GIS Groundwater Modeling Project
In 2018, students from Vanderbilt University conducting soil analysis at the Tennessee State University agricultural research farm discovered hydrocarbon contaminants in the soil. Since the initial discovery, several soil cores and water samples have been taken at the site, further confirming the presence of the pollution (Fig. 1). Based on initial analyses of the site, it is believed that groundwater plays a primary role in defining how the hydrocarbon is distributed throughout the area. The purpose of this analysis is to create a spatial model of groundwater flow at a contaminant site to better understand how the contaminant might have spread originally, and to help explain and predict the plume extent.
I became interested in this site when I was taking Vanderbilt’s geochemistry class last year. The cohort who had taken the class the year before I did discovered the plume without even looking for it: in contaminated areas, the soil just smells like gasoline. My class (2019) did some work on soil coring in this area and trying to map out the extent a bit more. I wanted to continue on with this research after the class, and have done some further work sampling soils and water.
This area is close to a community garden and to the Cumberland river, so figuring out how far it might extend has some real-world value. Additionally, understanding how the plume would have moved through the site is important for figuring out where the plume came from, and by extension, who is responsible for the spill.
It’s cool you’re continuing this past what we did in class. Hopefully you’ll be able to figure out where the contaminate came from. Based off of your understanding of the area, do you think the tool accurately predicted the flow?
I don’t know that I can really say with certainty how accurate it is, but I think it does make sense with some of the patterns that we’ve seen in the distribution of the plume. Earlier on, we were thinking about whether the fuel tanks (NW) or the fueling station (E) would be a more likely source, and I think both the coring and this model kind of imply that it came from the E or NE.
I think it is very interesting to investigate this real-world problem that is really near you. I love your way of research design as well as your cartographic/poster design. I think it could be useful to decrease your cell size so that you can generally estimate/decide which direction the contaminate comes from. Another way of increasing the prediction accuracy of the flow is to make the porosity value and transmissivity value anisotropy, not a constant through the surface, in order to predict the hydraulic head more accurately. Thanks!
Thank for the suggestions! This was a bit unclear in my poster description, but I did include some variation for porosity and transmissivity values over the area, based on the boundaries of different soil series.
Well, it looks like this turned out really well! I love the symbolization of the flow direction raster with the arrows. That’s super cool! How did you do it?
I also like that even your first-pass model seems to predict both a likely source and explain the current extent of the contamination to the south and west. That seems like a pretty good result, and will only improve as you improve your data sources. Will you be sharing this with the professor from the class? If you (or other classes) continue to do work out there, you’ll have to let me know how your model compares to the new data coming in. Excellent work.
It also occurs to me that there is a tool in GRASS GIS named r.drain, that may be interesting to you. It takes as inputs a “cost” raster, a flow direction raster, and starting points. Its output is a line that shows a pathway from that point following the flow. For you “cost” would just be a uniform raster (value 1 or something like that), flow direction you have from your model, and you could make starting points along the road and/or places where your cores show contamination. This would show you more clearly the paths the contamination is moving along and where it might be headed next.
I couldn’t find an equivalent tool in ArcMap unfortunately, but GRASS is free if you decide you want to play around with it. Just something to explore if you decide you want to keep working with this in the future. Never hesitate to reach out!
Thanks James! The output for the flow direction raster was in compass degrees, so I switched the symbology from stretched to vector field. The professor from my geochem class is my research advisor, so I’ll be sharing this with her. Thanks for the suggestion on the GRASS tool, I’ll check that out!
Really well presented, and it appears very useful results in and of themselves, and also for directing where next cores should be taken–I can imagine one or two more iterations of coring to specify further the contaminant source and dispersion based on these initial results. Great that you incorporated tools beyond what we covered in class and derived reproducible estimates. A local topographic survey would also likely improve the model. That could be done by surveying with total station or RTK GPS, just shooting a lot of elevation points, then interpolating them into a DEM, as we did for the Malata viewshed practicum. It will be interesting to hear what your professor thinks and hopefully you can continue this project–looks like the beginnings of a thesis to me.