NAM and RAP are two different models run by NOAA's NCEP (National Center for Environmental Prediction) and intended for slightly different purposes.  BLIPMAPs post-process the output from each to forecast parameters which are more useful to gliding pilots than the standard meteorological parameters. 

The NAM and RAP models are similar in that they both solve the fundamental meteorological equations as described in the How Does a Meteorological Model Work ? webpage.  In detail, however, their differences abound (and these differences are what meteorologists study and argue and write papers about!)  Most of those differences lie in the approximate treatment of processes which cannot be modeled exactly, hence there is no exact procedure upon which everyone can agree. 

For this reason the models do produce different forecasts - though often similar, their forecasts can also differ significantly from each other.  Each has different advantages and disadvantages, which are often revealed only through comparison between their predictions and the actual weather.  From a meteorologist's point of view they can complement each other, which is why NCEP runs both models operationally!  For example, comparing NAM vs RAP forecasts helps evaluate the uncertainty of a day's forecast - if they agree then the uncertainty is smaller and the forecasts are then considered more reliable than on those occasions when the model forecasts differ significantly. 




Click here for
NAM forecasts

Click here for
RAP forecasts

Forecasts out to 84 hours

Available only for two hours each day

Updated at 6 hr intervals

For same forecast period,
NAM available ~1hr later

12 km horizontal resolution

Poorer vertical BL resolution

Forecasts out to 12 hours

Available for five hours through day

Updated at 3 hr intervals

For same forecast period,
RAP available ~1hr earlier

20 km horizontal resolution

Better vertical BL resolution

Numerous model differences of detail, such as their parameterization of clouds, may lead to one model providing better forecasts of certain soaring parameters for your location, but that can only be determined through actual flight experience.

From an operational point of view, the fundamental difference is that the RAP is intended to provide more timely information, as suggested by its name, since RAP stands for "Rapid Refresh".  Hence every 3 hours it assimilates the most recent observations and provides updated forecasts - these forecasts are for 3 hour increments out to 24 hours, with additional hourly forecasts being provided out to 3 hours.  The NAM model ("North American Mesoscale") is intended to provide longer-term forecasts, out to 84 hours - but it assimilates new observational data and provides updated forecasts only every 6 hours, giving forecasts at 3 hour increments.

One important difference is that the horizontal resolution of the NAM model is much better than that of the RAP model, i.e. 12 km vs 20 km respectively.  Thus the NAM has the potential to provide more accurate forecasts when horizontal resolution is important, which is true for topographically-forced phenomena such as terrain-induced flow and wind convergences.  However, I do not have access to the full vertical resolution of NAM, due to NCEP's switch to data files of degraded vertical resolution, so its vertical resolution is not as good as for RAP. 
Cloud predictions are always difficult for a model, since clouds are often much smaller than the grid resolutions used.  The NAM possesses the advantage of allowing partial cloudiness to occur within a grid cell, whereas the RAP requires a cell to be either completely clear or overcast.

Processes which cannot be simulated explicitly by the fundamental equations, due to a lack of sufficient resolution, must be "parameterized", i.e. estimated from approximated (and often empirical) formulas.  One important example of such parameterization is cloud formation.  The models use different parameterizations and thus different forecasts are to be expected - but whether one is better than another can only be determined by empirical evaluation, which for BLIPMAPs means comparing to actual flight experience.  It is quite likely that one model may prove to be more accurate at one location than the other - and even possible at a given location that one model may be more accurate for one parameter and the other more accurate for a different parameter.  After all, if there were to be a parameterization which always produced better results then all models would be using it! 

General Differences:
Generalizing the discussion of the last section, any model difference can result in forecasts of one model being more accurate than the other.  For example, the hydrological cycle can be quite different between the models since they have different precipitation parameterizations, different treatments of the movement of ground water in the soil, etc..  Since thermal production is strongly dependent upon the amount of soil moisture used by the model, quite different thermal strength forecasts can result.  Soil moisture is very model dependent since there are no observations of that quantity to be assimilated into the model so there is no adjustment toward known conditions (Whereas for atmospheric moisture there are observations from radiosonde and satellites which can be used to drive the model toward a known reality.)

BLIPMAP Processing:
Because the NAM has higher resolution, and its grid domain is much larger than that of RAP, the files that BLIPMAP must download and process are 14 times larger!  This has obvious processing implications, among them being that the time between the availability of a model forecast and the appearance of the associated BLIPMAP being much longer for NAM than for RAP.

Having two model forecasts to compare allows one to better judge the uncertainly of any forecast - if forecasts from the two models agree then one has much more confidence in that forecast than if the two models differ.

Advantage of Multiple Models:
In addition to providing an evaluation of uncertainly, having two forecasts available allows each to provide a backup for the current day forecasts should the other not be available.

"But I just want to know which model is best"
Each model has different advantages and neither is always "best".  (Just as no one glider is "best" in all regards )  Those who have read NWS Forecast Discussion texts will have noted that they often say "model A says X and model B says Y" and then go on to come up with some synthesis forecast based upon their experience and their knowledge of each model's weaknesses and strengths - if one model was _always_ "best" then they would simply stop running the other model!  Here it can not be known apriori which will work best for a given soaring case, i.e. which model strengths/weaknesses are important in each circumstance, so that will have to be empirically determined.  Some people will say "but I only want to look at a single forecast", so what they can do is look at the RAP since that has some known history.  Those who are more interested in weather forecasting or are looking for day-before forecasts will use the NAM and I expect to hear of their experiences so that we all can gain a better appreciation of what model is best used under a given circumstance. 
Feb 2005 Addendum:  Based on the facts noted in the "Resolution" section addendum plus the fact that I've received more pilot reports citing cases where RAP forecasts have proved more accurate than I have for NAM I now recommend using RAP forecasts to evaluate the soaring potential of the current day.  But the "Resolution" concerns cited are theoretically-based preferences, i.e. not based upon model vs observation comparisions, and differences in the internal workings of the models can make NAM generally preferable to RAP for a specific location. 

What specific parameters differences might be expected ?
Theoretically, topographic influences, such as wind flow patterns, and convergence, whether forced by topography or other phenomena such as sea breezes, should be more accurately predicted by the NAM model due to its finer horizontal resolution.  But that is with "everthing else being equal" and there may be other factors involved.