Most bacterial genomes have roughly 8 copies of the 16S rRNA gene, I don't think the copy number varies a lot.  But I would be more concerned with how you count the different sequences obtained, and with differences in efficiency of amplifying the gene from different species, than the starting copy number. 

If you go to "The Ribosomal RNA Database" at https://rrndb.umms.med.umich.edu/  you will see the complexity of the problem you face. Just looking at the Bacteria (not Archaea) you get the following figures:

• Number of genomes analysed: 2865
• Range of 16S copies: 1 - 15
• Mode: 2
• Median: 4
• Mean: 4.12
• Standard deviation:2.75

So, you could decide to assume that each bacterial cell carries 4.12 copies on average. However, different composition of bacterial communities will skew the distribution from the one shown above. For instance if the bacteria are mainly from the family Bacillaceae, the mean number of 16S copies for that family is 10.55; conversely, the family Halobacteriaceae has a mean of 1.89 copies. Perhaps you could work out the figures if you had information on the composition of the communities you are testing.

As Brian says above, you also will have issues with amplification efficiency for real-time PCR.

I can imagine methods based on digital PCR, either digital droplet PCR or digital chamber PCR. That would get round issues of efficiency of amplification. However, it requires access to the hardware; these machines are not yet very common, so that might be an issue. You also still need to know average copies of the target per cell.

Thanks for correcting me, Vincent.  I had not gone to the rRNA database and I am only familiar with a dozen or species of bacteria with complete genomes done.   I should not have said "most bacteria" when I was only looking at a very few species.