Lesson 3 continued...

3.3. Pump and Treat Remediation

The most often used method for contaminant remediation in groundwater is pump-and-treat. Figure 3.21 shows the Saint Joseph site with two vinyl chloride plumes, one going toward Lake Michigan and the other toward Hickory Creek. The plume that was modeled by Tiedeman and Gorelick (1993) was the one heading toward Lake Michigan (outlined in red with the concentration contours shown). The goal of their analysis was to hydraulically contain this plume while minimizing the total groundwater extraction rate. In order to reach this goal, Tiedeman and Gorelick (1993) designed optimal plume containment strategies with the steady-state model, MODFLOW, and the nonlinear optimization solver, NPSOL. They designed scenarios with 90, 10, 2, and 1 extraction wells with each well assigned a specific location and extraction rate. A similar approach will be taken here for the 10 and 2 extraction well solutions only. For detailed information on Tiedeman and Gorelick’s (1993) work, review their paper.

Figure 3.21. The plume of concern is outlined in red, heading towards Lake Michigan (Tiedeman and Gorelick, 1993).

Model Setup

To replicate Tiedeman and Gorelick’s (1993) optimization of the 10 and 2 extraction well solutions,we used the IGW model with the final baseline values (Table 3.3) determined in Section 3.2. The vinyl chloride concentration contours shown in Figure 3.21 were added to IGW to represent the plume (see the Tutorial for details). A given retardation coefficient (R) of 1.40 (Tiedeman and Gorelick, 1993) was used to calculate the solid-liquid partitioning coefficient (K_d ) for vinyl chloride using Equation 2.1. K_d was calculated as 4.50x10-8 m3/g using the given IGW bulk density value of 2.65 kg/L. This K_d value was then added to the K1 zone.

Equation 3.1. Retardation coefficient (Bedient et al., 1999)

The retardation value indicates slowed transport of the contaminant with groundwater because it is sorbed onto aquifer solids. A value close to 1 indicates that the contaminant will travel at the groundwater velocity, while a value of 2 indicates that the contaminant velocity would be half of the groundwater velocity.

When optimizing the extraction rate at a given well, one must take into consideration a set criterion for drawdown. Extensive drawdown within a well is of concern because there will be a loss in extraction efficiency and an increase in the cost of remediation. Tiedeman and Gorelick (1993) set a drawdown criteria of maintaining at least 1 meter of head within each well screen above the aquifer bottom. In this lesson we also use this value as the drawdown criteria. One difference between MODFLOW and IGW is that in IGW there is full penetration within the water table aquifer system while in MODFLOW the wells penetrate only the lower 5 m layer of the aquifer. According to Tiedeman and Gorelick (1993), the difference in fully or partially penetrating wells has little impact on the pump-and-treat remediation of the plume.

The general location of the extraction wells for the two and ten well solutions are shown in Figure 3.22 with the corresponding total extraction rates shown in Table 3.5. The reliability level shown in this table is a measurement of the potential failure of the system and was determined with Monte Carlo simulations (Tiedeman and Gorelick, 1993). For example, a 90% reliability level for the ten well solution indicates that at least 90% of all simulations performed with the specified extraction rates contained the plume when Tiedeman and Gorelick (1993) varied any parameter.

Figure 3.22. Location of the wells used for the two and ten well solution (Tiedeman and Gorelick, 1993). See text for explanation of symbols.

For the ten well solution, the wells were placed where the filled (extraction wells 13 and 14) and unfilled (extraction wells 1 through 5 and 15 through 17) circles are located. For all reliability levels, extraction wells 1 through 5 accounted for 92% of the total extraction rate, 13 and 14 were not used, and 15 through 17 extracted only 8%. For the two well solution, the wells were placed where the boxes (extraction wells 12 and 17) are located. For a reliability level of 50% and 60%, extraction well 12 extracted about 97% and 88% of the total extraction rate, respectively, with extraction well 17 pumping the rest. Since the exact position of each extraction well was not given by Tiedeman and Gorelick (1993), the location of the each extraction well in IGW was approximated using the map shown in Figure 3.2.

Next, a sensitivity analysis was completed on time interval (t) and mesh size (x) that would provide the best accuracy and simulation time. To determine the appropriate t and x values, the Courant Number (Cr; see also Lesson 2) number was calculated using a conservative velocity value of 0.77 m/day, which is within the top 75 percentile of the highest groundwater velocities found within the location of the plume. Three simulations were conducted for the ten well solution with a 50% reliability level with different t and x values. The simulation results of the concentration breakthrough curves for each monitoring well were then compared. It was determined that a t = 10 days and x = 9 m (NX=268) produced a relatively accurate and fast simulation. Therefore, these values were used for the IGW pump-and-treat simulations.

Table 3.6 shows the data input for the ten and two well solutions. Each well has a corresponding monitoring well to collect the concentration breakthrough curve data necessary to determine the total mass collected and the initial and final head data needed to confirm whether the drawdown criteria were met. The extraction well and corresponding monitoring well are placed in the same exact location and given the same screen interval to ensure that the aquifer is fully penetrated. For detailed information about the pump-and-treat setup and procedure refer to the Saint Joseph site tutorial. Details on the simulations of the ten and two well solution will be discussed throughout the section.

