Methods considerations

I think this is a really interesting proposal, and well thought out. It's great to see the inclusion of metagenomics and antibiotic-resistance targeting. I think this will yield insights about our management of these facilities the way it is designed. Since you have already answered the broader concerns in another comment (Concerns with projected outcomes), I'll focus on a few methods issues that can be clarified or improved.

• You are targeting the 16S V4 region, but with the newest round of Illumina amplicon sequencing, we've begun extending primers from F515-R806 to F319-R806. This still allows some overlap between paired ends and gives a potentially more informative amplicon length. And I might recommend looking into the Fadrosh et al. 2014 method for adding a spacer to the beginning of the primers - this avoids the wasteful phiX spike necessitated by Illumina clustering algorithms. We've been experimenting with a version of this, and results are good so far. However, if you plan to run the amplicons and metagenomes in the same run, then nevermind - that probably fixes the same problem.
• You are going to use UniFrac for beta-diversity analysis. This is a great idea, and should certainly be part of the workflow. However, keep in mind that UniFrac excels at finding broad, phylogenetically-relevant habitat differences (e.g., acidic vs basic, or gut vs skin).
  But quite often microbial ecologists miss out on really interesting subtle habitat or source differences by eschewing traditional ecological beta-diversity metrics simply because they are not based in phylogeny. For example, when looking at skin samples from men and women, UniFrac sees very little difference because the same clades are represented in both sets. However, we keep seeing that minor taxonomic swaps hold the key to distinguishing between groups. Men might have one type of Corynebacterium enriched on their skin and women have another Corynebacterium. UniFrac overlooks this (by design, and for good reason), while another metric, like Canberra, will emphasize this difference. That is not to say that you should use one over another, but choose metrics carefully based on the differences you expect to see, or else risk missing out on interesting results.
• Minor point - QIIME is now up to version 1.8 and will likely change again before your sequences come out, so it is probably sufficient to just cite Caporaso et al. 2010, instead of version number.
• Aesthetic markdown point: The markdown editor seems to have hijacked your underscores in the pick_otus_through_otu_table.py description. Put backticks around it to avoid the italics.

This is a very cool idea, and I'm happy to help out where I can, or review a future version.