I think I can explain the reason for the discrepancy from the lower pessimistic estimates. Most of the numbers in the 2 to 3 trillion range come about from Hubbert Linearization estimates taken as we near the peak.
That takes it closer to the 2.1 to 2.3 trillion barrels normally quoted by Peak Oil experts than the 3 trillion expected by more optimistic analysts.However, since HL does rely on a symmetric profile, out year data becomes inconsequential. You can see this if I try to do a Hubbert Linearization of the Discovery/Shock model data (actually the Discovery part only):
The red line extrapolation near the "halfway" point puts the URR at a little less than 2.6 trillion but the "hockey stick" or "dog leg" up pushes the actual URR out to 3.4. Note that this purely reflects the discovery model, which shows long tails caused by the apparent reserve growth discoveries.
To answer another question: in the previous post, I fit to the data set shown in this figure that Khebab posted at TOD, solving for CumulativeDiscoveries(2005) - CumulativeDiscoveries(1945) = D(2005) - D(1945), which gave me the 3.4 trillion.
Now that I see how close I am to the USGS estimates, and knowing how overly optimistically many analysts treat their interpretation, I get a little worried on the veracity of my own estimates. But then again we have to recall that the long tails have absolutely no impact on hitting a peak at this instant; it can only serve to delay the downturn if we can crank up the extraction rate. But does the proverbial world have the technical and political wherewithal to accomplish that? ... or do we instead choose a soft landing?