Daily Freeze–thaw Cycles Affect the Transport of Metals in Streams Affected by Acid Drainage

Stream flow rates with seasonal, daily, or hourly cycles due to freezing and thawing can control downstream chemical processes by changing the mixing ratio of reactive flows. The extent of these hydrologic-chemical interactions has not been fully realized yet. This work explored the link between daily freeze-thaw cycles and the fate and transport of metals at a model stream impacted by acid drainage. We characterized hydrological and physicochemical parameters at the confluence between the Caracarani River (pH ~8.6) and the Azufre River (pH < 2), in northern Chile. Hourly water depth, temperature, and electrical conductivity monitoring coupled with pH and turbidity measurements revealed that maximum flow rates from thawing of the Azufre River induced characteristic daily drops in pH and turbidity. Shifts in pH controlled the precipitation and dissolution of arsenic-rich iron and aluminum reactive phases. Thus freeze-thaw processes are critical at streams receiving acid drainage where iron and aluminum phases form and are likely to impact the fate and transport of toxic metals in the system. This work highlights the importance of assessing the hydrological controls on flows and mixing ratios when studying the chemical reactivity and fate of contaminants at systems affected by acid drainage.


Introduction
Water quality may be strongly affected by annual, seasonal, and daily periodicity in stream flow rate [1][2][3][4][5].Changes in water quality can be caused by incoming waters that can dilute or concentrate chemical species [1][2][3][6][7][8] or incorporate sediments or nutrients into the rivers [1,[9][10][11].These processes can exert an important control in chemical speciation.Snow melting and freeze-thaw cycles are an example of processes that produce important changes in the flow rate and quality of surface waters.In terms of physical processes, freeze-up slows down or fully detains flow rate [7,12].In terms of chemical processes, dissolved salts are displaced during the freezing process to the liquid phase.This is especially critical in systems affected by streams of acid drainage (pH < 3).The melting of acid drainage produces trace metals, abrupt peaks in the concentration of anions (SO 4 ´2 and Cl ´), and location, extreme weather conditions, and safety protocols, manual sampling and operator-dependent measurements were taken only during the daytime (9 a.m.-5 p.m.).Such constraints did not compromise pursuing the objectives of the study.Continuous measurements of flowrate, temperature, and electric conductivity were performed with data-logging sensors deployed in situ (described below).
Water 2016, 8, 74 3 of 27 dependent measurements were taken only during the daytime (9 a.m.-5 p.m.).Such constraints did not compromise pursuing the objectives of the study.Continuous measurements of flowrate, temperature, and electric conductivity were performed with data-logging sensors deployed in situ (described below).Water 2016, 8, 74 3 of 27 dependent measurements were taken only during the daytime (9 a.m.-5 p.m.).Such constraints did not compromise pursuing the objectives of the study.Continuous measurements of flowrate, temperature, and electric conductivity were performed with data-logging sensors deployed in situ (described below).

Hydrologic Measurements
The rainfall and ambient temperature were recorded every 15 min by the meteorological station located 8.5 km from the confluence at an elevation of approximately 3900 m (Figure 1).The station was equipped with a data logger (Em50 Data Logger, Decagon Devices, Pullman, WA, USA) connected to a temperature sensor (ECH2O sensor, Decagon Devices, Pullman, WA, USA) and a rain gauge (ECRN-50, Decagon Devices, Pullman, WA, USA).
Water pressure, water temperature, and EC were recorded hourly at the Azufre River and Caracarani River using a data logger (Aqua TROLL 200 Data Logger, InSitu Inc., Bingen, WA, USA).The data presented corresponded to records from January to November 2011.The sensors were installed 700 m upstream of the confluence at the Azufre River and 100 m upstream at the Caracarani River (points a and b in Figure 1, respectively).Conductivity sensors were calibrated with a KCl solution 0.01 M and 1.41 mS¨cm ´1 at the Caracarani River, and a KCl solution 0.1 M and 12.8 mS¨cm ´1 at the Azufre River through several field campaigns.Water pressure values in both rivers were corrected using atmospheric pressure values recorded with a barometric sensor (Baro Troll Data Logger, InSitu Inc., Bingen, WA, USA) and utilized to further calculate water depth and estimate stream flow rates through rating curves [35].These curves were computed after several field campaigns between September 2010 and March 2013, which allowed for obtaining eight points of the rating curve for the Azufre River and four points for the Caracarani River.Stream flow velocity (ms ´1) and cross section area (m 2 ) were measured using a handheld acoustic Doppler velocimeter (Flowtracker Handheld ADV, Sontek, San Diego, CA, USA) equipped with an interface to estimate stream flow rate (m 3 ¨s´1 ).Finally, analytical expressions were fitted to these curves in order to automate the estimation of the flow rates from the measured water levels.