Ten Well Solution

Figure 3.23 shows the basic setup for the ‘ten well pump-and-treat solution. As noted above (Figure 3.22), the optimal solution ends up using only 8 extraction wells (Tiedeman and Gorelick, 1993).

Figure 3.23. Placement of eight wells for the ten well solution.

The plume contours are shown with 800 (red), 100 (light blue), and 1 (dark blue). The three extraction wells up-gradient of the plume (extraction wells 15 through 17) were placed so that they establish head gradients that force the plume toward the five down-gradient extraction wells (wells 1 through 5). This is a good arrangement because the three up-gradient extraction wells both remove contamination at high concentrations and direct the remaining plume toward the other extraction wells. The five extraction wells are placed down-gradient at the nose of the plume to reverse the natural gradient and to capture the rest of the plume.

The optimal total extraction rates for the ten well solution with a reliability level of 50% and 90% (Table 3.5) were simulated using IGW. Notice that the reliability level increased from 50% to 90% when the total extraction rate was increased by almost 4%. The increase in flow rate increases remediation cost, but may be reasonable considering the added insurance of capturing the plume.

50% Reliability Level

Figure 3.24 illustrates the head contours and flow distribution produced from the extraction rates used for each well with the ten well solution and a reliability level of 50% (Table 3.7). The simulation was used to create a cross-section of extraction wells 1 through 5 (red) and another cross-section of extraction wells 15 through 17 starting at Lake Michigan and extending to the lagoon (brown) (Figures 3.25 and 3.26). The red line represents lines of constant head and the blue arrows represent the direction of groundwater flow.

Figure 3.24. Head contours and flow distribution with well cross-section locations for the 50% reliability level.

In Figure 3.25, notice that the head contours get closer together and the groundwater velocity vectors (i.e., blue arrows) increase in size when approaching the wells, indicating that the extraction wells are extracting groundwater at a significant rate. The bottom of the aquifer is represented as grey with black diagonal lines.

Figure 3.25. Cross-section of extraction wells W1 - W5 showing head profile and direction of groundwater flow for the 50% reliability level.

Starting on the right hand side of Figure 3.26 and moving left, take note of the following features. Near the lagoon, the groundwater velocity vectors flow towards the bottom of the aquifer. This is due to the influence of recharge. Near extraction wells 15 through 17, the water table goes down (head value decreases) and there is a vertical downward velocity gradient between the wells. Just to the left-hand side of extraction well 15, a stagnation point develops and the head decreases rapidly flow approaches extraction wells one through five. Another stagnation point develops down-gradient of the five wells -- this causes the plume to be extracted by the five wells and not flow into the lake.

Figure 3.26. Cross-section showing extraction wells W1-5 and W15 through W17 starting at the lagoon and ending at Lake Michigan for the 50% reliability level.

Figure 3.27 illustrates the remediation of the plume with the ten well solution and a reliability level of 50%. As noted above, extraction wells 1 through 5 significantly influence the head contours and groundwater flow and cause a groundwater divide to form down-gradient from the wells. This causes the plume to travel towards the five wells and away from Lake Michigan. Notice the change in contaminant concentration through the years. The plume travels directly into extraction wells 1 through 5, while extraction wells 15 through 17 seem to have little effect on the plume. However, as noted above, the up-gradient wells extract enough groundwater to cause the head contours to direct the plume toward extraction wells 1 through 5. The majority of the contamination with high concentration is remediated after 8 years, with an additional 4 years of extraction required to meet drinking water quality standards of 5 for vinyl chloride. It took 12 years to remediate the plume using the ten well solution with a reliability level of 50%.

Table 3.7 displays the extraction rate used for each well and the resulting head. The data collected from the concentration breakthrough curves for each monitoring well was used to calculate the mass removed by each extraction well. The total mass removed from the aquifer was determined to be 92 kg compared to the initial total mass within the aquifer of 77 kg. The reason for this difference is unknown, but might be associated with the accuracy of the t and x values used for the simulations. The pumping wells cover such a small area that the x may be too large to accurately produce the concentration breakthrough curves.

To determine the influence of the extraction wells on the head gradient and groundwater velocity, tracking particles were added to each extraction well. The particles' pathway is perpendicular to the head contours and parallel to the groundwater velocity. Figure 3.28 illustrates the pathway of a particle starting at a given extraction well and then running the simulation in reverse time (i.e., backwards arrow in IGW in the palette button section). At time 0, the particles start at each extraction well. At time –4 years, notice that the distance traveled by the particles starting from the three up-gradient extraction wells is about 3 times shorter than the distance traveled by the particles starting at the five down-gradient extraction wells. This occurs at all times because the extraction rate of the five wells is about 6.5 times more than the three up-gradient wells. The five extraction wells also have a larger area of influence due to their high extraction rates, ensuring that the plume will be captured. The three up-gradient extraction wells influence a smaller area, but their main purpose is to direct the plume towards the five down-gradient wells by changing the head gradient. It takes about –24 years for the particles to reach the lagoon. The result indicates that if the lagoons and vadose zone under the lagoons continued to release contamination, the contamination heading towards Lake Michigan would be effectively captured.

