was a better adsorbent for NOM than calcium carbonate (CaCO 3 ) (which precipitates at

Enhanced softening is the removal of natural organic matter (NOM) during precipitative softening, with an objective of reducing the formation of disinfection by-products (DBPs). A BY SHAY RALLS ROALSON, JIHYANG KWEON, DESMOND F. LAWLER, AND GERALD E. SPEITEL JR. moderately hard water source (Lake Austin, Austin, Texas) was softened using a wide range of lime dosages and various other chemical additions (magnesium, iron, phosphate, silica, and sulfate). The results for hardness ions (calcium and magnesium), NOM (indicated by dissolved organic carbon and ultraviolet absorbance at 254 nm), and 21 DBPs formed after chlorination at ph 9 are reported. Magnesium hydroxide, which precipitated above ph 11, was a better adsorbent for NOM than calcium carbonate (CaCO 3 ) (which precipitates at lower lime doses), but removal of NOM by CaCO 3 was by no means inconsequential. DBP formation generally followed the NOM trends but was complicated by the formation of bromine-substituted DBPs because of the moderately high bromide concentration in this water. Enhanced Softening: Effects of Lime Dose and Chemical Additions Natural organic matter (NOM) is important in drinking water treatment because of its potential to react with chlorine and other disinfectants to form disinfection by-products (DBPs) that may have adverse health effects. The primary strategy of the drinking water industry for controlling DBP concentrations has been to remove NOM with conventional treatment processes. Much research has focused on enhanced coagulation the removal of NOM (precursors of DBPs) during coagulation with iron or aluminum salts. In contrast, little attention has been placed on enhanced softening the removal of DBP precursors during the softening process. Thus, this research was undertaken to better understand the role of calcium precipitation, magnesium precipitation, metal coagulants, secondary anions (phosphate, sulfate and silicate), and sludge recycle on NOM and DBP precursor removal in enhanced softening. The research was performed on water from Lake Austin in Austin, Texas, which is an example of a typical surface water source for softening plants. BACKGROUND Softening chemistry. Hardness is the concentration of polyvalent cations in water and is generally dominated by calcium (Ca) and magnesium (Mg) cations (Ca +2 and Mg +2 ). Softening consists primarily of the precipitation of Ca as calcium ROALSON ET AL PEER-REVIEWED 95:11 JOURNAL AWWA NOVEMBER 23 97

carbonate (CaCO 3 ) and Mg as magnesium hydroxide [Mg(OH) 2 ], according to the following equations: CaCO 3 (s) = Ca +2 + CO 3 2 K SO,Ca = 1 8.48 (1) Mg(OH) 2 (s) = Mg +2 + 2 OH K SO,Mg = 1 11.16 (2) in which K SO is the solubility product for each solid. These two reactions make clear that the carbonate speciation and the hydrolysis of water are involved: H 2 CO 3 = HCO 3 + H + K 1 = 1 6.35 (3) HCO 3 = CO 3 2 + H + K 2 = 1 1.33 (4) H 2 O = H + +OH K w = 1 14 (5) in which K 1, K 2, and K w are the equilibrium constants for the respective reactions. It is common to express the concentrations of each carbonate species as a fraction of the total. For example: [CO 3 2 ] = 2 C T (6) in which C T is the total carbonate concentration in mol/l, defined as C T = [H 2 CO 3 ] + [HCO 3 ] + [CO 3 2 ] (7) The preceding reactions and expressions include all of the species that directly participate in softening reactions (ignoring complexation), but softening takes place in waters that contain many other species as well. These species do not have to be accounted for individually and explicitly, but they must be included in lumped parameters such as alkalinity. Alkalinity can be expressed in a few different ways that are relevant to the softening process. Alkalinity is the difference between the proton condition of a water in its initial state and the proton condition of a water when it has been titrated to an equivalent carbonic acid (H 2 CO 3 ) solution. Assuming the water has no other weak acid/base species besides carbonate, or: Alk (eq/l) = 2[CO 3 2 ] + [HCO 3 ] + [OH ] [H + ] Alk(eq/L) = (2 2 + 1 ) C T + Kw [H+ ] [ H+] (8a) (8b) Using the right side of Eq 8b as a basis, a charge balance can be rearranged to find: Alk = 2[Ca +2 ] + 2[Mg +2 Other strong ] +[ [ All strong base cations] acid anions ] (9) in which other strong base cations includes sodium, potassium, and others, and strong acid anions includes sulfate, chloride, and perhaps others (all in equivalents per litre). This equation assumes that complexes such as CaOH + are negligible and that any other weak acid/base species (including NOM) are negligible in comparison to the species shown explicitly. To accomplish softening, a base is added to precipitate CaCO 3 and possibly Mg(OH) 2 ; for virtually all source waters, precipitation of CaCO 3 is initiated at a lower ph than that required to precipitate Mg(OH) 2. The cheapest base is lime (CaO); ironically, calcium is added to water in order to reduce its concentration. Sufficient carbonate is required to precipitate both the calcium that was initially present and that which was added as lime; if carbonate is not present in the water naturally, it can be added as soda ash (Na 2 CO 3 ). The reduction in the calcium concentration can be limited by the available carbonate. Softening reduces alkalinity from the point at which precipitation is initiated up until the remaining carbonate is no longer sufficient to precipitate further additions of calcium as lime. Most softening plants in the United States add lime and one or more coagulants (iron, alum, polymer, with iron being the most prevalent), whereas only a small fraction (<15%, according to data obtained for the Information Collection Rule [ICR]; USEPA, 2) add soda ash. Sludge recirculation is quite common. Based on the relative ratios of the calcium and magnesium concentrations, the typical hardness removal goals, and the substantial settling and sludge dewaterability difficulties associated with Mg(OH) 2 relative to CaCO 3, most plants operate with the intention of only precipitating CaCO 3. According to the ICR data, approximately two thirds of the softening plants had less than 4% removal of magnesium. Approximately 1% of the softening plants in the ICR database have split treatment, and all of those remove magnesium on one side of the plant. Some removal of magnesium occurs during CaCO 3 precipitation (Thompson et al, 1997). Natural organic matter. NOM in surface waters consists of a complex mixture of humic and nonhumic substances. The distribution of these compounds can vary substantially among water sources as well as temporally within a water source, making it difficult to use a single measure to quantify NOM. Total or dissolved organic carbon (TOC or DOC) is often used to characterize the overall NOM concentration, whereas ultraviolet (UV) absorbance at a wavelength of 254 nm (UV 254 ), hydrophobic DOC, and specific ultraviolet absorbance (SUVA) (the ratio of UV 254 to DOC) are commonly used to characterize the NOM constituents that are particularly reactive with respect to DBP formation. Preferential removal of the more-reactive NOM constituents within treatment processes can be of considerable benefit. NOM removal in lime softening is widely thought to occur through adsorption onto the surfaces of calcium and magnesium precipitates or by coprecipitation in which 98 NOVEMBER 23 JOURNAL AWWA 95:11 PEER-REVIEWED ROALSON ET AL