Hydrochemical Measurements and Sampling
Turbidity and pH were measured hourly from 9 a.m. to 5 p.m. during a field campaign in August 2012 at points a and b and at the stream segment downstream of the confluence indicated in Figure 1 (from the confluence to 1 km downstream).Downstream measurements were done at 90, 180, 320, and 1000 m from the junction.A Portable Turbidity Meter (HI 98713, Hanna Instruments, USA) was used and calibrated daily with standardized solutions (0.1 NTU, 1 NTU, 10 NTU, and 100 NTU).The pH and conductivity were measured with a multi-parameter probe (Hach HQ40d portable meter, Hach Company, Loveland, CO, USA).The calibration of the pH sensor was performed using buffer solutions of pH 4, 7, and 10 (Hach Company, Loveland, CO, USA).Correct functioning of the pH electrode at pH < 4 were corroborated by plotting pH measurements against potential measurements (mV) as done by Nordstrom et al. [36].
Turbidity was selected as an in situ proxy measurement for particle formation [37,38].Field measurements at the site between 2008 and 2012 showed that the concentration of suspended solids larger than 0.45 µm (y value) ranged between 1 and 43 mg¨L ´1, and were significantly correlated with turbidity (x value) through a mathematical regression of y = 0.71x ´0.56 (coefficient of determination R 2 = 0.72, with a level of significance of α = 5% according to a standard normal distribution two-tailed statistical test) [39].
Suspended solids were sampled through several field campaigns (see Table 2).These were collected by using a portable barrel filter of 2 L capacity (Barrel Filter Holder, Geotech Environmental Equipment Inc., Denver, CO, USA).Suspended solids were retained onto pre-weighed cellulose acetate membranes (>0.45 µm from the same company).Aqueous samples (non-filtered samples) were collected in the field, then filtered and stored at 4 ˝C when conditions allowed.

Analytical Methods
Membranes containing suspended solids were dried at 40 ˝C for 24 h and then placed inside Teflon tubes along with 12 mL of aqua regia (HCl and HNO 3 ) for microwave digestion (Mars Microwave System, MarsXpress-5, CEM Corporation, Matthews, NC, USA), according to the EPA protocol number 3015A [40].Blank membranes were also digested for quality control.
Digested aqueous and solid field samples were analyzed for iron, aluminum, and arsenic through Inductively Coupled Plasma-Optical Emission Spectrometry (ICP-OES, Optima 7300 V, Perkin Elmer, USA).

Diurnal Cycles in Stream Flow Rates, Temperature, and Electric Conductivity
The flow rates of the Azufre River (Q Azufre ) and Caracarani River (Q Caracarani ) showed daily cycles, as illustrated in Figure 4, where flow rates measured in August 2011 are shown (additional data is presented in the Supplementary Materials).The periodicity of the time series was identified visually and by means of the autocorrelation function, which determines how the variable x measured at time t correlates with the measured value at time t-k, where k = 1,2, . . .n is the so called "lag-time" [40].This test was performed using the Signal Processing Toolbox of Matlab.A strong 24 h periodicity for Q Azufre and Q Caracarani was found, as well as for the temperature and the EC of the Azufre River (T Azufre and EC Azufre , respectively).As a consequence, the proportion of Q Azufre , defined as Q Azufre /(Q Azufre + Q Caracarani )), exhibited the same periodicity.
