A RIVERBANK FILTRATION DEMONSTRATION PROJECT ON THE KALI RIVER, DANDELI, KARNATAKA, INDIA

A small scale RBF system was installed in a village near the Kali River in the state of Karnataka to evaluate the performance of riverbank filtration (RBF) under the hydrogeological and climatological conditions of southern India. A series of hydraulic and tracer tests were carried out along with periodic biological and geochemical monitoring of various water sources in the study area. Hydrogen and oxygen isotopes highlight the impact of evaporation and irrigation at nearby rice paddies on the RBF production well. Dissolved silica data used to determine the relative contributions of surface and groundwater indicate that this RBF system derives approximately 28% of its water from the river. Even with nearly 3⁄4 of the RBF water coming from groundwater, bacteria and metals data indicate that groundwater dilution does not appear to play a major role in pollutant reduction. Instead, other RBF removal processes, such as biodegradation and redox chemistry, are at work in the system. Bacteria levels demonstrate at least 88% to >99% removal over currently used source waters. Despite this, Indian drinking water standards for E. coli are not consistently met and total coliform standards are never met in the RBF system. Bacteria levels are higher during the three month monsoon season. Average dissolved metal levels meet Indian standards for all metals analyzed. A community survey carried out before and after RBF installation shows significantly improved health indicators amongst RBF water users. In summary, this pilot-scale project demonstrates an RBF system that is welcomed by the host community and provides water of higher quality than other water sources in this study area.

This project investigates the suitability of a small River Bank Filtration (RBF) system for providing improved water to rural communities in developing countries.
The RBF study site is located in the tropical monsoon climate of rural southern India near the perennial Kali River, which receives polluted effluent from many sources, including municipal discharge and a large paper mill. At the beginning of the project, for drinking water, local residents relied on the polluted river water, unprotected and unimproved dug wells (Open Wells), Bore Wells, or water delivered from upstream of industrial and municipal inputs. These established water supply systems provide unsafe water and are unreliable, sometimes breaking down for months at a time (Patil, 2009).
This study characterizes basic chemical constituents in the local groundwater, Kali River water, and the RBF water. Measurements of metal and bacteria concentrations are used to determine RBF's capacity to alleviate contamination and to determine whether RBF treatment under these conditions can meet the Bureau of Indian Standards limits for drinking water. As well, water chemistry is used to determine the percentages that groundwater and surface water contribute to the RBF water to further understand to what degree the treatment mechanisms of RBF, such as groundwater mixing, biodegradation, or redox chemistry, are at work. Dissolved silica (Hooper and Shoemaker, 1986;Wels, Cornett and Lazerte, 1991) and environmental isotopes (Sklash and Farvolden, 1979) of hydrogen and oxygen are used for this purpose.
2 SIGNIFICANCE OF THE PROJECT: Safe Drinking Water: More than one billion people in the world -18% of the global population -do not have access to safe drinking water (UN, 2006). India is included in the group of countries with the maximum percentage of citizens who experience health problems due to unsafe water ( Figure 1) (Nature, 2000). As a result of unsafe drinking water, the World Health Organization reports that 4,000 children under the age of five die daily from diarrheal diseases worldwide. This is 90% of the total deaths due to diarrhea in the developing world (WHO, 2005;WHO, 2010). India is in the highest tier, with 4 to 7.9% of its total disease burden due to unsafe water (Nature, 2000) The CIA World Factbook ranks India 51 st out of 224 countries for infant mortality rate, with 51 infants dying by their first year of age per 1,000 live births. This is more than eight times the reported rate in the United States (CIA, 2009). Some common water-related diseases are those caused by infection from hepatitis A, typhoid, giardia lamblia, cryptosporidiosis, poliomyelitis, cholera, amebic dysentery, cyclosporiasis, and Escherichia coli (E. coli) (Centers for Disease Control and Prevention, 2003).
These diseases can cause gastro-intestinal infections, which in weak populations, such as the very young, can lead to death. Additionally, repeated diarrheal episodes can impair health by causing chronic malnutrition, increased infections, and reduced growth and development (Ejemot et al, 2009).
Many of India's perennial rivers are heavily polluted by the discharge of untreated sewage and effluents from industrial facilities. This water should not be used without treatment, even for irrigation. However, due to lack of other options and weak enforcement of existing water quality regulations, contaminated surface water serves many uses, including drinking.
Access to improved drinking water is estimated to reduce the occurrence of diarrhea by 25% (WHO, 2005). Others, when reviewing 38 studies on the topic, have found a 15 -43% reduction in diarrheal diseases due to hygiene, sanitation, water supply, and water quality interventions (Fewtrell et al, 2005). Therefore, the reduction of diarrheal diseases in the developing world requires a multi-pronged approach including availability of sanitary toilet facilities, access to safe drinking water, handwashing education, and safe storage of water (WHO, 2005). Low cost treatment such as RBF can be a part of this effort as it can potentially provide a clean and affordable source of drinking water to thousands of people in rural India.
Total coliforms are a group of bacteria that can survive and grow in both aerobic and anaerobic settings in warm-blooded hosts as well as in water and soil (WHO, 2006). Presence of total coliform indicates incomplete treatment or potential 4 contamination of drinking water (Feng, Weagant and Grant, 1998). E. coli are a subset of total coliforms that are adapted to the higher temperatures of human and animals' intestines (WHO, 2006). Therefore, E. coli are used as indicators of recent fecal contamination because they are unable to grow and reproduce outside of their host. In temperate environments their survival half-life outside of their hosts ranges from 1 day (in water) to 3 days (in soil). But in moist, warm, high-nutrient settings in tropical environments, E. coli can maintain free-living populations (Winfield and Groisman, 2003). The RBF field site is in a moist, warm tropical environment, but highly leached soils such as laterites have naturally low fertility (Baligar et al, 2004).
For this reason, E. coli is used as an indicator of recent fecal contamination at the RBF field site. Fecal contamination is a concern in drinking water supplies because it can carry pathogens causing diarrhea, meningitis, and other health problems (WHO, 2006).
Industrial Pollutants: Bacteria are not the only indicator for unsafe drinking water.