TABLE 1 Source water characteristics Parameter* Average Range ph 8.32 8.26 8.43 Calcium mg/l as CaCO 3 152 147 161 Magnesium mg/l as CaCO 3 79 75 84 Alkalinity mg/l as CaCO 3 181 169 19 DOC mg/l C 3.6 3.4 4.6 Hydrophobic DOC mg/l C 1.6 UV 254 cm 1.11.14.116 SUVA L/mg-m 2.8 2.4 3.1 DBPs µg/l UFC SUFC** DOX 52 53 THM 4 135 174 HAA 9 81 71 HAN 4 15 9.3 HK 2 1.7.69 CP.27.32 CH 15.5 16.3 *C carbon, CaCO 3 calcium carbonate, DOC dissolved organic carbon, SUVA the ratio of UV 254 to DOC, UV 254 ultraviolet absorbance at a wavelength of 254 nm Average DOC and hydrophobic DOC values are from a single sample with multiple analyses. DOX dissolved organic halogen, THM 4 four trihalomethanes, HAA 9 nine haloacetic acids, HAN 4 four haloacetonitriles, HK 2 two haloketones, CP chloropicrin, CH chloral hydrate Uniform formation conditions **Softening uniform formation conditions NOM is incorporated into the crystal structure of the precipitate. NOM removal by CaCO 3 is thought to occur early in the precipitation process when the CaCO 3 lattice structure is amorphous and the surface area is high. Within a few minutes, CaCO 3 solidifies into a negatively charged precipitate with a dense, crystalline structure and low surface area, characteristics that make it a poor adsorbent for negatively charged NOM (Liao & Randtke, 1986; Liao & Randtke, 1985). Liao and Randtke (1985) found that the more hydrophobic fraction of the NOM complexed Ca +2 more strongly, adsorbed to CaCO 3 better, and was removed to a greater extent than the more hydrophilic fraction. Thus, the same organic chemicals that are likely to react with chlorine to form DBPs (high-molecular-weight, oxygenated, aromatic compounds) appear to be preferentially removed by softening. NOM removal is also enhanced by the precipitation of Mg(OH) 2, probably because it is a positively charged precipitate with a loose structure and high surface area (Liao & Randtke, 1986; Liao & Randtke, 1985; Randtke et al, 1994; Randtke et al, 1982; Shorney & Randtke, 1994; Thompson et al, 1997). Thompson et al (1997) evaluated enhanced softening for the removal of NOM from nine natural waters. Their results showed that precipitation of CaCO 3 removed a modest amount of DOC. Removal increased dramatically at the point at which Mg(OH) 2 precipitation began, and the amount of DOC removed by Mg(OH) 2 increased with increasing initial DOC concentration. DOC removal was more effective for waters with larger SUVA values, and SUVA was a better quantitative predictor of the effectiveness of DOC removal than the percentage of hydrophobic organic carbon. Randtke et al (1982) found that the addition of other inorganic ions, specifically sodium and chloride, had no effect on the removal of organic compounds. The additions of iron and magnesium, on the other hand, improved removal. At low concentrations, the addition of phosphate increased the removal of organics, as long as it was simultaneously precipitated with CaCO 3. Sulfate at low concentrations (approximately 1 mg/l) dramatically increased the removal of humic acid, but higher doses mitigated this effect. In other work, low doses of iron improved removal, but higher doses had no additional effect (Randtke et al, 1994; Shorney & Randtke, 1994). To a large degree, this earlier work left open the question of the role of these inorganic constituents in NOM removal in softening. Liao and Randtke (1985) also found that sludge recycling increased the rate of CaCO 3 precipitation by providing seed crystals, thus increasing Ca +2 removal and improving softening. Sludge recycling also favors the formation of large calcite crystals, thus decreasing the adsorptive capacity and the removal of NOM. With a commercially prepared sludge, NOM removal decreased from 35% with no sludge recycle to 23% with a 2:1 ratio of recycled solids to fresh solids. Therefore, these limited results suggest that the common practice of sludge recycle may be detrimental to enhanced softening. Previous research provides considerable insight into the factors that may be of importance to the performance of enhanced softening. The objective of this research was to provide a comprehensive evaluation of these factors for a typical softening source water. Water chemistry and treatment practices may considerably affect enhanced softening. Both were evaluated in this research from the perspectives of softening performance, NOM removal, and DBP formation. MATERIALS AND METHODS Experimental procedures. At the beginning of the project, approximately 4 L of water was collected from Lake Austin, the drinking water source for the city of Austin, Texas. All experiments were conducted on this single batch of water, thus eliminating complications of source water variability in data interpretation. The water was stored in plastic carboys at 4 o C until used. The storage time between ROALSON ET AL PEER-REVIEWED 95:11 JOURNAL AWWA NOVEMBER 23 99

FIGURE 1 ph 12. 11. 1. 9. Effect of lime dose on ph 8. 5 1 15 2 25 3 CaO lime FIGURE 2 Ca +2 mg/l CaCO 3 2 15 1 5 Effect of lime dose on calcium ion concentration 5 1 15 2 25 3 Ca +2 calcium ion, CaO lime, CaCO 3 calcium carbonate collection and the experiments reported here varied from one to six months; no measurable water quality differences were found throughout this storage period. The experimental protocol was designed to simulate softening at the laboratory scale. Softening jar tests were conducted using a six-place gang-stirrer. 1 The jars, designed specifically for this project, were square (11.5 cm [4.53 in.] inside) with floating covers to minimize the amount of carbon dioxide that could dissolve into the water from the atmosphere, thus affecting ph as well as the precipitation of Ca +2 and Mg +2. The covers had an opening for the stirrer shaft and a very small tolerance around the edges and the shaft. Both the jars and covers were acrylic. The jars were designed with side ports to allow the addition of lime and other chemicals and the removal of the liquid after softening without removing the covers. The ports used stainless-steel fittings that were sealed with a septum, and chemicals were added by syringe injection. The withdrawal port was fitted with a plastic tube and a PTFE-coated valve, and suspension was withdrawn by gravity. Jar tests were performed on 9 ml of source water. The source water was added to the jars, and the floating covers were installed so that air bubbles between the cover and the sample were eliminated. For each chemical dose, the concentration of the chemical addition solutions was selected so that the volumes added to the jars would have a negligible effect on the total jar volume but could be easily and accurately measured with a 5-mL syringe. For each lime dose, the required amount of calcium hydroxide [Ca(OH) 2 ] was calculated and measured into a 15-mL plastic beaker, then mixed with 5 ml of distilled deionized (DDI) water 2 on a stir plate to create a slurry. The slurry was drawn into a 3-mL plastic syringe fitted with a 16-gauge needle. An additional 5 ml of DDI water was then added to the beaker to rinse it, and this additional volume was also drawn into the syringe. Rapid mixing was initiated at 15 rpm, and the appropriate volumes of chemical solution were injected into the jars simultaneously. After 1 min of mixing, the lime slurries were added to the jars simultaneously, and each syringe was filled and emptied from the jar to rinse it and ensure that all of the lime was transferred into the jar. Rapid mixing continued after lime addition for 3 min. The mixing speed was then slowed to 45 rpm for 3 min of flocculation, followed by an additional 3 min of quiescent settling. The velocity gradient (G) in the jars during mixing was calculated based on the work of Cornwell and Bishop (1983). G was approximately 35 s 1 for the rapid mix and 6 s 1 for the slow mix, values that fall within the wide range obtained in fullscale plants. Following settling, the liquid was decanted and passed through a.45-µm membrane filter 3 to simulate filtration in a treatment plant. To minimize leaching of organic carbon from the filter into the sample, the filter was prepared by rinsing with 5 ml of DDI water. The filtered water was analyzed for ph, calcium, magnesium, alkalinity, DOC, and UV 254. A portion of the filtered water was chlorinated under softening uniform formation conditions (SUFC), and the concentrations of dissolved organic halogen (DOX) and 21 specific DBPs were measured after chlorination. The uniform formation conditions (UFC) chlorination test was developed by Summers et al (1996) to represent the average conditions in US distribution systems. Conditions are as follows: 24±1 h incubation at 2.±1. o C and ph 8.±.2 with a chlorine residual of 1.±.4 mg/l. The SUFC test, developed as part of this research, is identical to the UFC test except that ph 9±.2 is used to better represent the conditions in distribution systems following softening plants. Analytical methods. Alkalinity and hardness (calcium and magnesium) were measured by titration and flame atomic absorption spectrometry, 4 respectively, in accor- 1 NOVEMBER 23 JOURNAL AWWA 95:11 PEER-REVIEWED ROALSON ET AL