Daily cycles experienced by Q Azufre were attributed to the freezing and thawing of the stream.The Azufre River typically remains at 0 ˝C from 9 p.m. to 10 a.m. the next morning when the ambient temperatures are negative and can reach down to ´13 ˝C during the colder months (Figure 2).The partial freezing of the Azufre River upstream causes the stream flow rate to drop while the water temperature remains at 0 ˝C.Once the temperature passes the 0 ˝C threshold, the melting causes a sudden peak of Q Azufre .This process occurred on a daily basis while the ambient temperature was low enough for the stream to remain partially frozen (Figure 4).EC Azufre also exhibited cyclic behavior.This process does not occur between December and April, when stream freezing does not take place and rainfall drives stream flows (see Supplementary Materials).
A temporal cross-correlation analysis (i.e., a lagged correlation analysis between two time series shifted in time relative to one another) [41] was used to detect the correlation between Q Azufre and T Azufre data, and between T Azufre and EC Azufre data, which also experience cyclic behavior (data based on the whole 2010-2011 record).Results show that T Azufre and Q Azufre correlated (r = 0.38) with a lag time of 7 h, (i.e., Q Azufre reaches its maximum value around 7 h after T Azufre reaches its maximum value, see Figure 5).Similarly, a cross correlation was observed between T Azufre and EC Azufre (r = ´0.63)with a lag time of 1 h (i.e., EC Azufre reached its minimum value around 1 h after after T Azufre reaches its minimum value).Note that higher values of r were not reached because the water temperature is constrained in the low range to 0 ˝C, and thus the variability of T Azufre is more limited than that of Q Azufre and EC Azufre .However, a standard normal distribution two-tailed statistical test indicated both cross-correlations were statistically significant for a level of significance of α = 5%.Overall, the temporal crossed correlations between these variables further demonstrate that Q Azufre was controlled by the freeze-thaw cycles.
Water 2016, 8, 74 7 of 27 A temporal cross-correlation analysis (i.e., a lagged correlation analysis between two time series shifted in time relative to one another) [41] was used to detect the correlation between QAzufre and TAzufre data, and between TAzufre and ECAzufre data, which also experience cyclic behavior (data based on the whole 2010-2011 record).Results show that TAzufre and QAzufre correlated (r = 0.38) with a lag time of 7 h, (i.e., QAzufre reaches its maximum value around 7 h after TAzufre reaches its maximum value, see Figure 5).Similarly, a cross correlation was observed between TAzufre and ECAzufre (r = −0.63)with a lag time of 1 h (i.e., ECAzufre reached its minimum value around 1 h after after TAzufre reaches its minimum value).Note that higher values of r were not reached because the water temperature is constrained in the low range to 0 °C, and thus the variability of TAzufre is more limited than that of QAzufre and ECAzufre.However, a standard normal distribution two-tailed statistical test indicated both cross-correlations were statistically significant for a level of significance of α = 5%.Overall, the temporal crossed correlations between these variables further demonstrate that QAzufre was controlled by the freeze-thaw cycles.
The partial freezing of the Caracarani River upstream of the confluence occurs to a lesser extent than for the Azufre River (see Figures 4 and 5).Maximum QCaracarani values occur between noon and 3 p.m., when the temperature of the Caracarani River (TCaracarani) was also at a maximum.These cyclic variations also affect the water quality downstream from the confluence, as the supply of acidity from The partial freezing of the Caracarani River upstream of the confluence occurs to a lesser extent than for the Azufre River (see Figures 4 and 5).Maximum Q Caracarani values occur between noon and 3 p.m., when the temperature of the Caracarani River (T Caracarani ) was also at a maximum.These cyclic variations also affect the water quality downstream from the confluence, as the supply of acidity from the Azufre River is counterbalanced by the alkalinity of the Caracarani River (Q Caracarani is 2 to 12 times higher than Q Azufre ).
Maximum stream flowrates at both streams occurred within distinct times of the day.Indeed, 65% of minimum total flow rates in both the Azufre and Caracarani River occurred between midnight and 9 a.m., when both rivers tended to remain partially frozen according to temperature measurements (remained at 0 ˝C).Seventy-two percent of maximum Q Azufre values occurred between 8 a.m. and noon, when the temperature of the Azufre River passed the 0 ˝C threshold.These variations were coincident with the mixing proportion between the streams (Q Azufre /(Q Azufre + Q Caracarani )) (72% of maximum Q Azufre /(Q Azufre + Q Caracarani ) values occurred between 8 a.m. and noon).Finally, maximum flow rates in the Caracarani River, as 77% of maximum Q Caracarani values occurred between noon and 4 p.m.
