Journal of Undergraduate Research
Volume 7, Issue 1 - September/October 2005

A Method for Determining Groundwater Fluxes and Uncertainties in Measurement

Wendy Yankus

ABSTRACT

Groundwater is vulnerable to pollution from various sources, including underground storage tanks, landfills, and surface spills. Contaminants that leak into the groundwater form plumes, and their movement and concentration must be accurately measured before an effective remediation strategy can be put in place. Contaminant flux measurements are used to describe the rate of contaminant mass leaving a control plane perpendicular to the mean groundwater flow. Current methods of calculating contaminant flux in groundwater neglect spatial variations in contaminant concentration and groundwater flow. A new device referred to as a passive flux meter (PFM) has been developed by the faculty of the University of Florida to provide direct in situ measurements of both cumulative water and contaminant fluxes in groundwater. With this new technology it is possible to account for spatial variations and, using spatial statistical methods, calculate average flux and measurement errors.

INTRODUCTION

Groundwater contamination from underground storage tanks, landfills, and surface spills has been gaining more attention as contaminated water supplies and knowledge of their consequences have increased. Quality groundwater is essential, since it supplies much of the world’s water use, including drinking, irrigation, and industrial processing, and maintains the health of many ecosystems. As with the need to protect our water supply, there is also a need for better technologies to accurately describe and monitor subsurface pollution.

Contaminants such as trichloroethylene (TCE), perchloroethylene (PCE), petroleum products, and arsenic that leak into the ground move with the groundwater and disperse themselves in a plume. Once a contaminant plume exists, measurements of contaminant concentration and the movement of contaminants within the groundwater are needed to determine the most effective remediation strategy. Contaminant flux measurements are vital in predicting the risk of further spreading of the plume, as well as the progress of the cleanup effort. Currently, contaminant flux is approximated using calculated groundwater fluxes and depth-averaged concentration measurements gathered from monitoring wells. This method neglects spatial variation in contaminant concentrations and in groundwater flow, which reduces the reliability of model predictions. Also, contaminant flux values determined from monitoring well samples represent only current conditions rather than long-term trends.

By directly measuring both groundwater and contaminant flux, hydrologists could account for spatial variations and daily fluctuations in flow and contaminant concentration. This would be a great help in properly characterizing the site and implementing cost-effective remediation projects. The method of estimating contaminant flux from field measurements presented here involves spatially integrating point measurements of contaminant mass flux, Jc, as indicated in this equation:


Equation

where MQ is the contaminant mass discharge, Jc is the time-averaged mass flux per unit cross-section of aquifer, and dA represents an elemental area. This paper will discuss the method for determining contaminant mass flux and the uncertainties in interpolation techniques.

THE PASSIVE FLUX METER

The passive flux meter (PFM) is a new device designed to provide direct in situ measurements of both cumulative water and contaminant fluxes in groundwater. The device is a self-contained permeable unit that is inserted into a well or boring such that it intercepts groundwater flow but does not retain it (Figure 1). Inside the device is a matrix of hydrophobic and hydrophilic permeable sorbents that retain dissolved organic and inorganic contaminants present in fluid intercepted by the unit. The sorbent matrix is also impregnated with known amounts of one or more fluid soluble "resident tracers." These tracers are leached from the sorbent at rates proportional to the fluid flux.

The PFM is inserted into a well or boring and exposed to groundwater flow for a period ranging from days to months and then removed. Next, the sorbent is carefully extracted to quantify the mass of all contaminants intercepted by the PFM and the residual masses of all resident tracers. The contaminant masses are used to calculate cumulative and time-averaged contaminant mass fluxes, while residual resident tracer masses are used to calculate cumulative or time-average fluid flux. Depth variations of water and contaminant fluxes can be measured in an aquifer from a single PFM by vertically segmenting the exposed sorbent packing and analyzing for resident tracers and contaminants. Thus, at any specific well depth, an extraction from the locally exposed sorbent yields the mass of resident tracer remaining and the mass of contaminant intercepted, from which the local water and contaminant fluxes are calculated. Note that multiple tracers with a range of partitioning coefficients are used to determine variability in groundwater flow, which could range over several orders of magnitude. Existing monitoring technologies cannot provide cumulative water and contaminant fluxes without continuous, and therefore expensive, sampling.

The PFM possesses the advantage of providing a long-term monitoring solution that generates time-integrated estimates of groundwater and contaminant flux. Hence, transient fluctuations in contaminant concentrations and groundwater flows are not an issue of concern, compared with traditional monitoring methods, because such variations are directly integrated in flux estimates (Hatfield et al. 2004)

Schematic of a flux meter comprised of a permeable sock filled with a selected sorbent

Figure 1. Schematic of a passive flux meter (PFM) comprised of a permeable sock filled with a selected sorbent.

