It's not an assumption, it's an estimate, with the requisite error bars. But the estimate is based on analysis of data -- sometimes direct measurements, other times "economic" analysis. The air-sea flux is VERY well constrained by thousands of at-sea measurements, for example.
The trouble is, I have yet to see any isotope analysis that yields any numbers for the human component.
I found a few papers that allude to it, but its wrapped up in a bigger picture presentation and its complex. I'll keep looking. Looking...
I assume you found/saw this:
http://www.grida.no/climate/ipcc_tar/wg1/108.htm (can't wait to see the corresponding section in the new report!)
Here's a couple of references to peruse:
Isotopic Simple Global Carbon Model: the Use of Carbon Isotopes for Model Development (real interesting conclusion in this!)
Science reference, sorry, but maybe you can get it:
Oceanic Uptake of Fossil Fuel CO2: Carbon-13 Evidence (April 3 1992 issue)
Global Carbon Cycle (this is real simple, but the data in the lower-right hand corner of Figure 7.02 might be what you're looking for)
http://www.holivar2006.org/abstracts/pdf/T3-032.pdf (neat poster -- actual number, del13C decreased 0.44 per mil from 1979-2000, annual rate 0.0021, similar to Quay et al. referenced above)
Gotta keep this one in my pocket for drrocket: δ13C in CO2 at Mawson, Antarctica (wow, now I know how to do the del symbol!)
That's a start.
Some good links, although I would like to know Quay's assumption for carbon cycle time or seawater diffusion constant. I noticed in the PDF poster there is typo in your post: 0.021 decrease per year, not 0.002 like you and I thought. That's a lot more than the anthro contribution that I calculated (0.004 per year) from the NASA carbon numbers (quite simply, 6 parts anthro into 1560 parts seawater/atmosphere). This reversal means that seawater is absorbing a lot more anthro CO2 than I outlined in post 7.