dance with Standard Methods (1998). FIGURE 3 The alkalinity titration endpoint was determined from the slope of the ph 1 change. UV absorbance was measured using a UV/vis spectrophotometer 5 with 75 a 5-cm cell. A TOC analyzer equipped with an 5 autosampler 6 was used to measure 25 TOC, DOC, and hydrophobic DOC. All three measurements were taken for the source water, but only DOC was measured for the softened water. To measure hydrophobic DOC, a filtered sample was passed through a 1.1 cm (.43 in.) diameter column packed with FIGURE 4 a 1.5 cm (4.13 in.) depth of ion exchange resin 7 at a rate of approximately 15 bed volumes per hour to 2 15 remove the hydrophobic fraction of the organic carbon by sorption onto the 1 resin. The sample was then analyzed for organic carbon, and the value 5 detected was subtracted from the DOC concentration to find the hydrophobic DOC concentration. After termination of the incubation period for the UFC and SUFC tests, chlorine residual as free chlorine was measured by either amperometric titration 8 or spectrophotometry, in accordance with Standard Methods (1998). DOX was measured using a dissolved organic halogen analyzer 9 in accordance with Standard Methods (1998). The nine haloacetic acids (HAAs), four trihalomethanes (THMs), four haloacetonitriles (HANs), two haloketones (HKs), chloropicrin (CP), and chloral hydrate (CH) were measured by liquid liquid extraction and gas chromatography in accordance with methods 551.1 (USEPA, 1995a) and 522.2 (USEPA, 1995b) with minor modifications (Ralls, 1999). Three analyses were required to quantify all 21 compounds: one for HAAs; one for THMs, HANs, HKs, and CP; and one for CH. All DBP concentrations were measured using a gas chromatograph 1 equipped with an electron capture detector. RESULTS AND DISCUSSION Source water characteristics. As shown in Table 1, Lake Austin is a moderately hard water with a high percentage of the hardness (34%) occurring as magnesium. The high alkalinity provides sufficient carbonate for lime softening to remove calcium effectively without the addition of soda ash. The DOC concentration and UV 254 value are moderate, although some variation in the DOC values was observed among the different sample containers. The hydrophobic fraction of the DOC (44%) and the SUVA value are mid-range. Conventional wisdom holds that it is difficult to reduce SUVA below 2 Mg +2 mg/l CaCO 3 Effect of lime dose on magnesium ion concentration 5 1 15 2 25 3 CaO lime, CaCO 3 calcium carbonate, Mg +2 magnesium ion Alkalinity mg/l CaCO 3 Effect of lime dose on alkalinity 5 1 15 2 25 3 CaO lime, CaCO 3 calcium carbonate L/mg-m, suggesting that reducing the reactive fraction of Lake Austin NOM may be difficult. The DBP analyses showed the expected effect of ph on THM and HAA formation when UFC and SUFC results were compared. The source water is sufficiently reactive with respect to DBP formation to be of regulatory concern. Lime softening alone. Lime was added to 12 jars in doses varying from to 3 mg/l as CaO. As shown in Figure 1, the ph increased with increasing lime dose, reaching ph 11 and inducing Mg(OH) 2 precipitation at a dose of approximately 17 mg/l CaO. The change in calcium concentration is shown in Figure 2. Lake Austin water is calculated to be supersaturated with respect to CaCO 3, so even the smallest addition of lime caused precipitation and reduced the calcium concentration. That trend continued with doses up to 1 mg/l CaO. In the range of doses between 1 and 19 mg/l CaO, the soluble calcium remained essentially constant at its lowered value as the precipitation of CaCO 3 just matched the addition of Ca +2. At higher doses, further precipitation was limited by the lack of carbonate, so the calcium concentration increased; calculations from the ph and alkalinity data indicate that the available carbonate at these high doses was much less than the calcium addition. The magnesium concentration (Figure 3) decreased slightly up to a lime dose of approximately 17 mg/l CaO, presumably by adsorption or substitution of Mg +2 into the CaCO 3 precipitate. Then, as the ph increased to ROALSON ET AL PEER-REVIEWED 95:11 JOURNAL AWWA NOVEMBER 23 11

FIGURE 5 Effect of softening on Lake Austin water natural organic matter DOC FIGURE 6 Effect of softening on Lake Austin water natural organic matter SUVA 5. 3. 4. 2.5 DOC mg/l C 3. 2. SUVA L/mg m 2. 1.5 1. 1..5. 5 1 15 2 25 3 C carbon, CaO lime, DOC dissolved organic carbon. 5 1 15 2 25 3 CaO lime, SUVA specific ultraviolet absorbance 11 and higher, substantial magnesium removal occurred by precipitation of Mg(OH) 2. At a lime dose of 23 mg/l, almost all of the magnesium was removed. Theoretically, the alkalinity should change in the same way as the sum of the Ca +2 and Mg +2 concentrations (assuming no substantial change in the complexes such as CaOH + and MgOH + ). The results in Figure 4 approximated such a trend. Alkalinity decreased as calcium and magnesium concentrations decreased, reaching a minimum of approximately 5 mg/l CaCO 3 at a lime dose of approximately 13 mg/l CaO. The alkalinity was reasonably constant in the dose range from 1 to 23 mg/l CaO, reflecting the nearly constant Ca +2 and Mg +2 concentrations for the 1 17 mg/l doses and the offset of the rising Ca +2 concentration by the drop of Mg +2 concentration for the 17 23 mg/l doses. Finally, the alkalinity increased for the highest two doses because the calcium concentration increased even though the magnesium concentration was constant at nearly zero. The removal of NOM and DBP precursors is shown in terms of DOC in Figure 5. Careful inspection of the data indicates that the DOC concentration decreased in a linear fashion with increasing lime dose up to a dose of 19 mg/l CaO and then decreased more sharply between the lime doses of 19 and 21 mg/l, corresponding to the onset of significant Mg(OH) 2 precipitation. Finally, no additional DOC removal occurred at lime doses greater than 21 mg/l. DOC removal at the highest lime dose was 43%, which compares well with the 44% hydrophobic DOC content in the source water. At a dose of approximately 17 mg/l (just prior to Mg(OH) 2 precipitation), DOC removal was 29%. For waters with alkalinity values above 12 mg/l as CaCO 3 and TOC values of 48 mg/l, Stage 1 of the Disinfectants/Disinfection By-product Rule requires a 25% removal of TOC (USEPA, 1998). The DOC results suggest that the city of Austin might be able to meet this requirement and still avoid the voluminous sludge production associated with Mg(OH) 2 precipitation, although the minimum TOC removal was just barely reached in these results. These data