Water contaminated with heavy metals can cause stomach cramps (copper and zinc); anemia (chromium and zinc); diarrhea (copper); damage to the kidneys (cadmium and mercury), the nervous system (lead), brain functioning (manganese and mercury); and death (copper and lead) (Agency for Toxic Substances and Disease Registry, 2008).
One industry that may affect the research site is integrated pulp and paper mills, which produce many waste products, including heavy metals (US EPA, 2006).

5
Groundwater Depletion: Because surface water sources are often unreliable and unsafe for human consumption, as much as one third of the world population now relies on groundwater for drinking (Worldwatch Institute, 2000). Most groundwater, though, is used for irrigation, such as in northern India where 95% of groundwater used is for irrigating crops (Schiermeier, 2009). As a result of both agricultural and domestic uses, depletion of aquifers is an increasing threat to this water supply source.
The World Health Organization states that water usage has increased at twice the rate of population growth for the last 100 years (WHO, 2008).
The World Resources Institute's Pilot Analysis of Global Ecosystems (PAGE) predicts that the majority of the Indian subcontinent, as well as many other parts of the globe, will be experiencing water scarcity by 2025 (World Resources Institute, 2001) ( Figure 2). Here, water scarcity is defined as less than 2,500 m 3 of water/person/year.
By another measurement, per person annual water needs are 1 m 3 for drinking, 100 m 3 for other domestic use such as washing, and 1000 m 3 for food production, totaling approximately 1,100 m 3 of water/person /year (Allan, 2001). Even by this smaller measure, India's Krishna River basin-the second largest in India, covering nearly 260,000 km 2 , and neighbor to the Kali River basin-still falls short in WRI's estimated future supply (Bouwer et al, 2006). Excessive pumping from aquifers can lead to declining water tables, which can lead to problems such as well failure or changes in water chemistry. Excessive drawdown can also lead to irreversible compaction of the aquifer and land subsidence, which inhibits aquifer recharge. In addition, groundwater withdrawal can impact surface water levels which can become too low to provide habitat for aquatic life.
Beyond health and environmental effects, the greater cost of drilling deeper wells in the search for groundwater is an economic burden. Tushaar Shah, with the International Water Management Institute, states that over 25% of the farms in India are in danger of pumping their wells dry within the next few decades (Pearce, 2004).
In the state of Tamil Nadu, which borders Karnataka (the host state of the study site), 95% of small farmers' wells have already gone dry (Pearce, 2004). Additionally, 7 recent satellite imagery shows evidence of severe drawdown rates in northern India (Rodell, Velicogna and Famiglietti, 2009). High pumping and drilling costs can force small farmers, especially in India where many live at the subsistence level, to rely on rainfall to irrigate their crops. This can lead to diminished crop returns. The dramatic rise in Bore Well development in India that accompanied the 'Green Revolution' of the latter half of the 20 th century has lead to the current groundwater crisis. This has caused researchers to claim that "for the short term, drastic measures may have to be taken to ameliorate crisis conditions" (Narasimhan, 2006) and that there are "massive needs for investment in water supply systems for growing cities and for underserved rural populations" (Briscoe and Malik, 2006).
Proposed Solution: Riverbank Filtration (RBF) is one solution to the combined problems of contaminated surface water supplies and of aquifer depletion. RBF technology reduces withdrawal of groundwater, instead tapping into surface water which is currently underused in the research area due to contamination problems.
RBF draws infiltrating river water through the alluvium of a riverbed towards a well which is located up to a few hundred meters from the river ( Figure 3). Figure 3: Theoretical riverbank filtration (RBF) system diagram Map and cross-sectional views show path of infiltrating river water to RBF production wells (Kim, Corapcioglu and Kim, 2003) Similar to slow sand filtration systems, but with fewer ongoing labor needs, RBF uses the natural processes of sorption, ion exchange, redox reactions, precipitation, filtration, dilution, predation, and biodegradation to pre-treat drinking water (Hiscock and Grischeck, 2002;Kelly and Rydlund, 2006;Vogel et al, 2005a). RBF wells are best sited in sandy soils such as alluvial aquifers (Hubbs, Ball and Caldwell, 2006).
Much of the biological activity occurs within a few meters of the surface water interface. Here a biofilm of bacteria, fungi, algae, and protozoa embedded in a granular matrix lies just beneath the riverbed (Schmidt et al, 2003). Via this biologically active layer, referred to by its German name "schmutzdecke," RBF greatly reduces levels of pathogens, particles, and biodegradable compounds (Tufenkji, Ryan and Elimelech, 2002;Ray, 2004).
RBF systems have been used in European countries such as Germany, Holland, Hungary, France, Switzerland, and Finland for decades, and in some sites for over a century (Tufenkji, Ryan and Elimelech, 2002;Ray, Melin and Linsky, 2002).
Historically, RBF has been used mostly along rivers in temperate and cold climates such as Germany (Peel, Finlayson and McMahon, 2007). RBF is, however, relatively untested in monsoon climates, i.e. locations dominated by strong seasonal rains followed by a prolonged dry season. Because of the lack of studies on RBF's performance in these settings, municipalities in developing countries are reluctant to adopt this water treatment technology (Boving, 2007). As a response to extensive dysentery-related deaths (WHO, 2005) and increasing groundwater demand from population pressure -especially in the developing world-RBF is a well-suited lowcost, sustainable approach for producing safe drinking water in developing countries such as India.  (Radhakrishna and Vaidyanadhan, 1997).
