upwind observations. A very useful tool for assessing the surface DP field is the Unisys surface dewpoint contour plot.
The Model Sounding
The model sounding provides the basis for estimating the depth of the convective BL. It also gives a measure of the degree of instability within the BL. Both are quantified by constructing the surface adiabat passing through the model forecast surface temperature. Note that for the purposes of determining the depth of the BL and the degree of instability it is important to use whatever surface temperature is forecast by the model. For reasons I have explained elsewhere choosing some other value is meaningless.
The height at which the surface adiabat intersects the environmental temperture curve defines the top of the BL, and for thermal flight that is the maximum height to which a glider can climb. When cumulus cloud forms below the top of the BL the maximum height is cloudbase, at least in the USA where (legal) cloud flying is impractical.
The strength of the lift is determined both by the depth of the BL, and by the separation between the surface adiabat and the environmental temperature curve
There are many models and which to use is itself sometimes an art. The question of which is performing best is usually answered in the NWS discussion. For a forecast made in the morning of the day of interest the RUC model is preferred simply because it is run more often and at most times of the day is more current. If the 12 km NAM forecast enjoys the same currency as the RUC (which it does, twice a day at 0650Z and 1850Z, neither of which are particularly useful for soaring forecasts) then its higher resolution has some advantages. Forecasts out to 84 hours are provided by the 40 km NAM and the GFS models. A complete listing of model update times is available at http://twister.sbs.ohio-state.edu/models.html.
The two sites I make the most use of are the FSL Java site and the NOAA ARL READY site. The choice of model is in part dictated by how far out the forecast is to extend, in part by the assets and liabilities of the models, and in part by which of the models happens to be performing best.
With the groundwork in place, I next look at the RUC BLIPMAP and the ETA BLIPMAP (now officially " NAM" for North American Mesoscale). I find Dale Kremer's "BMapper" to be more useful than the standard BLIPMAP display. Not only does it overlay the RUC 25 km grid on the task area and waypoints, it also gives access to 11AM and 5 PM data (in addition to 2 PM data used in the BLIPMAPS), and it makes it very easy to generate a sounding at any grid point. To get the sounding from FSL all that is needed is to right click the grid point circle.
I am now ready to estimate the strength and height of the lift for the day, and to see if cumulus clouds will form and if so at what height. I start by determining of the depth of the convective boundary layer. This is something the BLIPMAP does well, but I always confirm those values by looking at several individual soundings. A little more work is required to do this. It is necessary is to construct the dry adiabat passing though the forecast surface temperature and note the height at which it intersects the temperature lapse rate line. The top of the convective BL is the maximum height to which we can expect to climb. This is because by definition the convective BL is that part of the atmosphere convectively mixed to the DALR. So, in the absence of cumulus cloud the top of the lift is just the top of the BL.
On many days of course cumulus clouds form. So, how do we know if they will form, and if so, at what height? One way is simply to click the bottom right end of the sounding on the FSL site and see where the black line appears: If it is to the right of the temperature lapse curve cu will form, if to the left, they will not. The problem with this approach is that the RUC generally performs more poorly in forecasting surface DP's than do the various MOS forecasts or Accuweather and this frequently leads to botched cumulus forecasts. I prefer make use of either MOS or Accuweather T and DP forecasts. I take the difference in forecast surface DP and T (both in Fahrenheit), divide that difference by 4.4, then multiply the quotient by 1,000 to get the height of the cumulus cloudbase (AGL!) in feet:
CLOUDBASE = ((T - DP) / 4.4) * 1,000 FT.
If that number is less than the height of the top of the BL cu will form, if greater, they will not. If close, it's a tossup.
It is also possible to enter a modified value of the surface DP on the FSL plot by right clicking after placing the cursor at the bottom of the superadiabatic layer (bottom right hand corner of the temperature profile). A dialog box then allows numerical entry of T and DP. The value of the surface temperature should be the default value output by the model. There seems to be an irresistible temptation to "tweak" the model value but it should be understood the model computes the values of the temperature in the convective boundary layer using the surface temperature as an input. Changes in the latter necessarily result in changes in the former. It might be objected that adjusting the value of the surface dewpoint is also not sensible but for the limited purpose of determining cumulus cloudbase it is quite legitimate.
BLIPMAPS forecast lift strength but I also like to take a look at other estimates which do not depend on the details of the BLIPMAP model. I do make some use of the empirical lift strength equations included in RAOB 5.6 but I have also developed my own coarser empirical guides for estimating lift strength.
My minimum criteria for workable lift is a separation of 2°F between the surface adiabat and the environmental lapse rate. I describe conditions corresponding to this small separation as "weak" and if pressed will commit to 200 fpm climb. A difference of 3-4 °F earns a "moderate" rating for lift strength, say 3 - 5 knots climb. Anything greater than 4°F gets a "strong" rating. My reservations on making more quantitative assessments of lift this is an area where high precision is all too often undone by low accuracy. Put non-technically, I prefer to say "2 - 4 kts" rather than "316 fpm". I continue to use degrees °F to assess lift strength and height only because so many surface forecasts use this archaic measure.
No soaring forecast is compete without winds. I generally rely on the RUC. I always include surface and 30,000 ft winds and of course winds throughout the convective BL. The former are of obvious importance, the latter control to motion of cirrus and are usually representative of the overall motion of weather systems. For ridge sites I always include winds at ridge height. I generally give winds at every 2,000 ft in the BL.
I don't attempt to forecast streeting for the simple reason that I don't know how to. I have also never heard a single complaint from a pilot who found streets where non were forecast. I appreciate of course, as do all soaring pilots, that if streeting is to occur, it will be in the direction of the local wind.
Whenever cumulus clouds do form there is always the possibility of overdevelopment. Overdevelopment can take either (or both) of two forms: Horizontal and/or vertical. I prefer to call horizontal O.D. "spreadout", reserving the term "overdevelopment for towering cu, or cu min. Both are easily predicted by inspection of the skewT. If the separation between the forecast temperature and the forecast dewpoint in the vicinity of cloudbase are less than 5°C then there is a possibility of spreadout, simply because the air surrounding the clouds is itself quite moist and so inhibits the dissipation of cloud. The smaller the spread in T and DP at or about cloudbase, the greater the chance of spreadout. Values of less than 2°C often result in an extensive stratocu deck. Vertical convective overdevelopment is also apparent by inspection of the skewT but the job is made even easier by the FSL site which computes and displays the appropriate saturated adiabat when the sounding is left-clicked at he surface temperature. If the sounding shows cloud growing to above the freezing level it is likely that rain will fall, or at least try to (virga). If cloud extends into altitudes where winds are strong those winds can get mixed down to the surface and cause problems.