also illustrate the relative NOM removal efficiencies of CaCO 3 and Mg(OH) 2. In the range of lime doses in which calcium removal dominated ( 1 mg/l CaO),.52 mg/l DOC and 125 mg/l of hardness as CaCO 3 were removed. In the range in which magnesium removal dominated (17 mg/l 21 mg/l CaO), approximately the same amount of DOC (.53 mg/l) was removed, with only 47 mg/l of hardness as CaCO 3. These results indicate that Mg(OH) 2 is a better adsorbent of NOM than CaCO 3, consistent with the findings of previous research. Nevertheless, the adsorption characteristics of CaCO 3 (or perhaps a mixed Ca Mg carbonate) are far from negligible and might be sufficient to achieve the USEPA s TOC removal standards for some waters. No additional DOC removal occurred at lime doses higher than 21 mg/l, presumably because no further precipitation of hardness occurred. Despite the improved removal of NOM with Mg(OH) 2 precipitation, existing softening plants that were designed for precipitation of CaCO 3 only and that treat water with a high fraction of their hardness coming from magnesium would generally prefer not to operate in this range. Mg(OH) 2 has a floc structure that includes a lot of water so that the particles are not nearly as dense as CaCO 3. Plants designed for CaCO 3 precipitation only are likely to experience substantial solid/liquid separation problems if the lime dose is increased to precipitate Mg(OH) 2 ; these problems include a higher solids carryover from clarifiers to filters, reduced filter runs, and sludge handling difficulties. Thus, operation at a dose just below that which induces Mg(OH) 2 precipitation appears to represent a good target for operating a plant for enhanced soften- 12 NOVEMBER 23 JOURNAL AWWA 95:11 PEER-REVIEWED ROALSON ET AL

FIGURE 7 DOX Concentration µg Cl /L 35 3 25 2 15 1 5 Total disinfection by-product formation in softened Lake Austin water under softening uniform formation conditions Concentration Yield 5 1 15 2 25 3 C carbon, CaO lime, Cl chloride ion, DOX dissolved organic halides 14 12 1 ing improving the NOM removal without drastically affecting the overall operation. SUVA, the ratio of UV 254 to DOC, represents the reactivity of the DOC with respect to DBP formation; as SUVA increases, NOM generally has greater aromaticity and is more hydrophobic in nature. If SUVA values are unchanged by a treatment process, then a decrease in UV 254 simply results from a decreasing NOM concentration. Decreasing SUVA values, on the other hand, indicate a relatively higher decrease in the reactive portion of the NOM. The SUVA values found in this experiment (Figure 6) decreased by half at the highest lime dose and dropped from 3.1 to 2.24 L/mg-m (28%) at the point at which magnesium precipitation began (approximately 17 mg/l CaO). Three distinct regions are apparent in the SUVA results. Up to a lime dose of about 1 mg/l CaO (where the Ca +2 concentration reached its minimum), SUVA decreased linearly with the dose. From a lime dose of 1 17 mg/l CaO, in which the Ca +2 and Mg +2 concentrations were stable, SUVA stayed relatively constant as well. As the lime dose increased from 17 mg/l CaO into the region of Mg +2 precipitation, SUVA dropped again, with most of the drop occurring by the 21 mg/l dose, coinciding with the Mg(OH) 2 precipitation. These results clearly indicate that the more reactive portion of the DOC was preferentially removed during Ca +2 and Mg +2 precipitation. The relative decreases in SUVA between Ca +2 precipitation (26%) and Mg +2 precipitation (37%) again point to Mg(OH) 2 as being the better NOM adsorbent, but the decrease associated with the Ca +2 precipitation is hardly insignificant. On the basis of the decreasing DOC concentration and SUVA, decreases in both DBP formation and yield would be expected as softening proceeds. The effect of lime softening alone on overall DBP formation, as characterized by DOX, is shown in Figure 7. DOX formation decreased dramatically as lime dose 8 6 4 2 DOX Yield µg Cl /mg C increased, resulting in a 55% drop from 316 µg/l Cl for the source water to 143 µg/l Cl at a lime dose of 28 mg/l CaO. Unlike the DOC and SUVA profiles, however, the DOX concentration decreased almost linearly through both the CaCO 3 and Mg(OH) 2 precipitation ranges. As with the softening and NOM results, the three highest lime doses yielded almost identical results as no further precipitation occurred. The DOX yield (i.e., the mass of DOX formed per unit mass of organic carbon) is also shown in Figure 7. This normalization is useful because waters with lower DOC concentrations generally produce lower DBP values. Although some scatter is evident (and one point is an apparent outlier), the data show a steady decrease (about 3%) in DOX yield with increasing lime dose, providing further evidence that enhanced softening preferentially removes the more reactive portion of the DOC. The effect of softening on subsequent formation of two individual THM species is shown in Figure 8. The chloroform (TCM) concentration decreased 55% (from 41 to 18 µg/l) as lime dose increased, whereas the dibromochloromethane (DBCM) concentration doubled, from 16 to 32 µg/l. In a similar fashion, bromodichloromethane (BDCM) showed a small increase from 27 to 31 µg/l, and the bromoform (TBM) concentration increased 2 times from.31 to 6.7 µg/l (data not shown). Thus, a clear trend toward increased formation of the bromine-substituted THMs was observed with increasing lime dose. A relevant factor is that Lake Austin water has a moderately high bromide concentration (16 µg/l), which makes the formation of bromine-substituted DBPs especially likely. Two additional trends can be discerned for each THM, one before complete Mg +2 removal (between lime doses of and 19 mg/l CaO) and one after almost all Mg +2 was removed (lime doses of 21 mg/l CaO and above). The TCM concentration dropped steadily in the first region, took a more precipitous drop between 19 and 21 mg/l CaO, and then leveled off at the high lime doses. The other three species (with only DBCM shown) all rose steadily in the first region, showed a small additional increase between 19 and 21 mg/l CaO, and leveled off at the high lime doses. Clearly, THM formation was closely tied to DOC removal; the formation and speciation of THMs stabilized only at the very high lime doses where no further Ca +2, Mg +2, or DOC removal occurred. The results for total THMs (TTHMs) are shown in both mass and molar units in Figure 8. On a mass basis, the TTHM concentration exhibited a small, steady rise (approximately 13%) up to a lime dose of 19 mg/l CaO, followed by a slight drop and plateau. The rise can be attributed to the changing THM speciation and the higher molecular weight of bromine (79.9 g/mol) compared with that of chlorine (35.5 g/mol). On a molar basis, the TTHM concentration stayed approximately constant up to a dose of 19 mg/l CaO, illustrating that the rising mass concentration was a molecular weight effect. At higher lime ROALSON ET AL PEER-REVIEWED 95:11 JOURNAL AWWA NOVEMBER 23 13