Variations in the time Q Caracarani reached its maximum might be explained by freezing-melting cycles occurring near its origin, with delayed effects due to the distance to the confluence.Maximum values of Q Caracarani /(Q Azufre + Q Caracarani ), however, were widely spread (63% of these values occurred between noon and midnight).This was due to decreased flow rates towards the evening.Maximum stream flowrates at both streams occurred within distinct times of the day.Indeed, 65% of minimum total flow rates in both the Azufre and Caracarani River occurred between midnight and 9 a.m., when both rivers tended to remain partially frozen according to temperature measurements (remained at 0 °C).Seventy-two percent of maximum QAzufre values occurred between 8 a.m. and noon, when the temperature of the Azufre River passed the 0 °C threshold.These variations were coincident with the mixing proportion between the streams (QAzufre/(QAzufre + QCaracarani)) (72% of maximum QAzufre/(QAzufre + QCaracarani) values occurred between 8 a.m. and noon).Finally, maximum flow rates in the Caracarani River, as 77% of maximum QCaracarani values occurred between noon and 4 p.m. Variations in the time QCaracarani reached its maximum might be explained by freezing-melting cycles occurring near its origin, with delayed effects due to the distance to the confluence.Maximum values of QCaracarani/(QAzufre + QCaracarani), however, were widely spread (63% of these values occurred between noon and midnight).This was due to decreased flow rates towards the evening.

Diurnal Cycles in Particle Formation Downstream from the Confluence
Diurnal cycles in the formation of suspended solids (detected as changes in turbidity) were also detected downstream from the confluence.Characteristic peaks and drops in turbidity and pH were observed within specific time frames (Figure 6).Global turbidity and pH peaks were consistently observed early in the morning.For example, temporal measurements taken on 20 August at 320 m downstream from the confluence showed a characteristic global turbidity and pH peaks early in the morning (reaching up to 45 NTU at pH 5.2); these peaks were not as clear in the other time series as measurements taken later in the morning.A sudden turbidity and pH drop was consistently observed between 9 and 11 a.m. at all post-confluence sampling points.Another turbidity and pH

Diurnal Cycles in Particle Formation Downstream from the Confluence
Diurnal cycles in the formation of suspended solids (detected as changes in turbidity) were also detected downstream from the confluence.Characteristic peaks and drops in turbidity and pH were observed within specific time frames (Figure 6).Global turbidity and pH peaks were consistently observed early in the morning.For example, temporal measurements taken on 20 August at 320 m downstream from the confluence showed a characteristic global turbidity and pH peaks early in the morning (reaching up to 45 NTU at pH 5.2); these peaks were not as clear in the other time series as measurements taken later in the morning.A sudden turbidity and pH drop was consistently observed between 9 and 11 a.m. at all post-confluence sampling points.Another turbidity and pH peak occurred in the afternoon (between noon and 2 p.m.).The consistent peaks in these parameters indicate hourly water quality changes, consistent with the hourly hydrologic changes, as observed in Figure 7. Minimum values of pH and turbidity occurring within the same period maximum values of Q Azufre /(Q Azufre + Q Caracarani ) were observed.Local peaks in turbidity and pH occurring within the same period values of maximum Q Caracarani were observed.
