Final Report: Riverbank Filtration Effectiveness in an Arid Environment

EPA Grant Number: R829009
Title: Riverbank Filtration Effectiveness in an Arid Environment
Investigators: Langford, Richard P. , Pillai, Surresh , Schulze-Makuch, Dirk
Institution: The University of Texas at El Paso , Texas A & M University
EPA Project Officer: Page, Angela
Project Period: September 1, 2001 through August 31, 2004 (Extended to August 28, 2005)
Project Amount: $437,418
RFA: Drinking Water (2000) RFA Text |  Recipients Lists
Research Category: Drinking Water , Water


The objective of this research project was to determine whether bank filtration in an arid environment is effective at removing microbial pathogens. The field site for the proposed studies is unique in that it is located in an arid region and utilizes the Rio Grande River system, which exhibits significant fluctuations in both water quantity and quality. The results from these studies can be significant for this region because the region relies almost exclusively on aquifers for its municipal drinking water.

Summary/Accomplishments (Outputs/Outcomes):

Site Characterization

A three-dimensional model was created and calibrated using a combination of cores, Ground Penetrating Radar. The geology of the aquifer consist ed entirely of fine- to coarse-grained sand that comprised two major aquifer units. Each is the fill of an ancient Rio Grande Channel and becomes finer grained upward, where the upper (unconfined) and lower (semi confined) aquifer units are separated by laminated fine and very fine-grained sand comprising a semi-confining (aquitard) layer. The coarsest and cleanest sands, which would make the best aquifer, are found in the lowest parts of each channel, in the area near the water table and at 5 to 6 m depth. Salts are concentrated in the upper 60 cm of the soil and are associated with the driest sediments. The coarser grained sediments will focus the flow of the water and are believed to provide conduits for faster transmission of pathogens in the bank filtration system.

Figures 1 and 2 illustrate the final characterization of the site. Key features are the high conductivity pathways illustrated in Figure 1 and the low conductivity horizon illustrated in Figure 2.

We used the results of the multi well tracer test to map paths of higher hydraulic conductivity (Figure 1). The result of tracer and aquifer tests demonstrated that only approximately one-half of the aquifer was acting as a conduit to the well. Therefore, even though the aquifer appeared to be relatively homogeneous, preferred pathways resulted in both more rapid transport of tracers and pathogens and also in dilution of water flowing to the well with water that had a much longer (weeks) residence time in the aquifer.

Figure 1. Fence Diagram of Radar Lines and Cored Wells Showing the Distribution of Channels and Facies at the Study Site.

Figure 1. Fence Diagram of Radar Lines and Cored Wells Showing the Distribution of Channels and Facies at the Study Site. The arrows in the map and fence diagram show preferred flow paths as determined from tracer tests conducted with plastic microspheres. Note that preferred flow paths lie in the orange and purple units.

Aquifer properties of specific zones at the test site were determined from multi well aquifer tests conducted in March 2002. The water within the shallow zone occurred under unconfined conditions, within the middle layer as leaky-semi-confined, and under semi-confined conditions in the lower part (Table 1).

Figure 2. Cross Section in the Study Site From the Stream to the Pumping Well Showing the Observation Wells and the Subsurface Layers

Figure 2. Cross Section in the Study Site From the Stream to the Pumping Well Showing the Observation Wells and the Subsurface Layers

The water table was near the base of the upper aquifer during the test so its properties could not be calculated. Laboratory tests, however, of vertical hydr aulic conductivity and the results of the tracer test indicate that its properties are similar to those of the lower aquifer unit.

The estimated aquifer properties of the lower part of the aquifer with a thickness of 20.8 m (68.3 ft) from 3.2 m (10.5 ft) to 24 m (78.7 ft) below ground surface using Neuman’s 1974 solution were: (1) transmissivity is 0.0064 m2/s (= 5972.9 ft2/day); (2) storativity is 0.004 (dimensionless); (3) specific yield is 0.05 (dimensionless); (4) horizontal hydraulic conductivity is 0.00031 m/s (= 87.52 ft/day); and (5) vertical hydraulic conductivity is 0.00015 m/s (= 43.84 ft/day).

