Question asked in this exercise:

How does ion principal component 2 at a well vary with the well’s distance from faults along the groundwater flow path?

In EX1, I used principal component analysis to evaluate how parameters accounted for variance between the wells I studied. Based on my knowledge of how chemistry varies as water flows through the basalt, ion PC2 accounts for variance caused by ion exchange between the basalt and the groundwater, with increasing sodium ion concentration/pH and decreasing calcium/magnesium ion concentrations as the water spends more time underground.

In this exercise, I estimated the groundwater flow directions in my study area using interpolation, calculated fault incidence direction, calculated angular difference between flow direction and fault direction, and then manual measurement of the distance between each well and the distance to the nearest fault segment that had flow direction within 45 degrees of parallel to the fault along the estimated flow path.

Name of tool or approach used:

Interpolation, reclassification, raster math, distance measurement in ArcGIS Pro

Methods:

Input data:

• 2018 static water levels in wells provided by Oregon Water Resource Department (OWRD)
• Well lithology from OWRD groundwater information system (GWIS)
• Well seal depth from well logs accessed through OWRD GWIS
• Fault polyline shapefile from Madin and Geitgey 2017
• Well locations from OWRD database, with ion concentration information based on my sampling in the summer of 2018

Steps:

1. Classified wells in the static water level dataset by the basalt formation that they were open to, based on lithology and seal depth. Excluded wells that lacked this information.
   1. Output: wells classified as open to Saddle Mountain Basalt (Smb) Wanapum Basalt (Wb), Grande Ronde Basalt (Grb), or both Wanapum and Grande Ronde Basalt (WbGrb). These classifications were joined to the static water level information.
2. Created an interpolation surface for static water levels of wells open to both the Wanapum and Grande Ronde. I ignored the Saddle Mountain formation, since the wells that I had sampled were not open to it. I used kriging with a cell size of 200 ft, and this created an estimated potentiometric surface for these wells. The interpolation was a bit ugly because my wells were not ideally distributed for this.
   1. I tried creating interpolations based on other combinations of formations to better approximate the potentiometric surfaces posited by past studies in the region. I ended up creating two potentiometric surfaces: one using wells that were only open to the Wanapum Basalt (Wb), and a second using wells that were open only the Grande Ronde Basalt as well as those that were open to both the Grande Ronde Basalt and Wanapum Basalt(WbGrb_Grbonly).
3. Calculated flow direction in the two aquifer groups – Wb and WbGrb_Grbonly
   1. Used the Hydrology toolbox to fill sinks and then calculate flow direction.
   2. This creates out a raster with eight possible values between 1 and 256.
   3. I then reclassified it so the values corresponded to the eight primary cardinal directions (N, NE, E, SE, S, SW, W, NW), which range from 0 to 360 degrees
4. Calculated fault incidence angle
   1. Added a cell to the polyline attribute table called “angle”
   2. Split the polyline at each vertex, which creates a new shapefile
   3. Used the field calculator and a python script to assign an angular value between 0 and 359 to each fault polyline segment
   4. Performed the polyline to raster function, using the angle as the cell value. I used a cell size of 200 ft.
   5. This created a raster where only pixels that include part of a fault polyline had values, and those values ranged between 0 and 359.
5. Subtracted the flow direction raster from the fault incidence angle raster, in order to create fault line pixels that had values that reflected the difference between the fault incidence angle and flow direction. I did this twice, once each for the Wb and WbGrb_Grbonly rasters
   1. Reclassified this raster so that pixels with fault direction within 45 degrees of parallel to the flow direction were 0, and the pixels with fault direction with 45 degrees of perpendicular to the flow direction were 1.
   2. I then added the WB and WbGrb_Grbonly results together, so that pixels with fault direction within 45 degrees of parallel to the flow direction in both were 0, and the pixels with fault direction within 45 degrees of perpendicular in either raster to the flow direction were 1, and pixels with fault direction within 45 degrees of perpendicular in both rasters were 2. I named this allwbgrb_ff.
6. On a map layout, I added my sampled sites with PCA data, the interpolated potentiometric surface for wells open to the Wanapum and Grande Ronde aquifers, and allwbgrb_ff.
   1. I added a field to the sampled sites with PCA data called “dist_from_fault”
   2. Using the measure tool, I measure the distance from each well along the path most perpendicular to potentiometric contours to the nearest allwbgrb_ff pixel with a value of 1 or 2.
   3. This was subjective because the potentiometric surface is imperfect because of the erratic spacing of wells. In areas where the potentiometric surface had noticeable glitches, I used my own judgement based on topography and literature about groundwater flow direction in the region.
7. I then graphed the “dist_from_fault” against the ionsPC2 category.