Along river courses, alluvial soils are also found.  Kali River Water Quality: The Bureau of Indian Standards (BIS) has desirable (ideal) regulatory goals and, frequently, permissible (less ideal) regulatory goals (Appendix 3). Previous studies on the water quality of the Kali River and its tributaries (Manjunatha et al, 2001;Bharati and Krishnamurthy, 1990;Bharati and Krishnamurthy, 1992;Chavadi and Gokhale, 1986;Krishnamurthy and Bharati, 1994;Krishnamurthy and Bharati, 1996;Subramanian, Biksham and Ramesh, 1987) have found pH levels ranging from 6.8 to 10.9 (BIS desirable goal: 6.5 -8.5 (Kelly and Rydlund, 2006;Schmidt et al, 2003;Tufenkji, Ryan and Elimelech, 2002;Boving et al, 2010;Vogel et al, 2005b;Schubert, 2002;Hoppe-Jones, Oldham and Drewes, 2010;Sontheimer, 1980;Grischek et al, 2010;Trettin et al, 1999;Kuehn and Mueller, 2000) and have shown successful RBF water treatment using wells 5 -250 meters away from surface waters sources. Travel times for these systems are from under 1 day to 270 days and were determined through various means, including temperature (Kelly and Rydlund, 2006;Schmidt et al, 2003;Vogel et al, 2005b;Grischek et al, 15 2010), dissolved oxygen (Vogel et al, 2005b;Hoppe-Jones, Oldham and Drewes, 2010), chloride (Boving et al, 2010;Sontheimer, 1980;Trettin et al, 1999), TOC (Hoppe-Jones, Oldham and Drewes, 2010), and groundwater modeling (Grischek et al, 2010). By dividing the distance of each study's RBF well to its adjacent river by the travel time reported, travel velocities were calculated. These velocities typically ranged from around 0.5 to 18 meters/day but in one case was as high as 250 meters/day (Trettin et al, 1999 Cl -) as well as the pesticides atrazine (at or below 3 ppb) and simazine (at or below 4 ppb) (pesticides: PurTest; all others were the same strip tests from 2008). These two pesticides are among those most commonly found in US surface waters (Gilliom et al, 2006). Additionally, one five liter water sample from the Kali River was sent to an independent laboratory in Bangalore, India, for analysis of dioxin, pesticides, petroleum hydrocarbons and a number of other compounds suspected to be potentially present in the Kali River water (Appendix 5). This analysis was also used to corroborate our lab results.
Bacteria: The RBF wells were sampled periodically for bacteria levels from January to November 2009. In preparation for this, all four wells were sanitized in early January with a solution of 1 part 5% sodium hypochlorite (NaOCl) bleach to 3 parts water.
The solution was left in the well overnight and then pumped out continuously for at least one day (Minnesota Department of Health, 2006).
For total coliform and E. coli bacteria testing, raw unfiltered water samples were collected in 100 mL sterile bottles and kept in coolers in the shade until analysis in the lab. Because the submerged pump could not be easily moved, the wells without the pump were sampled with bailers dedicated to each well to reduce the possibility of cross contamination. In village Open Wells and the Kali River, a plastic bucket on a rope was used to collect samples (Figure 8). At each new water source, the bucket was submerged underwater, effectively pre-rinsing it before sample collection.  Dividing Q by the unit cross sectional area (A) generates the specific discharge (q) (Eqn 2). The specific discharge divided by the porosity (n) results in the pore velocity (v) (Eqn 3). The travel time is derived from the distance of the RBF well to the Kali River (L) divided by the pore velocity (Eqn 4).
where Q = discharge (L 3 / T) K = hydraulic conductivity (L / T) A = cross sectional area (L 2 ) i = hydraulic gradient: h = drawdown in well (L) L = distance to river (L) n = porosity (dimensionless) q = specific discharge (L/T) v = pore velocity (L/T) As well, the semi-logarithmic curve of drawdown versus time graphed in Aqtesolv was examined for evidence of the river as a recharge boundary to the RBF well (Fetter, 1994).
Bacterial Data: The IDEXX system's detection range is <1 to >2,419.6 MPN per 100 mL. All bacteria data are reported to two significant digits. Minimum values were converted to 0.9 MPN / 100 mL and maximum values to 2500 MPN / 100 mL (Costa, 2010;US FDA, 2007;Eaton et al, 2005). These altered end points were used when plotting and averaging data for each sampling site. Coliform data were averaged by water source and these averages were compared across categories. Geometric means rather than arithmetic means were used because coliform data commonly range over many orders of magnitude and geometric means minimize the effect of outliers in the data set (Costa, 2010;Herron, 2007 testing of skew < |2| and kurtosis < |4|, so non-parametric significance testing was used (Fernandez, 2010). Results from Mann-Whitney U tests using 2-tailed asymptotic significance test statistics are shown in Table 5. Water sources to the left of '<<' are significantly less contaminated than those the right, and water sources to the left of '<' are less contaminated than those on the right, but not significantly so.
Supplemental Data: Data on local population numbers and current water usage was used for general information and not analyzed per se. Data from previous tests of Kali River water (Appendix 3) were used for comparison and to assess possible changes with time. Existing well logs were used for understanding the geology and water table of the surrounding area.
GIS Data: All suitable data, including household survey, water quality, and hydrogeology data was spatially referenced with Global Positioning System (GPS) readings. Spatial data was supplemented with maps and pictures taken in the field.
Qualitative parameters for each sampling location were tabulated and linked to internet-accessible maps of the study area in Google. This will provide a central place to store and access periodic updates from future monitoring of the RBF water.

28
Lab Data: Isotope Data: Data on stable isotopes of oxygen ( 18 O/ 16 O) and hydrogen (D/H) were examined to identify surface water and groundwater signatures as evidence for the origins of the RBF water. It was predicted that the Kali River water would have a heavier isotopic signature due to evaporation at the Supa Reservoir 10 km upstream and that the local groundwater, derived from infiltrated rainfall that had not been subjected to evaporative stress, would be isotopically lighter than the river water. The proportion of these two source waters in the RBF well would give an isotopic signature of the production well between these two end members.  (Rorabacher, 1991;Alfassi, Boger and Ronen, 2005). Test results were used to determine if the December samples qualified as outliers to the January data sets, i.e. if the relative contribution of surface and groundwater had changed over the 11 months of well operation.
Ion Data: Anion and cation percentages were tabulated and compared with other sample sets from around the world of impacted and unimpacted waters as well as previous studies on the Kali River.
Metals Data: Dissolved metals concentrations were tabulated and outliers were identified using Dixon's Q significance testing. Data with outliers removed was then used for comparisons between the RBF well water and the Kali River. Percent change was looked at as well as comparisons with water quality criteria laid out by the BIS.