The estimated aquifer properties of the intermediate semi confining layer from 2.4 m (7.9 ft) to 3.2 m (10.5 ft) below ground surface with a thickness of 0.8 m (2.6 ft) were estimated by averaging two values of estimation using Hantush-Jacob’s 1955 manual curve matching. The second estimate was from an automatic curve matching solution produced by AQTESOLV™ for Windows. The results were: (1) overall horizontal hydraulic conductivity is 0.000033 m/s (= 9.45 ft/day); (2) transmissivity is 0.00003 m2/s (24.76 ft2/day); (3) storativity = 0.000091 (dimensionless); (4) leakance is 0.0000000001 [(m/s)/s] or 1/s { 0.00016 [(ft/day)/ft] or (1/day)}; and (5) vertical hydraulic conductivity of the semi confining layer is 0.000014 m/s (4 ft/day).

The vertical hydraulic conductivity for bed layer underlying the stream (colmation layer) was measured using constant head method. The value is 0.000003 m/s (0.8 ft/day) by taking the geometric mean of samples from different depths in the clogging layer underlying the stream.

Table 1. Summary of Initial Estimates of Hydraulic Conductivities of Aquifer Layers for a Subsequent Groundwater Flow Model.

Model Layer

Vertical Hydraulic
by Laboratory
m/s and (ft/day)

Vertical Hydraulic
by Aquifer Test
m/s and (ft/day)

by Aquifer Test
m/s and (ft/day)

Shallow Aquifer Unit

5.0x10-6 (1.43)

1.5x10-4 (43.84)

3.0x10-4 (87.52)

Middle Low-Conductivity Layer

2.0x10-6 (0.57)

1.4x10-5 (4)

3.3x10-5 (9.45)

Deep Aquifer Unit

2.9x10-5 (8.25)

1.5x10-4 (43.84)

3.0x10-4 (87.52)


Geochemical parameters at the field site remained relatively constant during the 3-year sampling period. No significant change was observed during the weeks-long sampling events. The main solute in the ground water is sodium chloride and calcium sulfate with a total dissolved solid content slightly above 1000 mg/L. Biological content in well water is present, but typically low in biomass and not a result of fecal contamination. Nutrients are scarce in the ground water. The channel, however, is a source of high Kjeldahl nitrogen and elevated phosphorus, when channel water is drawn into the ground water. Overall, except for Well O, there was a slight long-term decrease in total dissolved solids over the sampling period. Well O, which showed an anomalous abundance of microspheres in the Year 2 tracer test, also exhibited episodic high salinities possibly caused by a salt-rich (NaCl) layer in the subsurface.

Determination of Dilution of Stream Water and Travel Time of Water to Stream


Many riverbank filtration sites were established with wells situated near a stream. Almost the entire flow from the wells was commonly assumed to have derived from the stream. Two key parameters in evaluating the effectiveness of riverbank filtration are the effects of travel time and dilution. Tracer tests, combined with modeling, can provide an estimate of travel along individual flowpaths. Temperature provides a better estimate of the overall travel time from the stream to the well and of the dilution of the stream water with groundwater. Recent studies, however, have show n that a highly varying proportion of the water produced from the well (up to 80%) was derived from groundwater rather than the stream. The overall effect o f reduction in numbers of pathogens caused by this process is insignificant beside the several orders of magnitude in reduction required for pathogens. The dilution, however, must be quantified to create a groundwater flow or contaminant transport model.

During a 285 hour continuously pumped tracer experiment, intensive monitoring was conducted during the 18-day pumping period . The average initial temperature difference was 3.3 °C. By 48 hours, the temperature difference had decreased to 3 °C showing an initial larger input of stream water. By 95 hours, the temperature difference was 50 percent of the original. By 137 hours the temperature difference stabilized with a difference of 0.5 degrees. Assuming that the temperature differences are proportional to the sources of flow and that no warming of the stream water occurs during its passage through the aquifer, a minimum estimate is that 85 percent of the water produced from the pumping well was derived from the stream, with 15 percent originating from the aquifer.

Lag Time

Another key factor is that of lag time. To assess how effectively the radial basis function (RBF) process improves water quality, one must determine the water travel time through the process. This step is critical in RBF studies because the correct determination of water travel time is the key to appropriate sampling procedures in which the quality of river water and riverbank filtrate can be directly compared.