of QAzufre/(QAzufre + QCaracarani) were observed.Local peaks in turbidity and pH occurring within the same period values of maximum QCaracarani were observed.A high correspondence between turbidity and pH was identified downstream from the confluence.This dependence was determined by pH controlled by incoming inflow ratios (QAzufre/(QCaracarani + QAzufre)), which caused the precipitation of mineral phases attributable to iron and aluminum oxides.Simultaneous measurements of pH and turbidity were significantly correlated, with correlation coefficients (rk) that ranged between 0.58 and 0.95 (see Figure 7).Additionally, turbidity versus pH showed a positive linear regression from pH 2 to approximately 5, as seen in Figure 8. Turbidity (suspended solids) was controlled by the precipitation of iron and aluminum phases, which are pH-controlled chemical reactions.Evidence of this is the presence of suspended solids, collected in the field, which showed high concentrations of iron and aluminum (Table 2).Furthermore, we developed a preliminary geochemical model that considered the formation of amorphous iron hydroxide (Fe(OH)3(a)) and amorphous aluminum hydroxide (Al(OH)3(a)) at different mixing ratios between the Azufre River and Caracarani River (see Supplementary Materials).This simplified model showed that Fe(OH)3 was feasible at pH > 3, while Al(OH)3 was feasible at pH > 5.It is also known that in natural systems Fe(OH)3(a) becomes saturated at pH > 3 A high correspondence between turbidity and pH was identified downstream from the confluence.This dependence was determined by pH controlled by incoming inflow ratios (Q Azufre /(Q Caracarani + Q Azufre )), which caused the precipitation of mineral phases attributable to iron and aluminum oxides.Simultaneous measurements of pH and turbidity were significantly correlated, with correlation coefficients (r k ) that ranged between 0.58 and 0.95 (see Figure 7).Additionally, turbidity versus pH showed a positive linear regression from pH 2 to approximately 5, as seen in Figure 8. Turbidity (suspended solids) was controlled by the precipitation of iron and aluminum phases, which are pH-controlled chemical reactions.Evidence of this is the presence of suspended solids, collected in the field, which showed high concentrations of iron and aluminum (Table 2).Furthermore, we developed a preliminary geochemical model that considered the formation of amorphous iron hydroxide (Fe(OH) 3 (a)) and amorphous aluminum hydroxide (Al(OH) 3 (a)) at different mixing ratios between the Azufre River and Caracarani River (see Supplementary Materials).This simplified model showed that Fe(OH) 3 was feasible at pH > 3, while Al(OH) 3 was feasible at pH > 5.It is also known that in natural systems Fe(OH) 3 (a) becomes saturated at pH > 3 and that aluminum hydroxide precipitates at pH > 4.5 [42].Although there are other iron and aluminum phases involved (such as hydrosulfates [43]), the analysis is consistent with how turbidity simultaneously increases with pH.These increases in turbidity downstream are not caused by incoming of suspended solids from upstream of the confluence.The turbidity measured at the Azufre River and Caracarani River fluctuated between 2 and 14 NTU, respectively.One hundred percent of turbidity values measured downstream were higher than 4 NTU (69% were higher than 14 NTU), indicating that turbidity increases were mainly due to pH-dependent chemical reactions.
Water 2016, 8, 74 10 of 27 aluminum phases involved (such as hydrosulfates [43]), the analysis is consistent with how turbidity simultaneously increases with pH.These increases in turbidity downstream are not caused by incoming of suspended solids from upstream of the confluence.The turbidity measured at the Azufre River and Caracarani River fluctuated between 2 and 14 NTU, respectively.One hundred percent of turbidity values measured downstream were higher than 4 NTU (69% were higher than 14 NTU), indicating that turbidity increases were mainly due to pH-dependent chemical reactions.The peaks of pH occurring before noon can be attributed to lower total flow rates, which triggered higher turbidity values (no particular patterns in QAzufre/(QAzufre + QCaracarani) were found).Additionally, lower flowrates (higher residence times [44,45]), caused increases in turbidity.The subsequent drop in pH observed between 9 a.m. and noon (Figure 7) was caused by peaks in QAzufre/(QAzufre + QCaracarani) (Figures 4 and 6), thus diminishing turbidity.Diminished suspended solids can be attributed to dissolved precipitates.Finally, the local peak in pH observed in the afternoon (Figure 7) was coincident with higher values of QCaracarani (Figures 5 and 6).Despite the fact that the alkalinity of the Caracarani River was moderate (20-232 mg CaCO3•L −1 , see Table 1), its discharge contributed to increase pH, thus an increase in turbidity.Overall, our results indicate that diurnal flow rate behavior controlled downstream pH.Downstream pH triggers the precipitation of iron and aluminum mineral phases at streams affected by acid drainage [42,43,46].In consequence, diurnal cycles in the formation of iron and aluminum phases, known to control the cycling of arsenic in the environment [23,42,43,[46][47][48], also constitute an important factor in the fate and transport of this toxic element.