RESULTS:
In all graphs of data from this study, there each water source is represented in a consistent manner. In bar or line graphs, the colors remain consistent. In scatterplot graphs, the colors and shapes remain consistent. Samples from the Kali River are shown as yellow triangles. RBF Wells 1, 2, 3, and 4 are shown as lavender, turquoise, black, and dark green squares. The Open Well in Kariyampalli is shown as maroon circles. The Mainal Bore Well and Local Bore Wells are shown as royal blue squares or diamonds. Indian Groundwater is shown in light blue. STABLE ISOTOPES: In total, 18 samples were collected from the study area, including the RBF wellfield (n = 11) the Kali River (n = 3), the Kariyampalli Open Well (n = 3), and the Mainal Bore Well (n = 1). The range of these data points was -1.99 to -0.1‰ δ 18 O and -10.98 to 3.49‰ δD. Samples from the study area correlate closely with the meteoric water line from Belgaum (Kumar et al, 2010) (approximately 70 km NNW of the study area), with data points from the Kariyampalli Open Well deviating the most from the regional average precipitation (Appendix 8).   recovery test and two pump tests -Appendix 10) were analyzed using the Cooper-Jacob method in the Aqtesolv program and yielded hydraulic conductivities of 9.0, 13.6, and 15.1 m/day) via visual matching ( Figure 9 and Appendix 10). Total coliform: The data set (n = 95), along with rainfall distribution over the study period, is shown in Figure 10. Of all samples, 25 (26%) were at or beyond the upper detection limit of the method. Most of these highly contaminated samples were taken from the Kali River, where the actual total coliform concentration was at or greater than the detection limit of 2500 MPN / 100 mL in 11 out of 15 samples (73%). Four data points were removed from the data set due to various reasons. These reasons include high readings attributed to the pump probably coming into contact with bacteria while being moved from Well 3 to Well 4, possible data transcription errors, and suspicion of outside contamination of samples. Removing these four data points did not dramatically affect the results (Appendix 15). Eight Bore Well samples analyzed by the Dharwad District Health Lab were included for comparison. In figures and tables here, these are included when the phrase 'Local Bore Wells' is used.
Five of these showed total coliform concentrations at the minimum detection limit, which did not correlate with the measurements we took. Additionally, data from India's Central Pollution Control Board show total coliform concentrations ranging up to 15,000 MPN / 100 mL in all water sources and up to 6,000 MPN / 100 mL in water sources whose labels specify bore wells or groundwater (  This near equivalence of total coliform and E. coli concentrations is not seen anywhere else in data from this or other studies and the E. coli data was therefore deemed suspect. In order to match the methodology used in this study, three data points that were above 2,500 MPN / 100 mL were changed to 2,500 MPN / 100 mL.  Aggregate Annual Data: Considering the entire total coliform data set from this RBF study, the average percent change from the Kali River's geometric mean of 1700 MPN / 100 mL ranged from -44% (at KOW, MOW, and MBW) to 95% (RBF Well 3). The data are graphically summarized in Figure 12. Note that negative percentages indicate a bacteria concentration that is higher than the Kali River water, whereas positive numbers are indicative of bacterial loads lower than the Kali River. Because the 38 KOW and MOW had total coliform levels that were worse than the river, their maximum percent changes are zero. Again, because the actual total coliform concentration in the river exceeded the upper detection limit in 11 of 15 samples, the actual removal percentage calculated based on the aggregate annual data is underestimating the performance of the RBF system.
Results from non-parametric Mann-Whitney U tests on the total coliform and E. coli levels in RBF Wells 3 and 4, the Kariyampalli Open Well, and the Kali River at the RBF wellfield sampling site are shown in Table 5. In all instances, water from RBF Well 3 was the least contaminated. For the wet season, there was only enough data to perform Mann-Whitney U tests on the bacteria concentrations of the Kali River and Well 3 (see Appendices 16, 17, and 18 for greater detail).
Seasonal Data: The total coliform data set was divided into two seasonal data sets for the dry (n = 75) and wet (= Monsoon; n = 20) seasons (Table 6). Overall, the Kali River showed almost twice as much total coliform concentration during the dry season relative to the monsoon (2100 versus 1200 MPN/100 mL). The opposite trend is seen at the production well. The total coliform concentration at Well 3 was less than half as much during the dry season relative to the monsoon (66 versus 140 39 MPN/100 mL). The removal efficiency of RBF Well 3 was 97% during the dry season and 88% during the monsoon. Independent of the season, the Kariyampalli Open Well (the principal water supply for the villagers prior to the RBF installation) was always equally polluted or more polluted than the Kali River.
n/a n/a n/a n/a 0% Figure 15: Total coliform -monsoon data Percent change relative to the Kali River is shown above each column. Error bars show upper and lower range of geometric mean standard deviation. n/a = not analyzed, KR = Kali River, W1 = RBF Well 1; KOW = Kariyampalli Open Well; LBWs = Local Bore Wells Table 6: Total coliform concentrations and removals Shown are all data combined and organized by dry and wet seasons. Note that whenever ≥2500 is used, the upper detection limit has been exceeded. As such, the average and maximum RBF removals versus the river are underestimated. Indian groundwater ("Indian GW") data is a combination of CPCB 2006 data (n = 127), Dharwad District Health Lab data (n = 8) and data from this study (n = 1

Avg % Change
Max % Change  : E coli-Monsoon data. Percent change relative to the Kali River is shown above each sample 54% 97% n/a n/a n/a n/a 0% Figure   In contrast to the total coliform data, the E.coli concentration did not change as much by season. The Kali River maintained approximately the same total coliform concentration during the dry season relative to the monsoon (470 versus 440 MPN / 100 mL). The total coliform concentration at Well 3 did not change substantially between the dry season relative to the monsoon (1.8 versus 13 MPN / 100 mL).
