I wouldn't be so certain. Of course any estimate would have some (huge) error bar, but to be useful it would be enough to have some argument for why the error stays constant between the calculations for two different years.

Then one could say whether it increases/decreases, and over time build some model off of that.