doses, THM formation on a molar basis dropped by approximately 11%. Thus, the shifting THM speciation did not affect TTHM formation until the fraction of the DOC associated with Mg(OH) 2 precipitation was removed. Although not of current regulatory concern, the shifting THM speciation in bromide-containing waters could be of importance if future regulatory schemes consider the concentration of individual THMs, rather than only the total concentration. The THM yield is shown in Figure 1 for TCM and DBCM and in Figure TABLE 2 11 for TTHMs. The TCM yield decreased slightly, whereas the BDCM (not shown), DBCM, and TBM (not shown) yields increased with increasing lime dose. The net result was an increasing yield of TTHMs with increasing lime dose. Although an increasing THM yield may seem surprising because of the decrease in DOC and SUVA (and even DOX) values with increasing lime doses, it is consistent with the anticipated effect of significant bromide concentrations (Symons et al, 1993). As the lime dose increased, the DOC concentration (and therefore the chlorine demand) decreased, and the chlorine dose in the SUFC test was decreased. Because bromide is not removed in softening, the Br :Cl 2 and Br :DOC ratios increased with increasing lime dose, thereby causing the shift in speciation and yield. HOCl is consumed very quickly by Br to form HOBr and Cl. An increasing Br :Cl 2 ratio, therefore, favors an increased relative concentration of HOBr in comparison to HOCl. HOBr attacks more sites on NOM and reacts with the sites faster than HOCl does. In addition, as the Br :DOC ratio increases, the water becomes more precursor-limited, and the faster-reacting HOBr may consume many or most of the active NOM sites before HOCl can react with them. A comparison of the concentrations and yields for DOX (Figure 7) and THMs (Figures 8 11) shows that THM formation behaved differently from that of DBP formation as a whole. The DOX measurement is essentially a molar measurement converted to mass units (i.e, all halogens are counted as Cl), so that the THM contributions to DOX concentration and yield follow the molar trends in Figures 9 and 11, not the mass trends. The DOX concentration decreased substantially with increasing lime dose, whereas the THM molar concentration was unaffected by lime dose until the onset of Mg(OH) 2 precipitation, and even then the decrease in THM formation was modest. Similarly, the DOX yield decreased with increasing lime dose, whereas the THM molar yield increased. These differing trends can be reconciled somewhat by noting that the THM contribution to the DOX was between 2 and 4% for all experimental conditions. Clearly, however, THM formation was not affected DBP* formation at selected lime doses DBP Concentrations and Yields Lime HAAs HANs HKs CP CH Dose mg/l µg/l µg/mg C µg/l µg/mg C µg/l µg/mg C µg/l µg/mg C µg/l µg/mg C 64 14 1.9.53.67.14.51.11 16 3.4 1 42 13 2.7.85.36.11.17.5 9.8 3. 19 3 12 3.4 1.4.43.16 ~ ~ 7.8 2.9 235 21 9.5 5.9 2.7 NA NA NA NA NA NA *CH chloral hydrate, CP chloropicrin, DBP disinfectant by-product, HAAs haloacetic acids, HANs haloacetonitriles, HKs haloketones 15 mg/l for HKs, CP, and CH Data not available as much by enhanced softening as that of other DBPs; THM formation is known to increase at high ph, so it is reasonable to think that the ph 9 of the SUFC favored the formation of THMs more than other DBPs. As noted earlier, the moderately high bromide concentration might also have played a key role in the THM response to enhanced softening. The concentrations and yields of HAAs, HANs, HKs, CP, and CH are shown in Table 2 for lime doses of, 1, 15, 19, and 235 mg/l. As shown in Figures 2 and 3, the 1 mg/l lime dose produced CaCO 3 precipitation, the 19 mg/l dose resulted in the onset of significant Mg(OH) 2 precipitation, and the 235 mg/l dose provided nearly complete Mg +2 removal. In contrast to the THMs, both the HAA concentration and yield decreased with increasing lime dose. As with the THMs, the concentration of chlorine-substituted HAAs decreased with increasing lime dose, but unlike the THMs, the concentration of bromine-substituted HAAs remained relatively constant over the range of lime doses, except for the bromochloroacetic acid concentration, which decreased. Throughout the range of lime doses tested, dichloroacetic acid was the dominant HAA species, followed by bromochloroacetic acid and trichloroacetic acid (Ralls, 1999). The HAN concentration and yield increased with increasing lime dose; as with the THMs, both increases are attributable to greater formation of bromine-substituted HANs with increasing lime dose. Data for HKs, CP, and CH are only available up to a lime dose of 15 mg/l. The concentrations decreased with increasing lime dose as would be expected from the results for the other DBPs, which showed a decrease in the formation of chlorine-substituted species with increasing lime dose. The CP and CH yields decreased with increasing lime dose, whereas the HK yield was stable. To summarize, DOX and the chlorine-substituted THMs, HAAs, and HANs, all decreased as the lime dose increased, as did HKs, CP, and CH. Conversely, brominesubstituted THMs and HANs clearly increased with increasing lime dose, in terms of both concentration and yield, whereas bromine-substituted HAAs showed no 14 NOVEMBER 23 JOURNAL AWWA 95:11 PEER-REVIEWED ROALSON ET AL

TABLE 3 Hardness and natural organic matter removal in the silicate screening test Silicate Added mg/l Si Lime Dose Constituent mg/l CaO* 11 33 Ca +2 mg/l CaCO 3 16 43 4 31 26 93 163 167 Mg +2 mg/l CaCO 3 16 6 31 2 26 2.5 1.6.2 DOC mg/l C 16 2.5 2.7 2.8 26 2. 2.3 2.8 SUVA L/mg-m 16 2.2 2.4 2.2 26 1.8 2. 2. *Lime Source water concentration approximately 5.5 mg/l Si C carbon, CaCO 3 calcium carbonate, DOC dissolved organic carbon The ratio of UV 254 to DOC change. The DBP formation trends with respect to bromine-substituted compounds would likely be different in waters containing low bromide concentrations. Magnesium screening test. In the lime-softening-alone experiments, DOC removal reached a maximum at the point at which nearly all of the magnesium was removed, thus begging the question of whether additional magnesium removal could promote further NOM removal. For the magnesium screening test, magnesium was added in varying concentrations to evaluate the effect of magnesium concentration on NOM removal at six lime doses (, 8, 125, 17, 25, and 26 mg/l CaO). At each lime dose, magnesium (in the form of magnesium chloride) was added in concentrations of, 18, and 36 mg/l Mg +2 before lime addition. These doses were selected as zero, one, and two times the source water magnesium concentration. Magnesium addition had no effect on the DOC, UV 254, or SUVA values attained after softening. This result has two possible implications. The first is that waters with higher magnesium concentrations or higher magnesium-to-calcium concentration ratios are not necessarily better candidates for enhanced softening. This seems unlikely, however, as other research has linked DOC removal strongly to magnesium removal (Liao & Randtke, 1986; Liao & Randtke, 1985; Randtke et al, 1994; Randtke et al, 1982; Shorney & Randtke, 1994; Thompson et al., 1997). The second possible implication is that no further DOC removal was possible because the DOC fraction susceptible to removal by softening (i.e., the hydrophobic fraction) had already been removed. This explanation seems more likely than the first, especially because the results from the lime-softening-alone test indicated that total DOC removal (43%) approximately equaled the hydrophobic fraction of the DOC (44%). Iron screening test. The addition of iron is commonly practiced in softening plants to improve solid liquid separation by coagulation. In plants that primarily practice coagulation, iron is known to precipitate as Fe(OH) 3 and reduce NOM concentrations. In this research, the addition of iron (added as ferric sulfate) was evaluated for its effect on softening and NOM removal at two lime doses, 13 and 21 mg/l CaO. Fe(OH) 3 is expected to precipitate at the ph conditions of these softening tests. These lime doses were selected to represent softening conditions with and without magnesium precipitation. Six jars were tested at each lime dose. At the low lime dose, iron was added in concentrations of.72, 1.44, 2.88, and 5.76 mg/l Fe to the first four jars before lime addition. At the high lime dose, iron doses of 1.44, 2.88, 5.76, and 11.52 mg/l