Discussion and Results:

I hypothesized that ion PC2 would increase with decreasing potentiometric surface elevation. An increased score in ionPC2 indicates an elevation sodium concentrations and pH and a decrease in calcium and magnesium caused by a progressive ion exchange reaction between the groundwater and the basalt. Because water flows from higher potentiometric elevations to lower potentiometric elevations, I would expect water samples from lower potentiometric elevations to show chemical evidence of increased interaction with the basalt. If this process of down-gradient groundwater flow were the only process influencing the ion exchange reactions, the well symbols on the map below would become progressively darker as the interpolated well level surface elevation decreased and the wells were further from the up-gradient recharge zone.

However, upon examining a map of ion PC2 values this is not the case – there are anomalously low values of PC2 in the valley, where one would expect to see increased values if groundwater flowed uninterrupted from the up-gradient recharge zones. In this study I introduced the variable of distance from faults in order to test another hypothesis: that faults compartmentalized groundwater flow, blocking lateral flow through the aquifer while promoting vertical permeability and modern recharge into the down-gradient aquifer. I also hypotehsized that if a fault was a barrier, PC2 values up-gradient of the fault would be elevated as the fault trapped water behind it. This would result in more chemically evolved groundwater backed up behind the fault, and less chemically evolved groundwater down-gradient of the fault.

The results of this study tentatively support the conceptual model of fault compartmentalization. In particular, water samples from wells in the valley down-gradient of the fault zones have evidence of less exposure to the basalt than wells further towards the recharge zones. 15 of the 18 wells sampled show a positive correlation between distance from a fault along a flow path and ion PC2 score, especially when graphed points are compared to their up-gradient and down-gradient neighbors (i.e. 57946 being up-gradient from 57235 and down-gradient from 54277).

Because of the width of the raster cells indicating faults and their flow direction in this model, four wells ended up have a distance from faults of 0. This does not seem unreasonable, because examination of exposed fault zones in the area indicated that many are up to a couple hundred meters wide. Additionally, while geological studies of the area indicate that the faults are close to vertical, their exact dip angles are unknown and this introduces a certain amount of uncertainty about their location at depth. Ion PC2 values of wells with dist_from_fault = 0 show an interesting dichotomy of either very high (4167, 4179) or very low (3929, 3962) values. I believe this indicates that wells 4167 and 4179 are up-gradient of faults that are acting as barrier, while 3929 and 3962 are slightly down-gradient or within the fault damage zone itself.

Critique of Method:

This is admittedly a crude method of estimating groundwater flow direction. However, a certain amount of estimation is necessary to model a system with relatively few and unevenly spaced measured water level points in a hydrogeologic regime with many individual water-bearing layers of lava interflow zones that are unpredictably connected. I wish I had been able to find an automated way to measure well distance from faults along the groundwater flow path, which would have taken away some of the subjectivity of measuring the flow paths by hand the old-fashioned way.

Because of the uneven spacing of wells used for interpolation, the flow direction rasters that I created have glitchy areas and some of these areas of physically improbable values influenced the Fault and GW Flow Direction raster. I reclassified that raster into broad categories to avoid creating a false sense of precision.