Figure 3.28. Particle tracking that shows the capture zones that develop with a ten well solution with a 50% reliability level.

90% Reliability Level

Figure 3.29 illustrates the head contours and flow distribution produced from the extraction rates for the ten well solution and a reliability level of 90%. The total extraction rate for this solution is higher, with most of the increase occuring at the five down-gradient extraction wells. The simulation was used to create a cross-section of extraction wells 1 through 5 (red) and another cross-section of extraction wells 15 through 17 starting at Lake Michigan and extending to the lagoon (Figures 3.30 and 3.31).

Figure 3.29. Head contours and flow distribution with well cross-section locations for a 90% reliability level.

The IGW cross sections allow for a comparison of the head countours and flow directions for the 50% (Figure 3.25 and 3.26) and 90% reliability solutions (Figure 3.30 and 3.31). Figure 3.30 shows an increase in drawdown and a shift in the head contours closer to the extraction wells compared to the 50% reliability solution (Figure 3.25). Figure 3.31 also shows a shift in head contours between the right-hand side near the lagoon and the three up-gradient extraction wells. Also, it shows a sharper decline in the water table between the three up-gradient wells and the five down-gradient wells and the development of a more pronounced stagnation point to the left of the five extraction wells. All of these changes are caused by the higher total extraction rate.

Figure 3.30. Cross-section of extraction wells W1 - W5 showing head profile and direction of groundwater flow.

Figure 3.31. Profile of extraction wells W1-5 and W15 through W17 starting at the lagoon and ending at Lake Michigan.

Figure 3.32 illustrates the remediation of the plume for the ten well solution with a 90% reliability level. A comparison of this figure to the results of the 50% reliability level (Figure 3.27) shows that the total remediation time decreases from 12 to 10.5 years and that there is an increase in the total extraction volume from 1.9 to 2.3 million m³. When comparing Figure 3.32 at 10.5 years to Figure 3.27 at 12 years, notice that the head contours have changed significantly around the five down-gradient extraction wells. In Figure 3.32 (10.5 years), the head contours have shifted further west than in Figure 3.27 (12 years) and there is an added head contour around wells 2 through 4, indicating that there is an increase in the total extraction rate. These two figures also show that the stagnation point to the west of the five extraction wells has shifted even further west. In Figure 3.32 down-gradient of the plume, the groundwater velocity (blue arrows) tends to flow towards the five extraction wells in comparison to Figure 3.27, which shows more groundwater by-passing the wells and going into lake.

Both the head contours and groundwater velocity show that with the increase in reliability level from 50% to 90%, there is added assurance that the extraction wells will captures the plume. Figure 3.32 illustrates that the majority of the contamination with a high concentration is extracted within 7 years with an additional 3.5 years necessary for concentrations to meet drinking quality criteria. Interestingly, the plume seems to be completely remediated around 10 years but some contamination with a concentration higher than 5 was detected and then removed a half a year later.

Figure 3.32. Remediation of the plume with a ten well extraction solution with a 90% reliability level.

The final head left in each well and the mass removed is shown in Table 3.8. With an increase in reliability from 50% to 90%, the total flow rate increased from 442 to 613 m³/day with a slight increase in total mass removal of 92 to 94 kg, respectively. Notice the increase in the amount of mass removed from the center wells compared to the wells on the edge of the five down-gradient extraction wells, shown in Figure 3.32 and Table 3.8. Well 15 removed more mass in comparison to the three up-gradient wells, because it is located within the highest concentration contour and has more time to extract more of the contaminant.

Figure 3.33 illustrates the particle tracking for the ten wells solution with a 90% reliability level. The pathways of the particles coming from each well are similar to Figure 3.28. The most noticeable difference is the increase in area that the particles cover (i.e., capture zone) at –24 years in Figure 3.33 compared to –24 years in Figure 3.28. When taking a close look at the head contour line in each figure, notice that the head contours in Figure 3.33 bend further inward towards the five down-gradient extraction wells, causing the pathway of the particles to spread out and creating a larger capture zone.

Figure 3.33. Particle tracking that shows the capture zones that develop with a ten well solution with a 90% reliability level.

The best reliability level to use for the ten well solution depends on the objectives and criteria developed for the site. For example, if reliability level and time were the most important factors, than the 90% reliability level should be chosen. However, the 50% reliability level extracts 171 m³ of groundwater less per day than the 90% reliability level option. Therefore, if the cost of groundwater treatment is the primary concern, the added 1.5 years of remediation acceptable, and intrinsic bioremediation proven, than perhaps the 50% reliability level would be the best solution.

Continue to the Two Well Solution

Back to top