Fe were added to the first four jars before lime addition. For both lime doses, iron doses of 1.44 and 5.76 mg/l Fe were added to the fifth and sixth jars after lime addition to evaluate whether the order of chemical addition would have any effect on softening performance or NOM removal. These iron doses were selected as multiples of the iron concentration currently added by the city of Austin. As with magnesium, the addition of iron did not significantly affect softening or NOM removal, and the order of chemical addition appeared to have no effect. At both lime doses, the addition of iron decreased the ph slightly (ferric sulfate is an acid), resulting in small increases in calcium concentrations. Magnesium concentrations stayed approximately constant over the range of iron doses for both lime doses. For the low lime dose, the DOC concentration remained approximately constant. For the high lime dose, the DOC concentration decreased slightly with increasing iron dose, from approximately 2. to approximately 1.7 mg/l. UV 254 stayed relatively constant at the low lime dose and decreased slightly with increasing iron dose at the high lime dose. No trend was discernible in the calculated SUVA values. Therefore, the use of iron appears to have, at best, a small effect on NOM removal in enhanced softening. In operating plants, the primary benefit of the addition of iron is to improve the removal of small particles. To the extent that these particles have adsorbed NOM, iron addition can improve NOM removal. In this research, the effect of iron on settling and deep bed filtration was not investigated. Phosphate screening test. The addition of phosphate (as potassium phosphate) was evaluated for its effect on softening and NOM removal at two lime doses, 145 and 26 mg/l CaO. Six jars were tested at each lime dose. At both lime doses, phosphate (P) was added in concentrations of 1, 3, 6, and 1 mg/l P to the first four jars prior to lime addition. These doses were selected as one, three, six, and ten times the finished water phosphate concentration (1 mg/l). In the fifth and sixth jars, phosphate ROALSON ET AL PEER-REVIEWED 95:11 JOURNAL AWWA NOVEMBER 23 15

doses of 1 and 3 mg/l P were added after lime addition to evaluate whether the order of chemical addition would have any effect on softening performance or NOM removal. For both lime doses, the addition of phosphate (and the order of chemical addition) had a negligible effect on softening and NOM removal. Sulfate screening test. The addition of sulfate (as sodium sulfate) likewise was evaluated for its effect on NOM removal at two lime doses, 16 and 26 mg/l CaO. At each lime dose, sulfate was added in concentrations of 4 and 12 mg/l SO 4 2 before lime addition. These doses were selected as one and three times the source water sulfate concentration. Aside from the suggestion that sulfate slightly improved the precipitation of magnesium, no other effect on softening or NOM removal was apparent. Silicate screening test. Scale that has formed in the city of Austin distribution system contains substantial fractions of silicate and magnesium, with silicate-to-magnesium ratios ranging from.84 to 1.45, and averaging 1.23 (Camet Research Inc., 1999). Because silicate is known to exacerbate scaling in Austin and is known to interact with Mg +2, an understanding of its effect on softening and NOM removal is important, especially if consideration is given to operating at a ph high enough to induce Mg(OH) 2 precipitation. The addition of silicate (added as sodium silicate) was evaluated for its effect on NOM removal at two lime doses, 16 and 26 mg/l CaO. At each lime dose, silicate (Si) was added in concentrations of 11 and 33 mg/l Si prior to lime addition. These doses were selected as two and six times the source water silicate concentration. The experimental results are summarized in Table 3. At the low lime dose, the calcium concentration decreased somewhat as the silicate dose increased, whereas at the high lime dose, the calcium concentration was substantially higher with the extra silicate addition. The magnesium concentration decreased substantially at both lime doses: from 6 to 2. mg/l as CaCO 3 at the low dose and from 2.5 mg/l to.2 mg/l as CaCO3 at the high dose. These results strongly indicate that silicate forms a precipitate with magnesium capable of removing nearly all of the magnesium. Results from the water chemistry computer program 11 (Westall et al, 1976) indicate that the most likely precipitate (from a viewpoint of thermodynamic stability) is tremolite (Morel & Hering, 1993): 2CaO 5MgO 8SiO 2 H 2 O + 14H 2 O 2Ca +2 + 5Mg +2 + 8H 2 SiO 3 + 14OH K SO = 1 133 Kinetic considerations often limit the formation of mixed solids, and several other magnesium silicate precipitates are possible, such as forsterite, enstatite, and diopside (Morel & Hering, 1993). Nevertheless, it is clear from the magnesium results that increased silicate concentration increased magnesium removal, making it likely that at least one of the possible magnesium-silicate precipitates is forming. In the absence of additional analysis of the sludge formed during this screening test, the exact composition of the precipitate formed is unknown. To explore the equilibrium chemistry at work in the precipitation of magnesium-silicate mixed solids, the following discussion focuses on the formation of tremolite. For tremolite to form, the reaction product ([Ca +2 ] 2 [Mg +2 ] 5 [H 2 SiO 3 ] 8 [OH ] 14 ) must exceed the solubility constant, 1 133. Because of the stoichiometry, as the silicate dose increases, the reaction product increases exponentially. With the addition of 11 mg/l Si (yielding a total concentration three times that of the source water), the reaction product increases by a factor of 3 8, and by adding 33 mg/l (yielding a concentration seven times that of the source water), the reaction product increases by a factor of 7 8. Also, the reaction product increases by a factor of 1 14 for every unit increase in ph. The magnesium concentration in the source water is approximately 75 mg/l CaCO 3, or.75 mmol. The silicate concentration is 5.5 mg/l Si, or.2 mmol, yielding a molar silicate-to-magnesium ratio of.27:1. Ignoring any complexation of the constituents of tremolite, the source water concentrations suggest that the water is nearly at equilibrium with tremolite. Because precipitation usually does not occur until the solubility product is exceeded by a substantial amount, tremolite is not likely to form in the source water. With a silicate addition of 11 mg/l Si, the silicate concentration increases to.59 mmol and tremolite formation is more likely. Precipitation of all the silicate as tremolite would require.49 mmol Mg +2, leaving.26 mmol Mg +2 in solution. This value corresponds well with the results for the low lime dose of the silicate screening test, in which.31 mmol Mg +2 (31 mg/l as CaCO 3 ) was left in solution at the low silicate dose. With the silicate addition of 33 mg/l Si, the silicate concentration increased to 1.3 mmol. Precipitation of all the silicate as tremolite would require.98 mmol Mg +2, leaving mmol Mg +2 in solution. Again, this corresponds well with the results for the low lime dose, in which.2 mmol Mg +2 (2. mg/l CaCO 3 ) was left in solution at the high silicate dose. These results indicate that source water silicate concentrations may be very important to softening. For many years before initiation of recarbonation at the treatment plants, the treated water in the city of Austin was generally supersaturated with respect to one or more magnesium silicate solids, leading to precipitation and scaling in the distribution system. With respect to NOM removal in softening, the SUVA value at each lime dose was relatively unaffected, but silicate addition had a slight effect on the DOC concentration at the low lime dose. At the high lime dose, the DOC concentration increased from 2. mg/l for the condition in which no silicate was added (interpolated from the lime-only data) to 2.8 mg/l when 33 mg/l Si was added. These results suggest that silicate addition, rather than improving enhanced softening, actually impairs it. This 16 NOVEMBER 23 JOURNAL AWWA 95:11 PEER-REVIEWED ROALSON ET AL

