The Public Use Statistics Office, which manages a program for collecting, auditing, and analyzing visitation data provided each month by individual National Park System units, has released the Statistical Abstract 2010.  The 74-page document, which details National Park System visitation in 2010 as well as forecasts for 2011 and 2012, is  available as a pdf file.

This Natural Resources Data Series publication provides visitation statistics that are broken down by NPS unit, region, state, and several other categories, including unit type and population center type.  Also included are 11 tables (routine fare) and five figures (very legible).

This well-organized abstract is a treasure trove of attendance data that is about as accurate -- all things considered -- as you might reasonably expect.  The standard caveat applies, though; you just naturally have to be cautious when using attendance data.  It's probably not a good idea to put too fine a point on it.
 
While users will typically consult this statistical abstract on an as-needed basis, seeking this or that attendance figure as circumstances demand, data hounds (like yours truly) will want to take time to dig into it for interesting comparisons and odd facts or projections.

As your eyes sweep over the stats, question after question leaps to mind. The statistical abstract shows that visits to Mississippi's national parks declined by nearly 45% last year. Why was that?   Recorded attendance at Arkansas' Pea Ridge National Military Park surged by almost 68% (to 115,128) last year. Was that solely because the park switched from hand counting to using a traffic counter?  Last year the number of backcountry stays in Yosemite National Park was nearly triple that of Yellowstone National Park. Wow; I wouldn't have guessed that! And so on.

Try it for yourself.