The Fate and Transport of Metals Controlled by Freeze-Thaw Cycles
The effect diurnal changes in pH due to incoming flow ratios have on the fate of iron, aluminum, and arsenic can be visualized by comparing analyses of samples taken before and after the thawing of the Azufre River (Figure 9).A sample taken at 10:35 a.m.indicated a pH value of 5.2 and turbidity value of 45 NTU, dropping to pH 2.3 and turbidity of 13 NTU at 11:10 a.m.Total concentrations of iron, aluminum, and arsenic were considerably lower at 10:35 a.m.than those measured at 11:10 a.m.(5.4 vs. 22.2 mg•L −1 of total Fe, 14.7 vs. 58.5 mg•L −1 of total Al, and 0.18 vs. 0.63 mg•L −1 of As).Moreover, particulate fractions were much higher when higher pH values were measured, as shown in Figure 9.The abrupt increase in metal concentrations occurred within the same time range when the higher inflow of the Azufre River was observed.The sudden inflow of the Azufre River was not only higher in flow rate, but also carried higher concentrations of chemical species, explaining the higher concentration of iron, aluminum, and arsenic.This explains why higher total flux occurred within a matter of minutes in a pulse-like manner.Additionally, low pH promotes the dissolution of iron and aluminum mineral phases (iron and aluminum hydroxides), increasing concentration of dissolved metals, a condition that does not favor arsenic sorption [47,48].This comparison of samples, taken with only an hour from each other, showed that important changes in water quality are caused by diurnal variations in stream flow rate, particularly due to freeze and thaw cycles.The peaks of pH occurring before noon can be attributed to lower total flow rates, which triggered higher turbidity values (no particular patterns in Q Azufre /(Q Azufre + Q Caracarani ) were found).Additionally, lower flowrates (higher residence times [44,45]), caused increases in turbidity.The subsequent drop in pH observed between 9 a.m. and noon (Figure 7) was caused by peaks in Q Azufre /(Q Azufre + Q Caracarani ) (Figures 4 and 6), thus diminishing turbidity.Diminished suspended solids can be attributed to dissolved precipitates.Finally, the local peak in pH observed in the afternoon (Figure 7) was coincident with higher values of Q Caracarani (Figures 5 and 6).Despite the fact that the alkalinity of the Caracarani River was moderate (20-232 mg CaCO 3 ¨L´1 , see Table 1), its discharge contributed to increase pH, thus an increase in turbidity.Overall, our results indicate that diurnal flow rate behavior controlled downstream pH.Downstream pH triggers the precipitation of iron and aluminum mineral phases at streams affected by acid drainage [42,43,46].In consequence, diurnal cycles in the formation of iron and aluminum phases, known to control the cycling of arsenic in the environment [23,42,43,[46][47][48], also constitute an important factor in the fate and transport of this toxic element.

The Fate and Transport of Metals Controlled by Freeze-Thaw Cycles
The effect diurnal changes in pH due to incoming flow ratios have on the fate of iron, aluminum, and arsenic can be visualized by comparing analyses of samples taken before and after the thawing of the Azufre River (Figure 9).A sample taken at 10:35 a.m.indicated a pH value of 5.2 and turbidity value of 45 NTU, dropping to pH 2.3 and turbidity of 13 NTU at 11:10 a.m.Total concentrations of iron, aluminum, and arsenic were considerably lower at 10:35 a.m.than those measured at 11:10 a.m.(5.4 vs. 22.2 mg¨L ´1 of total Fe, 14.7 vs. 58.5 mg¨L ´1 of total Al, and 0.18 vs. 0.63 mg¨L ´1 of As).Moreover, particulate fractions were much higher when higher pH values were measured, as shown in Figure 9.The abrupt increase in metal concentrations occurred within the same time range when the higher inflow of the Azufre River was observed.The sudden inflow of the Azufre River was not only higher in flow rate, but also carried higher concentrations of chemical species, explaining the higher concentration of iron, aluminum, and arsenic.This explains why higher total flux occurred within a matter of minutes in a pulse-like manner.Additionally, low pH promotes the dissolution of iron and aluminum mineral phases (iron and aluminum hydroxides), increasing concentration of dissolved metals, a condition that does not favor arsenic sorption [47,48].This comparison of samples, taken with only an hour from each other, showed that important changes in water quality are caused by diurnal variations in stream flow rate, particularly due to freeze and thaw cycles.The mass rate of elements depends on both the concentration of total suspended solids (TSS) and the concentration of metals.This was determined by estimating the rate of iron, aluminum, and arsenic in g h −1 , as follows: where TSS: concentration of total suspended solids, Element: concentration of the element (Fe, Al or As); Q: stream flowrate of the Caracarani River downstream from the confluence.