might seem surprising, given the increased (and dramatic) removal of magnesium with increased silicate addition. However, increased concentrations of silicate appear to increase magnesium removal through precipitation of magnesium-silicate solids, which apparently do not adsorb NOM as readily as Mg(OH) 2. Perhaps the presence of silicate is the reason NOM removal during lime softening of Lake Austin water is not improved with increasing magnesium removal as dramatically as reported by other researchers (Liao & Randtke, 1986; Liao & Randtke, 1985; Randtke et al, 1994; Randtke et al, 1982; Shorney & Randtke, 1994; Thompson et al, 1997). Lime softening with sludge recycle. Sludge recycle is commonly used in softening plants to improve the rate and extent of precipitation and settling in sedimentation basins. To evaluate the effect of sludge recycle on softening, NOM removal, and DBP formation, an experiment was performed using three lime doses (, 125, and 25 mg/l CaO), three sludges (one collected from Austin s Ullrich Water Treatment Plant and two that were created in the laboratory using the two different lime doses), and four levels of sludge recycle (, X, 2X, and 4X). The variable X was intended to represent two thirds of the city of Austin s sludge recycle condition (X target =.6 g/l), but at the time of the softening jar tests, the solids concentrations of the sludges were unknown. The volumes of sludge added to the jars to achieve the desired solids concentration were estimated, and the solids concentration in each jar was calculated later, after the solids content of each sludge was determined. At 125 mg/l CaO, the solids doses were,.67, 1.3, and 2.7 g/l. At 25 mg/l CaO, the solids doses were,.5, 1., and 2. g/l. At both lime doses, the calcium concentration was unchanged with increasing solids doses. The magnesium concentration was unchanged at the low lime dose, although at the high lime dose, the magnesium concentration dropped from 13 to 2 mg/l CaCO 3 at the first increment of sludge recycle, suggesting that sludge recycle is more beneficial for a plant that is removing Mg +2 than for a plant that is removing Ca +2 alone. At both lime doses, the DOC concentration varied by less than 1%, and SUVA values varied by less than 12%. These results suggest that sludge recycle has essentially no effect on DBP precursor removal, at least under the conditions of the jar tests. Chlorination and subsequent analysis for DBPs were performed on a subset of samples. At the low lime dose, the DOX concentration decreased with increasing solids dose, but only by 7%, indicating that the addition of sludge recycle has little effect on DBP formation. In contrast, sludge recycle increased THM formation by 21%, from 8 to 97 µg/l, at the low lime dose and by 6%, from 72 to 76 µg/l, at the high lime dose. Sludge recycle had an insignificant effect on the HAA and HAN concentrations. The total HK concentration remained unchanged at the low lime dose but decreased at the high lime dose because TCP was not detected at the higher FIGURE 8 THMs µg/l 5 4 3 2 1 Formation of individual THMs in softened Lake Austin water under softening uniform formation conditions Chloroform Dibromochloromethane 5 1 15 2 25 3 CaO lime, THMs trihalomethanes FIGURE 9 TTHM Concentration µg/l 1 8 6 4 2 TTHM formation in softened Lake Austin water under softening uniform formation conditions Concentration Yield. 5 1 15 2 25 3 Lime Dose mg/l C carbon, CaO lime, TTHMs total trihalomethanes solids concentration. The CP concentration dropped from.18 µg/l with no sludge recycle to below the detection limit at the first addition of solids; no CP was detected at the high lime dose. The CH concentration decreased at the low lime dose, from 13.5 to 9.9 µg/l, but remained unchanged at the high lime dose. In sum, although some variation was seen in the DBP results for lime softening with sludge recycle, most of the changes were small, and no clear trend emerged showing any advantage to recycling sludge in these jar tests. The primary advantage of sludge recycling in softening plants is to increase the rate of precipitation and the rate and extent of flocculation, with associated advantages in sedimentation. The design of these jar tests was not intended to test those advantages (which are well established), but only to see if there were additional effects with respect to DBP formation. The fact that no effects, positive or negative, on DBP formation were found does not detract from the operational advantages in treatment plants. 1..8.6.4.2 TTHM Yield µmol/l ROALSON ET AL PEER-REVIEWED 95:11 JOURNAL AWWA NOVEMBER 23 17

FIGURE 1 THM Yield µg/mg C 2 15 1 5 Yield of individual THMs in softened Lake Austin water under uniform softening formation conditions 5 1 15 2 25 3 Lime Dose mg/l C carbon, CaO lime, THMs trihalomethanes FIGURE 11 TTHM Yield µg/mg C 5 4 3 2 1 Chloroform Dibromochloromethane Yield of TTHMs in softened Lake Austin water under uniform softening formation conditions Mass basis 5 1 15 2 25 3 Molar basis C carbon, CaO lime, TTHMs total trihalomethanes SUMMARY AND CONCLUSIONS The use of enhanced softening in drinking water treatment to reduce DBP formation is complex. Waters that require softening typically have higher ph and alkalinity values than other source waters, and the softening process raises the ph even further to induce calcium precipitation and, in some cases and at even higher ph, magnesium precipitation. THM formation generally increases at higher ph values, whereas the formation of DOX and other DBPs generally decreases. Waters with a high bromide concentration (such as Lake Austin) can form bromine-substituted DBPs preferentially and more rapidly than chlorine-substituted DBPs. Because bromine-substituted DBPs have higher molecular weights than their chlorine-substituted counterparts, waters with a higher bromide concentration may produce a higher mass concentration of DBPs, thus affecting the ability of treatment plants using such waters to meet DBP regulations. In addition, brominesubstituted DBPs are thought to be of greater health risk than chlorine-substituted DBPs (Bull, 1995; Pegram, 1995). The effect of enhanced softening on DBP speciation in such waters is therefore important..5.4.3.2.1 TTHM Yield µmol/mg C Enhanced softening performed on Lake Austin water improved NOM removal, achieving a maximum DOC removal of 43%. As others have observed, Mg(OH) 2 was more effective than CaCO 3 in NOM removal, although this difference was not very dramatic. Enhanced softening also lowered SUVA values, decreased DOX formation, and shifted DBP speciation toward bromine-substituted compounds. The concentrations of HAAs, HKs, CP, and CH decreased with increasing lime dose as a result of both a decreasing DOC concentration and preferential removal of the more reactive DOC fraction (i.e., SUVA decreased). The THM and HAN concentrations, however, increased with increasing lime dose, because the decreasing concentrations of chlorine-substituted species were more than offset by the increasing concentrations of bromine-substituted species. The Br :Cl 2 and Br :DOC ratios increased with increasing lime dose, thereby causing greater formation of bromine-substituted species. Results from screening tests that examined the effects of changing the source