IONS: Cation (n = 80) and anion (n = 63) analyses of samples from the wellfield and the surrounding region are summarized in Table 4 with supplemental information provided in Appendices 12 and 20. Summaries of previous studies on the Kali River (Bharati and Krishnamurthy, 1990;Krishnamurthy and Bharati, 1994    147 mg/L) in two previous studies of the Kali River at Dandeli (Bharati and Krishnamurthy, 1990;Krishnamurthy and Bharati, 1994), whereas this study found 0 to 60 mg/L using test strips (avg: 20 mg/L). Hence, previous studies showed more than 7 times greater alkalinities than reported here. Further, comparison of electrical conductivity readings from this study with the sum of the anions and cations (Lenntech BV, 2011) shows wide variability in comparisons, with 39% with lower ion totals and 48% with higher ion totals than expected (Appendix 12). This suggests also that the ± 20 -30 mg/L error for the bicarbonate test strips is evident in data from this study. Altogether, the potentially underreported anion (alkalinity) concentrations limited the usefulness of this data set. were also performed on chloride levels. Other anions were not uniformly reported in previous studies. None of this study's anion data qualified as outliers (Appendix 23).    -2006(Panchayat Raj Engineering, 2006 This study: On Piper plots, samples from this study (n = 47) showed a broad range of compositions while staying at or below 30% magnesium. Most samples (76%) are calcium to neutral cation type waters, while 24% were Na/K type waters, including 2008 Kali River samples (n = 9) and Moulangi and Kerwad Town Tap.
Kali River samples were neutral to Na / K type waters in the cation field ( Figure 27).  Table 11 shows the Bureau of Indian Standards (BIS) and the World Health Organization (WHO) limits for dissolved metals in drinking water and whether sample categories from this study exceeded those standards. The majority of cases that exceeded WHO drinking water standards were above aesthetic rather than health standards (Table 11). For sample averages (without outliers) from the RBF wellfield area, Figure 29 shows the water quality relationships between the Kali River in Kariyampalli, the Results from this study can be broadly categorized as contributing understanding to two questions concerning this RBF system. The first question is where the water in the production well originates. Isotope, silica, and aquifer test data was used to address this research question. These results are discussed in the section 'Source of Water in the RBF Well.' The second major question asked during this study is the degree to which the RBF production well delivers high quality drinking water. This is approached by analyzing field parameters and bacterial, metals, and ions data and elaborated on in the section 'Performance of the RBF System.' Source of Water in the RBF Well: Stable Isotopes: Isotopic variation amongst the samples was not great enough to use these values for distinguishing surface water from groundwater inputs to the RBF water. The slope of the local meteoric water line (BMWL) recorded in Belgaum by Kumar et al., (2010) is 7.78. Two out of three Kali River samples plot close to the BMWL meteoric line, indicating that the river is fed by precipitation water. Closely paralleling this is the best fit line for data points from this study that are not impacted by evaporation (KR, MBW, RBF W 1, 2, 3; n = 13). That slope is 5.43 ( Figure 9).
In contrast, data points (n = 7) from the Kariyampalli Open Well and the nearest RBF well (Well 4) form a line with shallower slope (2.93). These data show the effect of isotopic fractionation due to evaporation at the rice paddies surrounding the Kariyampalli Open Well (Kendall and McDonnell, 1998). All of these lines are shown in Figure 9.

60
The isotopic signature of groundwater not influenced by precipitation, evaporation, or river recharge could not be determined. Although the isotopic data could not be used for the originally intended purpose -distinguishing between surface and ground water inputs -it did prove useful when combined with the dissolved silica data to show the source of water in the RBF production well.
Silica: A two-end member mixing model was created using silica concentrations as a proxy for groundwater percentage in the samples. In this model, the Kali River water was set at 0% groundwater and the Mainal Bore Well at 100% groundwater, under the assumption that it is entirely fed by groundwater. Based on that, the dissolved silica levels, which ranged from 21 to 39 mg/L, corresponded to 38% groundwater (KOW) to 72% groundwater (RBF W3) in January. After 11 months of pumping in which the pump was running approximately four hours per day, another set of samples was taken. These showed 27% (RBF W4) to 73% (RBF W3) groundwater contributions.
With particular reference to RBF Well 3, the mixing model shows an average of 28% river water contribution (Table 12). This falls within the range seen in other studies of RBF systems, which show 13 -75% surface water (Appendix 1). Silica concentration demonstrates that the percentage of Kali River water drawn into RBF Wells 1, 2 and 4 increased with time pumping, whereas Well 3 (the pumping well) shows a constant percentage of the ratio of river water to groundwater. Results from Dixon's Q testing show the December increase in RBF Well 2 to be significant at the 95% confidence level. None of the other wells show a statistically significant change between the January and the December silica data at the 95% confidence level (Appendix 24). The change at Well 2 is hypothesized to be the result of more river water being drawn into 61 the RBF wellfield during the 11 months of pumping. This mixing model would benefit from a greater number of samples to give greater confidence to its results. When isotope data are combined with silica concentration data, effects from evaporation at the rice paddies is corroborated. In theory, samples would have increasing silica concentrations with distance from the river because groundwater has more time for water-rock interaction. Instead, as Figure 38 shows, only RBF Wells 1 and 2 fall on a mixing line between river water and groundwater and the evaporative effect of the rice paddies is shown on the Kariyampalli Open Well and RBF Well 4.
Additionally, the rice paddies around the Kariyampalli Open Well were sometimes irrigated with Kali River water and the effect of this is seen in the silica data in which RBF Wells 3 and 4 as well as the Kariyampalli Open Well begin to reverse the trend seen along mixing line A such that mixing line B trends back towards the silica signature of the Kali River ( Figure 30). Thus evaporative enrichment and irrigation with Kali River water of the rice paddies are influencing shallow groundwater in that area and acting as a secondary surface water source to the RBF system. This leads to a three-end member model of effects on the samples in this study rather than a two-end member model that was originally conceived between the surface waters of the Kali River and the 88 meter deep groundwater of the Mainal Bore Well. pump tests on Well 3 suggest the existence of a recharge boundary (Fetter, 1994). The existence of a recharge boundary further supports the idea that the RBF production well is receiving water from the Kali River and, possibly, the nearby rice paddies ( Figure 10 and Appendix 10).
Performance of the RBF System: Field Parameters: pH levels at the production well ranged from 6.1 to 7.3 with the average being 6.6. Nearly a quarter (12 of   cell membranes, ruining cell walls and killing the microbes (Suslow, 2004). In this study, the highest ORP value seen was at the Kerwad Town Tap Figures 13 and 18). However, BIS standards for total coliform are regularly exceeded while E.coli standards are met most the time.