Indeed, arsenic reached its maximum concentration in the solid phase at pH 3.8 (Table 2); however, the observed concentration of total suspended solids was only 9 mg/L.The highest concentration of suspended solids (81 mg/L) occurred at pH 5.9.Results showed that the mass rate of arsenic, iron, and aluminum in the solid phase, at a constant flowrate of 200 L•s −1 , increased with a higher concentration of suspended solids, as shown in Figure 10.This means that the higher the concentration of total suspended solids, the higher the mass rate of elements in solid phase will be.Values of particulate and dissolved fractions of arsenic downstream from the Azufre River-Caracarani River confluence were presented in Guerra et al. [39].Particulate arsenic concentration ranged between 75% and 95% within a wide range of pH values (pH 3 to 7); there was no increase or decrease of particulate arsenic concentration with increasing pH.This indicates that particulate arsenic transport, in g•h −1 , was controlled by the concentration of total suspended solids, which is a result of the pH of the stream.Immobilization of arsenic is promoted through the precipitation of mineral phases, reflected in the increased concentration of suspended solids.This is favorable in terms of site remediation: when iron and aluminum are part of the solid phase, these have the potential to settle on the stream bed.As arsenic sorption occurs, this metalloid will also be immobilized.Additionally, arsenic and iron were highly correlated (Figure 10), reaffirming the role of iron phases downstream from the confluence.A controlling factor in the fate of arsenic is the sorption or co-precipitation onto freshly precipitated iron and aluminum mineral phases (hydroxides and hydrosulfates) [42,45].In the case of the Azufre River-Caracarani River confluence, diurnal changes in flowrate (in consequence, in the proportion QAzufre/(QAzufre + QCaracarani) control downstream pH.pH triggers reactions of precipitation of iron and aluminum phases.In consequence, particulate arsenic (sorbed or co-precipitated onto mineral phases) is also prone to experience diurnal cycles.During the hours when QAzufre is at peak values, iron hydroxides are dissolved, and, as a consequence, arsenic is again released into the aqueous phase.For this reason, the estimation of the mass rate of The mass rate of elements depends on both the concentration of total suspended solids (TSS) and the concentration of metals.This was determined by estimating the rate of iron, aluminum, and arsenic in g h ´1, as follows: Mass rate of elements, g¨h ´1 " rTSS, mg¨L ´1s ˆrMetal, mg¨kg ´1s Q, where TSS: concentration of total suspended solids, Element: concentration of the element (Fe, Al or As); Q: stream flowrate of the Caracarani River downstream from the confluence.Indeed, arsenic reached its maximum concentration in the solid phase at pH 3.8 (Table 2); however, the observed concentration of total suspended solids was only 9 mg/L.The highest concentration of suspended solids (81 mg/L) occurred at pH 5.9.Results showed that the mass rate of arsenic, iron, and aluminum in the solid phase, at a constant flowrate of 200 L¨s ´1, increased with a higher concentration of suspended solids, as shown in Figure 10.This means that the higher the concentration of total suspended solids, the higher the mass rate of elements in solid phase will be.Values of particulate and dissolved fractions of arsenic downstream from the Azufre River-Caracarani River confluence were presented in Guerra et al. [39].Particulate arsenic concentration ranged between 75% and 95% within a wide range of pH values (pH 3 to 7); there was no increase or decrease of particulate arsenic concentration with increasing pH.This indicates that particulate arsenic transport, in g¨h ´1, was controlled by the concentration of total suspended solids, which is a result of the pH of the stream.Immobilization of arsenic is promoted through the precipitation of mineral phases, reflected in the increased concentration of suspended solids.This is favorable in terms of site remediation: when iron and aluminum are part of the solid phase, these have the potential to settle on the stream bed.As arsenic sorption occurs, this metalloid will also be immobilized.Additionally, arsenic and iron were highly correlated (Figure 10), reaffirming the role of iron phases downstream from the confluence.A controlling factor in the fate of arsenic is the sorption or co-precipitation onto freshly precipitated iron and aluminum mineral phases (hydroxides and hydrosulfates) [42,45].In the case of the Azufre River-Caracarani River confluence, diurnal changes in flowrate (in consequence, in the proportion Q Azufre /(Q Azufre + Q Caracarani ) control downstream pH.pH triggers reactions of precipitation of iron and aluminum phases.In consequence, particulate arsenic (sorbed or co-precipitated onto mineral phases) is also prone to experience diurnal cycles.During the hours when Q Azufre is at peak values, iron hydroxides are dissolved, and, as a consequence, arsenic is again released into the aqueous phase.For this reason, the estimation of the mass rate of elements in solid phase (g¨h ´1) is relevant, in order to determine at which moments of the day arsenic transport is promoted.