water characteristics through chemical additions (iron, phosphate, sulfate, and silicate) and the addition of recycled sludge did not indicate any ways to improve enhanced softening. The silicate addition experiments suggested that large concentrations of silicate might have a detrimental effect on NOM removal, presumably through the precipitation of less desirable magnesium silicate solids at the expense of magnesium hydroxide formation. The results of this research have important implications for water treatment plants that currently use softening. First, the success of enhanced softening for NOM removal suggests that some utilities might be able to meet the Stage 1 requirement for TOC removal by increasing lime doses above that required for softening (the point at which calcium concentration is at a minimum) but not so high as to induce magnesium precipitation (and thus avoid the sludge handling problems associated with Mg(OH) 2 ). For example, the Lake Austin water tested in this research appears to meet the regulations, but just barely. Second, enhanced softening preferentially removes certain fractions of the NOM, such that the NOM left after softening will have different characteristics than the NOM in the source water, which in turn affects the formation and speciation of DBPs. In addition, plants that operate in the range of CaCO 3 precipitation only (i.e., that avoid Mg(OH) 2 precipitation) might be able to remove sufficient DOC to meet current regulations for TOC or DBPs, but further removal of NOM is possible with Mg(OH) 2 precipitation. Thus, the effect of enhanced softening on DBP formation and speciation for such a plant would be less than the effect for a plant operating in the range of Mg(OH) 2 precipitation. This result is especially true of waters with high bromide concentrations, such as Lake Austin. ACKNOWLEDGMENT This research was funded by the US Environmental Protection Agency, Office of Research and Develop- 18 NOVEMBER 23 JOURNAL AWWA 95:11 PEER-REVIEWED ROALSON ET AL

ment, Water Supply and Water Resources Division, under Cooperative Agreement CR826395, Keith Kelty, Project Officer. The views expressed herein are those of the authors alone and do not necessarily reflect policies of the USEPA. Additional funding was provided by the city of Austin. ABOUT THE AUTHORS: Shay Ralls Roalson is an engineer and project manager with Turner, Collie and Braden Inc. in Austin, Texas. At the time of this research, Roalson (then known as Shay Ralls) was a graduate student at the University of Texas at Austin where she received her master s degree in Environmental and Water Resourcces Engineering. Roalson was awarded the 23 Texas Section AWWA Young Professionals Maverick Award and is a member of AWWA, WATER For PEOPLE, and the Water Environment Federation. JiHyang Kweon is an assistant professor in the Department of Environmental Engineering at Konkuk University in Seoul, Korea. Desmond F. Lawler 12 is the W.A. Cunningham Professor in Engineering at the University of Texas at Austin, Department of Civil Engineering, 1 University Station C1786, Austin, TX 78712-273; e- mail dlawler@mail.utexas.edu. Gerald E. Speitel Jr. is the John J. McKetta Professor in Engineering in the Department of Civil Engineering at the University of Texas at Austin. FOOTNOTES 1 779-4, Phipps and Bird, Richmond, Va. 2 MILLI-Q Water Purification System, Millipore Corp., Bedford, Mass. 3 HA WP, Whatman, Clifton, N.J. 4 Atomic absorption spectrometer, Model 238, PerkinElmer, Wellesley, Mass. 5 Lambda 38 UV/Vis spectrophotometer, PerkinElmer, Wellesley, Mass. 6 DC-18, Dohrmann/Envirotech, Santa Clara, Calif. 7 Supelco 2278, XAD-8, Sigma-Aldrich Inc., St. Louis, Mo. 8 CL Titrimeter 397, Fisher Scientific International Inc., Hampton, N.H. 9 TOX-1 with a TXA-2 adsorption module, Mitsubishi, Cypress, Calif. 1 589A gas chromatograph, Hewlett Packard, Avondale, Pa. 11 MINEQL: A Computer Program for the Calculation of Chemical Equilibrum Composition of Aqueous Systems, Department of Civil Engineering, Massachusetts Institute of Technology 12 To whom correspondence should be addressed If you have a comment about this article, please contact us at journal@awwa.org. REFERENCES Bull, R.J., 1995. Carcinogenic Properties of Brominated Haloacetates. Workshop Report, Disinfection By-Products in Drinking Water: Critical Issues in Health Effects Research. International Life Sciences Institute Health and Environmental Sciences Institute, Washington. Camet Research Inc., 1999. Elemental and Phase Characterization of Deposits From the Water and Waste Water Utility Distribution System of the City of Austin. Report to Mehrdad Morabbi, Water and Waste Water Utility, City of Austin. Cornwell, D.A. & Bishop, M.M., 1983. Determining Velocity Gradients in Laboratory and Full-Scale Systems. Jour. AWWA, 75:9:47. Liao, M.Y. & Randkte, S.J., 1986. Predicting the Removal of Soluble Organic Contaminants by Lime-Softening. Water Res., 2:1:27. Liao, M.Y. & Randkte, S.J., 1985. Removing Fulvic Acid by Lime-Softening. Jour. AWWA, 77:8:78. Morel, F.M.M. & Hering, J.G., 1993. Principles and Applications of Aquatic Chemistry. John Wiley & Sons Inc., New York. Pegram, R.A., 1995. Use of Mechanistic and Pharmacokinetic Data: Bromodichloromethane as a Case Study. Workshop Report, Disinfection By-Products in Drinking Water: Critical Issues in Health Effects Research, International Life Sciences Institute Health and Environmental Sciences Institute, Washington. Ralls, S.M., 1999. Enhanced Softening for Disinfection By-Product Precursor Control. Master s thesis, University of Texas at Austin. Randkte, S.J. et al, 1994. A Comprehensive Assessment of DBP Precursor Removal by Enhanced Coagulation and Softening. Proc. 1994 AWWA Ann. Conf., New York. Randkte, S.J. et al, 1982. Removing Soluble Organic Contaminants by Lime-Softening. Jour. AWWA, 74:4:192. Shorney, H.L. & Randtke, S.J., 1994. Enhanced Lime Softening for Removal of Disinfection By-Product Precursors. Proc. 1994 AWWA Ann. Conf., New York. Standard Methods for the Examination of Water and Wastewater, 1998 (2th ed.). APHA, AWWA, and WEF, Washington. Summers, R.S. et al, 1996. Assessing DBP Yield: Uniform Formation Conditions. Jour. AWWA, 88:6:8. Symons, J.M. et al, 1993. Measurement of THM and Precursor Concentrations Revisited: The Effect of Bromide Ion. Jour. AWWA, 85:1:51. Thompson, J.D. et al, 1997. Enhanced Softening: Factors Influencing DBP Precursor Removal. Jour. AWWA, 89:6:94. US Environmental Protection Agency (USEPA), 2. Information Collection Rule Auxiliary 1, Database Version 4, Office of Water, Washington. USEPA, 1998. National Primary Drinking Water Regulations: Disinfectants and Disinfection Byproducts. Fed. Reg., 63:241:69389. USEPA Office of Research and Development, 1995a. Method 551.1 Determination of Chlorination Disinfection Byproducts, Chlorinated Solvents, and Halogenated Pesticides/Herbicides in Drinking Water by Liquid Liquid Extraction and Gas Chromatography With Electron Capture Detection. Methods for the Determination of Organic Compounds in Drinking Water, Supplement III, EPA/6/R-95/131. USEPA Office of Research and Development, 1995b. Method 552.2 Determination of Haloacetic Acids and Dalapon in Drinking Water by Liquid Liquid Extraction, Derivatization and Gas Chromatography with Electron Capture Detection. Methods for the Determination of Organic Compounds in Drinking Water, Supplement III, EPA/6/R-95/131. Westall, J.C.; Zachary, J.L.; & Morel, F.M.M., 1976. MINEQL: A Computer Program for the Calculation of Chemical Equilibrium Composition of Aqueous Systems. Technical Note No. 18, Water Quality Laboratory, Department of Civil Engineering, Massachusetts Institute of Technology. ROALSON ET AL PEER-REVIEWED 95:11 JOURNAL AWWA NOVEMBER 23 19