67
Total coliform: The dilution levels calculated with the silica data were used to predict the total coliform bacteria levels at the research site. This was done to distinguish reductions in bacteria due to dilution versus removal due to other RBF processes.
Compared to the actually measured concentration (Figure 33), the predicted total coliform values were lower at Well 1, the Kariyampalli Open Well, and the Mainal Bore Well. RBF Well 1 probably shows a higher level of total coliform than expected due to a livestock holding area within 3 meters of the wellhead. The Kariyampalli Open Well most probably had higher coliform levels because users dip into that well with buckets that have been placed on the ground or have otherwise come into contact with contaminated surfaces. The well is unprotected and anything may fall into it.
Additionally, the rice paddies surrounding the Kariyampalli Open Well are plowed with cattle and water buffalos whose manure could contribute to bacteria levels in the Open Well. . These data imply that the 63% of samples in the Dharwad District Health Lab data that found total coliform levels below the minimum detection limit and 75% at or below 10 MPN / 100 mL is unusual in comparison to other data sets in India. Therefore it is perhaps less surprising that this study found such high total coliform levels at Mainal Bore Well than that the local health lab found such low levels in this and nearby bore wells. In 2008, the Mainal residents considered the Mainal Bore Well to be a good water source and the only reason they didn't use it at that point was that the pump was broken. In 2009 when the pump was working and samples were collected, it was presumably in use by local residents. Because of the discrepancies in bacteria concentration in the local groundwater, the models used in this study set these levels at 0 MPN / 100 mL.
On the other hand, dilution calculations predict higher levels of total coliform than were actually seen in RBF Wells 2, 3, and 4 ( Figure 33). At the production well (RBF W3), actual total coliform removal was more than five times greater than that attributed to dilution alone (dilution prediction of 490 MPN / 100 mL versus actual mean of 85 MPN / 100 mL). This indicates that factors beyond dilution, such as biological activity and filtering, are at work in removing bacteria in the RBF process.  can be attributed to these other factors. Mean total coliform detected at the Kali River (column 1) is multiplied by the percentage of river water (column 2) to produce the geometric mean predicted by dilution with groundwater at 0 MPN / 100 mL (column 3). In contrast, column 4 shows the actual coliform levels at each sample site.
Column 5 quantifies the other RBF processes, such as filtering and predation, at work in the system. Column 3 minus column 4 then divided by column 1 (the source water) yields column 5. Table 13 underestimates the non-dilution processes because it sets the groundwater at 0 MPN / 100 mL of total coliform, which is lower than the data indicate. As there was some discrepancy in what the actual bacteria level is of the local groundwater, this absolute minimum level was used here. Figure 33 gives a graphical display of columns 3 and 4 from Table 13.  E. coli: Dilution levels calculated with dissolved silica data predict from 1.4 times (at KOW) to 36 times (at W3) more E. coli than is actually seen in the data, indicating that dilution is not the only factor at work in E. coli removal ( Figure 34). Again, evidence of livestock holding near RBF W1 is seen in the data, as this well has a notably higher level of bacteria than the more distant RBF wells and even the river itself.
The geometric average of E. coli bacteria level in the production well (RBF W3) is at 3.6 MPN / 100 mL over 10 months of data collection. Although this is slightly above the BIS level of <1 MPN / 100 mL, it is greatly improved from the levels of the other local water sources in the area. Therefore, minimal disinfection will be required to get the RBF water to BIS standards, but it will be much easier to achieve treatment with water starting at 3.6 MPN / 100 mL than with Kali River levels 71 (Schmidt et al, 2003;Kuehn and Mueller, 2000). As an example, studies have shown that nanofiltration treatment membranes had to be replaced every 8 days when used with conventionally pre-treated surface water and replaced every 62-75 days when used with water pre-treated via RBF . Further, water with approximately 1-10 MPN / 100 mL is occasionally allowable according to BIS total coliform regulations, but may increase in contamination level due to regrowth during storage in the home. RBF-treated water, though, tends to have low assimilable organic carbon (AOC) content and therefore inhibits bacterial regrowth when compared with raw river water and water treated via ozonation or activated carbon (Schmidt et al, 2003). Finally, it is suspected that at least some of the bacteria originate from livestock manure inside the RBF well catchment area. This assumption is based on the observed higher bacteria concentrations at RBF W3 during the wet season (Tables 6 and 7), which indicate possible influx of bacteria contaminated seepage from the surface. Based on the results of this study, the owner of the cattle was advised to move the livestock away from the RBF wellfield.  the Kali River (column 1) is multiplied by the percentage of river water (column 2) to produce the geometric mean predicted by dilution with groundwater at 0 MPN / 100 mL (column 3). In contrast, column 4 shows the actual bacteria levels at each sample site. Column 5 quantifies the other RBF processes, such as filtering and predation, at work in the system. Column 3 minus column 4 then divided by column 1 (the source water) yields column 5. Table 14 underestimates the non-dilution processes because it sets the groundwater at 0 MPN / 100 mL of E. coli bacteria, which is lower than the data indicate. Because there was only one data point for the local groundwater, this absolute minimum level was used here as a precaution against possibly erroneous data. Figure 34 gives a graphical display of columns 3 and 4 from Table 14.  Kali River samples from previous studies show similar trends to samples from this study. Samples from upstream of the Halmaddi paper mill effluent are lower in ions and similar to rivers in rural areas or to a global average river (Duh et al, 2008;Hem, 1985). This study: Dixon's Q test results indicate that analytical errors associated with determining alkalinity in the field explains the apparent cation : anion imbalance shown in Appendices 12 and 22. Overall, the RBF production well delivers water with an ion content that falls between the river and local groundwater (Figure 24).
The RBF water, in comparison to the Kali River water, is more closely related to 'unimpacted waters,' such as groundwater, rural river water, and precipitation, seen in studies from other parts of the world (Duh et al, 2008;Hem, 1985;Gobel, Dierkes and Coldewey, 2007;Barber et al, 2006;Safai et al, 2004;Lee and Fetter, 1994).  Table 15 compares the dissolved metals levels at the production well with those found at the Kali River at the field site and with the levels at the most contaminated sources of drinking water in the area that were in use by villagers at the time of the sampling (Appendices 25, 26, and 27). Overall, the average RBF water quality is superior to the alternatives and all applicable drinking water standards for metals were met. Based on the silica mixing model data, the improvement of the RBF water quality must be caused by processes other than just dilution.   One exception to this is the chromium Bore Well value, which lacks data from Mainal.

