This really belongs over in the Ask/Tell forum... To better understand how they come up with the track, always read the discussion, not just the advisory. They often explain why they are deviating from model consensus. A lot of the time it is just because they know the models flip back and forth, so they apply averaging over time and they know some models do better than others in certain situations, and they compare previous model runs with what the storm actually does. Each model has its own personality. Some are better at some things than others. Many of them are fed (at least partially) from output of other models. The more detailed recon data input into the models, the more accurate they are. Models are also tweaked from time to time (some more often than others) or the data input into them is massaged differently. There are also times when there is something wrong with the data and the models don't get initialized properly, which results in inaccurate results, which in turn might get fed into other models.

Bill