Get the information on the glyphosate label, it should already have some chronic and acute testing relative to various non-target organisms. I would suggest spend at least as much time on BMPs to limit direct spraying of streams and open water. Technology exists if you have your areas for aerial spraying in GIS, that appropriate computer with GIS and GPS interface can control spray apparatus to turn off before streams if you can estimate their extent and they are not evident. Glyphosate binds quickly to soil and is seldom seen as a major fishery problem if application limits direct spraying of waters, which can affect aquatic plants. If you are working also with insecticides, you might look over the 1984 report of our monitoring of water quality relative to the spruce budworm treatment project where we treated 1/2 million acres and also reported on effects of several accidental spills (in researchgate). If you intend to do this for research or possible legal reporting, your lab should become certified for this testing, keep chain of custody of sampling, etc. I was well known for my expertise on using fluorescent dyes as a tracer for herbicide and insecticide aerial spraying operations, as helpful in estimating which samples should be sent to lab for testing. Some of my papers and letters of findings are quite old and have not been integrated into my researchgate. My papers using stream ordering were a result of my early work in Oregon in trying to predict which streams were flowing and needed protection in project planning for aerial spraying operations.