This was determined by increasing the temperature of the stream water points and shifting the curve and subtracting the pump temperature points until the residuals reached a minimum (Figure 3). The minimum occurred with a shift of 30 to 34 hours. This indicates that after 34 hours the majority of water produced from the well was derived from the stream.

Figure 3. Water Travel Time

Figure 3. Water Travel Time

Microbial Results

Four pump tests were conducted, one at a “low” pumping rates (0.005 m3/second, 79 gal/minute) and three at a “high” pumping rate (0.0025 m3/second, 40 gal/minute) (Figure 4). These studies were conducted to determine whether riverbank filtration was functioning at variable pumping rates. These studies were conducted without any additional injection of microbial tracers. The indigenous concentrations of target organisms in the canal were used as the background concentration. The addition of infectious organisms into the canal at concentrations necessary to observe potential transport was not a possibility. Moreover, based on background sampling, we detected selected indicator organisms within the canal. We postulated that the detection of these organisms in the monitoring wells during the pumping test would provide data necessary to understand the migration of organisms during high and low pumping rates.

The background samples from the canal contained detectable concentrations of the target indicator organisms and selected pathogens (Giardia spp.). We postulated that we could monitor the presence of these organisms in the different sampling wells during the pumping to determine the movement of organisms under the varying pumping flow rates.

The results suggest that riverbank filtration appears to be an effective tool to prevent the migration of protozoan cysts. Riverbank filtration, however, does not totally remove or retard the migration of bacteria. Bacteria migrated at nearly the same rate at both low and high flow rates. We monitored the migration of two bacterial indicator organisms, E scherichia coli and Enterococci, that appeared in reduced numbers in the pumping well as compared to the canal, suggesting that bacterial pathogens could breakthrough the barriers during riverbank filtration.

Figure 4. An  Example of the Results of Pumping Test 1.

Figure 4. An Example of the Results of Pumping Test 1. Notice the variability of indicator species in both the channel and pumping well.

Future Activities: The project is almost complete. We now are concentrating on writing and publishing our results.

Journal Articles on this Report : 2 Displayed | Download in RIS Format

Other project views: All 10 publications 3 publications in selected types All 2 journal articles
Type Citation Project Document Sources
Journal Article Abdel-Fattah A, Langford R, Schulze-Makuch D. Applications of particle-tracking techniques to bank infiltration: a case study from El Paso, Texas, USA. Environmental Geology 2008;55(3):505-515. R829009 (Final)
  • Abstract: SpringerLink Abstract
  • Journal Article Widmer KW, Srikumar D, Pillai SD. Use of artificial neural networks to accurately identify Cryptosporidium oocyst and Giardia cyst images. Applied and Environmental Microbiology 2005;71(1):80-84. R829009 (Final)
  • Full-text from PubMed
  • Abstract from PubMed
  • Associated PubMed link
  • Full-text: AEM
  • Supplemental Keywords:

    drinking water, groundwater, sediment, stream, bank filtration, Rio Grande, arid streams, pathogens, biology, geology, southwest, border, geographic area, health, waste, water, arsenic, biochemistry, contaminated sediments, environmental chemistry, groundwater remediation, health risk assessment, hydrology, risk assessments, Escherichia coli, E. coli, Giardia, arsenic exposure, arsenic mobility, chemical contaminants, contaminant transport, contaminated sediment,, RFA, Health, Scientific Discipline, Geographic Area, Waste, Water, Contaminated Sediments, Environmental Chemistry, Health Risk Assessment, Arsenic, Risk Assessments, Biochemistry, Environmental Monitoring, Ecological Risk Assessment, Drinking Water, Groundwater remediation, EPA Region, monitoring, pathogens, fate and transport, risk assessment, contaminant transport, exposure and effects, natural disinfection, contaminated sediment, exposure, chemical contaminants, E. Coli, Region 6, cryptosporidium , treatment, municipal water, microbial risk management, human exposure, arsenic mobility, water quality, groundwater contamination, drinking water contaminants, drinking water treatment, Giardia, water treatment, arsenic exposure, riverbank filtration, groundwater

    Relevant Websites: exit EPA

    Progress and Final Reports:

    Original Abstract
  • 2002 Progress Report
  • 2003 Progress Report
  • 2004 Progress Report