METALS:
Dissolved metals concentration at the Kali River sampled at the RBF wellfield site (column 1) is multiplied by the percentage of river water (column 3) and dissolved metals concentration at the nearby Bore Wells (column 2) is multiplied by the percentage of groundwater (100 minus column 3). The concentration of dissolved metals expected at the production well from dilution of the surface water and groundwater (column 4) results from the addition of these two products. In contrast, column 5 shows the actual metals concentrations at RBF W3. Column 6 quantifies the other RBF processes, such as sorption and precipitation, at work in the system.
Column 4 minus column 5 then divided by column 4 yields Column 6 (the percent change at W3 over the concentrations expected via dilution alone) (see Appendix 27 for complete calculations). From Table 17 we see that, as with the bacteria data, dilution itself (here, surface water is diluting groundwater) accounts for a portion, but not the entirety, of pollutant removal at the RBF wellfield. Well depths around 90 meters, water from these sources is likely more characteristic of the underlying metasedimentary rocks, rather than the laterized soils overlying them which are enriched in oxidized, and therefore soluble, iron and manganese.
Since this study limited sampling to less than a year, it is recommended to take additional samples for iron to determine if lower concentrations remain stable over time. As mentioned above, decreasing trends in iron content seen in previous studies since the early 1990's, as well as processing upgrades at the West Coast Paper Mill, imply that improved dissolved oxygen levels in the Kali River may lead to decreased iron levels at RBF Well 3 in the future, as more oxygen-rich river water leads to lower levels of dissolved iron and manganese in source waters. Similar results were seen in 79 RBF systems along Germany's Rhine River in response to improved water quality leading to higher dissolved oxygen levels in the river after environmental protection measures in the 1970's (Schmidt et al, 2003;Kuehn and Mueller, 2000). In summary, total coliform BIS guidelines were never met during the study period because although levels did dip below 10 MPN / 100 mL, they never did so on consecutive days. E. coli bacteria BIS guidelines were met 10 days out of the 46 sampling days (22% of the time). These occurred in both the dry (9 days or 29% of the time) and the monsoon (1 day or 7% of the time) seasons. BIS standards were met 100% of the time for chromium, copper, cadmium, zinc, and lead readings.

Entire Data Set with Std Error
Manganese and iron concentrations, as mentioned above, met the BIS guidelines once the flow field stabilized at the research site.

FUTURE STUDIES:
There are many processes at the field site that warrant further research. Some involve strengthening the current data set by continued sampling and analysis. Others involve looking into topics that have not yet been covered in this study. Kariyampalli village, south of Dandeli, Karnataka was in need of a reliable, safe drinking water supply and an ideal candidate for a riverbank filtration system.
When looking at the overall picture of both bacterial and metals concentrations, RBF water is safer than any of the other drinking water sources in the area. Additionally, using RBF water instead of Bore Wells reduces groundwater depletion.
Ion analyses support the observation that the Kali River upstream of the research site is impacted by municipal and industrial inputs.
Bacterial pollution appears to be a larger problem in this area than metal pollution of the drinking water as all water samples exceeded BIS standards more frequently for bacteria levels than for metals levels. Samples indicate average removal at RBF Well 3 from the Kali River of 95% (maximum >99%) for total coliform and 99% (maximum >99%) for E. coli bacteria. The actual removal percentages are even higher, but could not be precisely determined in the field. Silica data indicate that water from RBF Well 3 is 28% river water. Yet dilution with groundwater by itself does not explain the reduction in metals and pathogen concentrations. Of the total coliform change, 23% is attributable to processes other than dilution by groundwater.
As well, 27% of the change in E. coli concentration is due to non-dilution processes in the RBF wellfield. Stable isotope data indicate that the removal of both total coliform and E. coli levels by riverbank filtration could become even greater if the Kariyampalli Open Well is retired, as this might reduce the amount of bacteria introduced into the RBF wellfield. Additionally, if this project is replicated elsewhere in India in locations where the wellfield can be protected from irrigation water and 83 livestock manure, better pollutant removal can be expected, especially if the wells are set back further from the river, leading to travel times greater than the range of 8.7 -23 days seen at this site. Although bacteria levels seen in the production well are generally slightly above BIS regulations, the levels are a great improvement over currently used water supplies. Data gathered during the monsoon season indicates increased bacteria levels during this season, but further protection of the wellfield from surface contamination by livestock manure may improve this problem. As Bore Wells are associated with groundwater depletion, RBF provides water that best addresses the combined issues of both groundwater depletion and surface water contamination.
Average RBF production well concentrations of metals, like most other sampled sources, are all below BIS standards, although the data for iron and manganese did temporarily range above BIS aesthetic guidelines.
This study, including a community survey, demonstrated that RBF-treated water is welcomed by local residents, and, through dilution and other processes, reduces pollutants found in source waters by up to >99%, producing water that is near or below BIS standards and leading to significantly improved self-and proxy-reported family health by RBF users.    These represent 24.3% of the 37 total households (both RBF users and non-users) in the village.
Water: Survey respondents use the RBF water for household uses such as drinking and cooking. None reported using it for agricultural purposes. One third of RBF users reported water shortages that are "not too bad" for household use and the rest stated that they have no water shortage concerns. Similarly, one third of respondents claim that the RBF water supply is irregular, and the rest that the supply comes regularly every day. 89% of RBF users also use other water sources. There is incomplete data on which water sources are referred to in this question. The RBF water is reported to be 'clear' by 100% of users and to taste either sweet (11%), normal (56%), or bitter (33%). None of the RBF users reported purifying the RBF water before using it, with 56% being "satisfied" and 44% "very satisfied" with the water quality. Two thirds of users say that the water supply is within a 10 minute walk from their house. As a consequence, 78% state that they spend less than half an hour per day collecting water for the household. 33% pay nothing for the water and 67% pay between INR 30 -50 per month for the RBF water. 58% of those surveyed use Kali River water for agricultural purposes, but none use the river water within their household.
