Just wanted to update here with our results. We deployed our datalogging device in the NE Branch of the Anacostia River right next to USGS Site # 01649500 (http://waterdata.usgs.gov/usa/nwis/uv?site_no=01649500). We deployed here mainly so that we had an easy way to verify the accuracy of our sensors by comparing it to USGS’s data. We built the device for $275 + $6/month for cellular data. Our first deployment we were unable to get the stalker to sleep (not immediately sure why), so we just put a 15-minute delay in between sensor readings. The device lasted for 47 hours before running out of battery. You can see the data we collected at https://evh.herokuapp.com/devices/2. More info on the device itself can be found at: https://hackaday.io/project/8807-envirohub. We also wrote custom software for collecting and viewing the readings (which can be found here: https://github.com/kwyoung11/WaterLog).
If you compare our data to the USGS data for the site mentioned, you can see that our data is considerably off, although it looks like the trends in the data for the parameters collected are roughly similar. So next steps would include calibrating the accuracy of the sensors. We also accidentally fried the turbidity sensor we had because we were looking at the wrong data sheet, so that will hopefully be included in the next iteration. I am also planning on designing a custom enclosure for the device using tinkercad.com and 3D printing it. We ended up changing our design considerably from the sketch I posted and instead went with a floating water-proof box (using 4 small buoys). More pictures of this are in the hackaday link. However, this design is untested during high flows and its likely more visible and as such perhaps more susceptible to vandalism.