What they've done here is to combine the historical annual global average temperature record from the Hadley center (going back to 1850) with a recent MIT report on likely temperature increases through 2100 under projected emissions scenarios. The citation for that is A.P. Sokolov et al, "Probabilistic Forecast for 21st Century Climate Based on Uncertainties in Emissions (Without Policy) and Climate Parameters", Report No. 169 (Cambridge, Mass.: MIT Joint Program on the Science and Policy of Global Change, 2009), and can be downloaded from MIT (pdf format).
The reason for the upward-turning curve is the projection that, without an international policy agreement to limit fossil carbon emissions, coal and other fossil fuel use will continue to grow, particularly with economic growth in developing nations. Even if that growth didn't happen, our current rate of CO2 emissions is enough to keep temperatures rising for a long time to come, though on a lower curve than this one. And if we completely halted all CO2 emissions now, temperatures would still rise for a while because Earth is presently out of equilibrium, and there is still some warming expected "in the pipeline". Nevertheless, the sooner we get an international agreement in place, the sooner we can start bringing that crazy curve down, at least a little bit. One policy option is illustrated in figure 4 of the CBO report, to the right of this paragraph.
The report rightly emphasizes the point that uncertainty should move us to swifter action, because uncertainty in effects includes uncertainty on the side of things being that much worse than we expect. The consequences of no action will be devastating to the world our children inherit.