To clarify a little bit on Random Chaos' good post about the models...

The CMC is the Canadian global forecast model. It does well with the systems that tend to affect Canada most often -- the midlatitude systems and the extratropical transitioners. It's not as good -- in general -- as some of the other global models when it comes to tropical systems, however.

The WRF is a new, developmental model designed for both midlatitude and tropical forecasting. There are numerous "versions" of the WRF in development for various tasks, making it the future "end-all" model for quite a bit of research operations. It'll likely replace the MM5 in about 3 years, but as noted before, is still under development. But, it has shown good promise in all aspects of its abilities, so it should be watched for reference.

All of the major models -- GFS, ECMWF, NOGAPS, UKMET, CMC, GFDL, the mesoscale (MM5/WRF) models -- are dynamically-driven models. Simply put, they solve the governing equations of meteorology and produce a forecast from their solutions. Each model employs different data assimilation techniques, leading to different analyses and different forecasts. Any given forecast is only as good as the initial data you give it, of course, and in general the Navy's model (NOGAPS) and the European model (ECMWF) are the best at this task. (It should be noted, however, that the ECMWF is rarely used as a tropical model, but current statistics show that it is the best global model for midlatitude weather patterns...and has been for some time.) Further, each model has its own method of representing convection, rain, boundary layer effects, and so on within its input systems & equation mechanisms, leading to more of the spread you see between each model.

The FSU Superensemble model is unique in that it is purely statistical. It takes the output from a selection of global dynamical models and uses their past history (e.g. biases) in a complex statistical scheme in an attempt to obtain a better intensity and track forecast. More often than not, that goal is met. It has been the best performing track model in the Atlantic for two straight seasons, though the output wasn't always available in a timely manner for forecast purposes in 2003 and is thus not included in the official statistics.

Other models, such as the BAM-series, CLIPER, LBAR, and so on, fall into the statistical-dynamical or pure statistical series, but none are as complex in their statistical schemes as the FSU Superensemble. Most of these models are over 10 years old and do not perform as well as the dynamical models, though some do have their uses. The BAM series performs reasonably well in the tropics, given a storm of a reasonable intensity, while the CLIPER model does well for storms that follow just what it says -- CLImotology and PERsistence. In general, however, these are older models with limited uses at this point in time.

That said, what should you be looking at in the model output available from the similar PSU & FSU websites? Generally speaking, I like to use about 5 tools: sea level pressure (to track the surface features), 850mb vorticity (better representation of the tropical cyclones and allows you to better identify potential development concerns), 500mb vorticity (can occasionally give you an idea of the mid-level flow regime, but I don't use this that often), 850-200mb shear (also includes the average flow pattern between those levels, making it a good proxy for what you might see for a potential track as well as the ridge/trough pattern), and the 850mb theta-e (low-to-midlevel moisture, particularly when trying to forecast when a dust layer outbreak might end). Feel free to experiment around, however.

Hope this helps clarify some of the model questions...