SITE

Several field tests were conducted with the University of Waterloo to validate the new device. The flux data presented here was gathered during the second of three tests. In this field test, PCE and TCE are the primary groundwater contaminants. The location is the forested research site at Canadian Force Base Borden located 150 km north of Toronto, Ontario. Site geology features a surficial sand layer approximately 3.5 m thick, which overlies a clayey aquitard.

The Borden site is a unique research facility established by John Cherry and the University of Waterloo research group. The site originally suffered contamination from a landfill, prompting initial research investigations. The results present herein involve contaminants introduced as part of multiple past and concurrent research projects.

Flux measurements were taken in a controlled release plume in which the University of Waterloo released a DNAPL mixture consisting of 45% PCE, 45% TCE, and 10% chloroform by weight. This mixture was released April 9, 1999 from a single release point located 1.8 m below ground surface and 0.9 m below the water table. In the area of release, the aquifer was approximately 3 m thick, consisting of fine-to-medium grained sand. The DNAPL release generated a dissolved plume approximately 80 m long that at one time was discharging into a small stream (Hatfield et al. 2004).

For this test, PFMs were used to measure water, TCE, and PCE fluxes in fence-row wells located 1 meter down gradient from the release point. Local TCE and PCE flux-averaged concentrations were obtained by taking the ratio of PFM measured water and contaminant fluxes (Hatfield et al. 2004).

METHOD

Flux measurements were recorded from fifteen PFMs each with approximately ten vertical segments. Despite the number of sample measurements, many areas exist where data is unknown and must be predicted to properly describe the contaminant mass flow over the plume transect. However, each prediction flux at an unsampled location has an associated error or uncertainty that must be considered in the integrated contaminant mass flow.

Tools of geostatistics, including semivariance models and kriging, will be used to estimate flux values at locations where data is unavailable, as well as the uncertainties in those predictions. A semivariogram is a geostatistical model of spatial data that is a function of distance and the variance between two locations. Kriging is an interpolation technique that uses values from the semivariance model as weights to predict the values at unsampled locations. Together these techniques use information about spatial patterns in data to make predictions at unsampled locations. With these predictions comes a measure of uncertainty, as determined by the following equation of estimation variance:

σE2 = -ϒ[(K),(n)] - ϒ[(n),(n)] + 2ϒ[(K),(n)]

where σE2 is the estimation variance, ϒ is the semivariance, K is the coordinate of a point of known measurement, and n is the coordinate of a point of unknown measurement (Journel et al).

Semivariance models of PCE were generated for the horizontal and vertical directions using all the data points and then concentrating on just the data within the plume. All models were created using GS+ software. Sample points were chosen midway between known measurement points. A MathCAD program was used to evaluate the estimation variance equation and its square root. The result is a value that describes the spread of values around the expected value (mean). A 95% confidence interval (+/- 2σE) was used to determine the percent of uncertainty in the prediction values.

RESULTS

PCE flux measurements were collected at 71 points over a transect 10.4 meters long. A histogram was created from the data (Figure 2), and the mean was calculated to be 110.05 x 10-5 mg/cm2/hr. A semivariogram analysis showed that the data is correlated at distances less than 48 cm (Figure 3). The estimation variance and standard deviation were calculated using the above-referenced equation, and the resulting values were 1.30935 x 10-7 mg/cm2/hr and 36.185 x 10-5 mg/cm2/hr. Using a 95% confidence interval, the uncertainty in measurement was shown to be ±65.8% of the mean flux value.

Histogram of PCE data
Figure 2. Histogram of PCE data.

Semivariogram of PCE flux data

Figure 3. Semivariogram of PCE flux data.

REFERENCES

  1. Hatfield, K; Annable, M.; Cho J.; Roa P.S.C; Klammler, H.: A direct passive method for measuring water and contaminant fluxes in porous media. Journal of Contaminant Hydrology 75 (2004) 155-181
  2. Journel, A.J; Huijbregts, Ch.J: Mining Geostatistics. Caldwell, New Jersey: The Blackburn Press, 2003.
  3. Kitanidis, P.K.: Introduction to Geostatistics: Applications in Hydrology. Cambridge, UK: Cambridge University Press, 1997.
  4. Roa, P.S.C; Jawitz, J.; Enfield, C; Falta, R.; Annable, M.; Wood, A.: Technology integration for contaminated site remediation: clean-up goals and performance criteria. Groundwater Quality: Natural and Enhanced Restoration of Groundwater Pollution 275
    (2002) 571-578

--top--

Back to the Journal of Undergraduate Research