Health: Two thirds of surveyed residents describe their family's health as "good" and one third as "average," with none reporting "bad" health. Only 11% feel that the RBF water affects their health. Two thirds state they do not experience gastrointestinal distress. 11% state that they sometimes have episodes of diarrhea and 22% said that 92 they don't know. 56% of these users attended the water hygiene and sanitation workshops held the previous year. Of those that attended, they claimed these workshops were either "good" or "very good." Results comparing Kariyampalli residents' responses in 2008 (n = 16 -19) and 2010 (n = 9) were compared for the following parameters, which met normality standards:  Water: Cost considerations may explain why none of the survey respondents reported using the RBF water for agricultural purposes. Another reason may be that the water tank is too far away from their fields. The report of bitter-tasting water may not necessarily be a negative attribute as bitter flavor is considered to be one of the six basic flavors in a well-rounded Indian meal. In traditional Indian Ayurvedic medicine, bitter flavor is used to purify the body and aid in digestion (Gupta, ). The minimal concern for water shortage by RBF users implies that there is enough water in the house to use for hand washing, a crucial element in family health (WHO, 2005;Ejemot et al, 2009;Fewtrell et al, 2005).
Hand washing education was a component of the overall project in which this study was conducted. (Fewtrell et al, 2005) found, in reviewing 38 articles on various techniques for reducing diarrhea in developing countries, that a 15% reduction could be attributed to water quality improvements and a 33-42% reduction from hand washing education, although they do state that possible publication bias may lead to inflated numbers (Fewtrell et al, 2005). Another meta-analysis of five trials found a 30-32% reduction in diarrhea attributable to hand washing education (Ejemot et al, 2009). The World Health Organization has stated that a 25% reduction in diarrhea can be expected from water quality interventions and a 45% reduction from hygiene improvements (WHO, 2005). Therefore we can expect that approximately 15-25% of the diarrhea reduction seen in this study was attributable to the RBF system and 30-45% to the hand washing education workshops. Combined, these two factors then account for 45-70% of the reduction in diarrhea.  20011985-1986-1988-1988 3 -34 3 -34 Co (ppb) 0.  1987 -1988 1978, 1984 1983-1984 1983-1984 1983-1984 1983-1984 7.03 -8.51 7.2 7.6 -9.2 7.5 -9.1 6.9 -9.4 8.  1987 -1988 1978, 1984 1983-1984 1983-1984 1983-1984 1983-1984                  1/12/09 iButton testing of incubators Began sampling at RBF site Power went out at midday so the incubator was not heating up the whole time we were in the field this afternoon. It was warming up for an hour or two before we put the samples in. It looks like 37.9 °C on the top red readout is actually about 35 °C (according to thermometer inside) 1/13/09 gathered bacteria results Thermometer read 33.2 °C red display on incubator read 35.2 °C green display in incubator read 11.9 V 1/14/09 Meeting with Dr. Preeti Kudesia from World Bank -New Delhi. She suggests best to ask about acute diarrhea only in past 2 weeks -longer than that is too hard to remember. Also, she suggests we give plotting materials to the primary health center nurse i.e. red dots to put next to a child's name in a graph means they had diarrhea that week, green dot means no diarrhea. That will help us keep more accurate records.
Note: the power company released extra water today so that people can bathe in the Kali River in celebration of Sankranti, so the river is high today.
Collected samples at RBF site 1/15/09 Slug tests Read bacteria results in lab Incubator: red display: 35.8 °C, green display: 12.35 V, thermometer: 34 °C Note: these sterile sample bottles are hard to judge exactly where the fill line is for 100 mL so our first few bacteria test runs have ended up being a bit short of 100 mL so the Quanti-trays have not been completely full. When reading the bacteria results, therefore, I have noted under each sample location how many of the large Quanti-tray wells ended up being empty. Because of these empty Quanti-tray wells, this has introduced some error into our data, so results are reported as a range of MPN #'s, i.e. at least the # of Quanti-tray wells counted up to at most the added empty wells if they had been positive.
We moved the pump from well #4 to Well #3 this morning. This afternoon we ran the pump for about an hour, then checked water level = 7.135 m. Slight leakage (dripping) out of pipe beyond the meters.
Both meters running together: Right meter: 0.658 m 3 /10 minutes (This one sat in a cardboard box in Dandeli for 3 months without refrigeration, though it is pretty cool inside the classroom where the box was stored) Set up schedule for future sampling after I am gone 1/20/09 Collected samples at RBF site LevelTroll in the Kali River is out of the water (water level has gone down since it was hung out there) 3:30 pm: Levelogger LTC: I switched the conductivity measurement from 80.00 mS/cm to 30.000 mS/cm for more accuracy Kali River data logger: 2.4 meters from tip of branch down to the mud. We lengthened the rope holding the data logger so it is in the water now. 4:45 pm I started a new program with the 1 st log set for 21 cm water gauge reading. I think this may be the first time this meter was correctly set for Kali River actual conditions. I think the others have all been programmed wrong for various reasons.
Mr. Patil said 2001 no rain, Kali was very low. 2002 pipeline to Kariyampalli installed. Worked on and off until 2008 when it broke completely. It served Kariyampalli, Mainal, Kerwad, Harnouda, Saksali, and Halmaddi. 6 Lakh from West Coast Paper Mill (10 % of cost) other 90% from World Bank. 2001: drought: >100 cattle owned by Mr. Patil died b/c no clean water. "Kali River was just black." Lots of other people lost cattle then also.
April + May = summer Jun -September = rainy season October -February = dry West Coast Paper Mill paid 10% to supply water via pipelines to villages and World Bank supplied the other 90%, but work was not done properly so pipeline was constantly breaking down and getting repaired until 2008 when it broke down a year ago and still hasn't been fixed so people in Kariyampalli and Mainal must use the Open Well.

1/21/09 Collected samples at RBF site
Mainal Bore Well: this is one of 3 tanks from one bore ~200 m away from this one. This one is the one they drink from. Mr. Patil