Graph comparing vote growth to voter registration growth in two counties in Florida. The lower leg on each line represents votes and registrations in 2000; the upper leg shows the incremental change required to get to the current vote totals, and the incremental change in registrations. The upper segments point upward relative to the lower segment, which means disproportionately more votes going to Bush. The lower segment represents voters who cast their ballots in the 2000 election, most of whom are likely to vote the same way as last time -- dramatic shifts in such a large group could not go undetected (i.e., a president whose popularity is surging).
The point is that the supposed upsurge in Bush votes relative to Kerry would have necessitated extremely abnormal patterns of voting behavior (i.e., large numbers of Gore voters switching to Bush, independents falling very heavily towards Bush) or, more simply, vote rigging.
We can ask ourselves, "Is it technically POSSIBLE that this could happen?" TECHNICALLY possible, yes...but we also have to shut our eyes to everything else, like exit polls that didn't agree (fell short in NUMEROUS states, by the way), the most energized left-wing in (at least) 36 years, a president with falling approval ratings at 45%, the simultaneous lack of accountability in vote counting.
Graph below depicts top 9 Florida counties. All of these are suspect except Miami-Dade. Estimated total votes skimmed from these counties alone: 82,036; or a swing of 164,072.
The attached graphs use data I downloaded from the Florida Department of Election site. My hope is that this format makes the information more readily understood.