Bacterial counts on environmental samples such as the rhizosphere can be very tricky. Plate counts obviously underestimate, and will completely miss more fastidious genera. There is the difficult question of VBNC bacteria, intracellular and biofilm format bacteria. The tentative conclusion we've come to over the years is that an orthogonal, systems biology approach will help you define the degree of complexity, convergent data may narrow the total count, but the confidence limits on any single analysis will often be quite wide.

NB it makes sense to express counts in log 10 format

Hi Dale. First of all, I do not understand really well your problem. Are you measuring the bacterial abundance by NGS or by other methods? 

In any case, you have to consider that the bacteria abundance in a microbial community rarely follows a normal distribution. In fact, it is more close to a lognormal or a Mandelbrot distribution than to a normal.

If your concern is if you have to use absolute or relative counts, that's another kind of question. You have a lot of samples, and you can see if your absolute number of total bacteria seems to be more or less equal among them. If it is the case, you could use absolute counts.

But if you have really different absolute counts, you should use relative counts. The problem with this kind of data is that you could have compositional effects that can bias your conclusions. There is a bioinformatic tool that, a priori, can diminish those compositional effects:

http://psbweb05.psb.ugent.be/conet/microbialnetworks/sparcc.php

I hope this helps.