An integration of the hydrologic and chemical processes that were described can be presented through a conceptual model, as shown in Figure 11.When Q Azufre /(Q Azufre + Q Caracarani ) is high, pH is low and, as a consequence, so is the concentration of total suspended solids (surrogate for iron and aluminum mineral phases).At a low pH, total arsenic is high (Figure 9) and the low availability of solid phases indicates that the particulate arsenic mass rate is low (thus, dissolved arsenic will be high).When Q Azufre /(Q Azufre + Q Caracarani ) is lower, pH is higher and, as a consequence, the formation of mineral phases is promoted.At higher pH values, total arsenic is lower (Figure 9), and the higher availability of mineral phases promotes arsenic sorption.This indicates that the particulate arsenic mass rate is higher at these conditions.
This study shows how cycles in flow rate, due to daily freeze-thaw processes, can also constitute a controlling factor in the fate and transport of toxic elements downstream a confluence affected by acid drainage.It is also necessary to consider these freeze-thaw processes and their effect in arsenic transport at distances further downstream the confluence.Modeling of arsenic-rich sediments at a future reservoir located downstream, at the Lluta Valley, showed the potential desorption of arsenic from the sediments [49]; the cyclic transport of dissolved or particulate arsenic may also have effects on the chemical dynamics of reservoirs near areas affected by acid drainage.
Water 2016, 8, 74 13 of 27 elements in solid phase (g•h −1 ) is relevant, in order to determine at which moments of the day arsenic transport is promoted.An integration of the hydrologic and chemical processes that were described can be presented through a conceptual model, as shown in Figure 11.When QAzufre/(QAzufre + QCaracarani) is high, pH is low and, as a consequence, so is the concentration of total suspended solids (surrogate for iron and aluminum mineral phases).At a low pH, total arsenic is high (Figure 9) and the low availability of solid phases indicates that the particulate arsenic mass rate is low (thus, dissolved arsenic will be high).When QAzufre/(QAzufre + QCaracarani) is lower, pH is higher and, as a consequence, the formation of mineral phases is promoted.At higher pH values, total arsenic is lower (Figure 9), and the higher

Conclusions
The objective of this paper was to show how daily variations in stream flow rates due to freeze and thaw cycles control metals' transport downstream in streams affected by acid drainage.We characterized the hydrologic and geochemical behavior of the Azufre River (pH < 2) and Caracarani River (pH ~8.6) confluence (northern Chile) by measuring stream flow rates, temperature, electric conductivity, pH, and turbidity and analyzing the concentration of metals in the aqueous phase.We showed how cyclic hydrological variations control the fate of contaminants in a watershed affected by acid drainage.Coincidental diurnal patterns were observed between flow rates, turbidity, and pH.Important drops in pH occurred when peaks in QAzufre (QAzufre/(QAzufre + QCaracarani)) occurred, with consequential drops in turbidity.Similarly, peaks in pH occurred within the same time range when the maximum QCaracarani occurred, with consequential drops in turbidity.Changes in the flow rate of the Azufre River were caused by freeze and thaw cycles.The conclusions of this work are as follows: 1. Due to the below-zero temperatures reached at night, QAzufre was controlled by daily freeze-thaw processes.The daily pulse in flow rate in the acid drainage results in a pulse in the flux of dissolved metals downstream and in the dissolution of suspended solids.Although this process occurred for only a few minutes every day, it was sufficient to mobilize metals and degrade water quality downstream due to, for example, higher flux of dissolved arsenic.2. Diurnal behavior of water quality downstream from the confluence was controlled by diurnal changes in the upstream flow regime.Clear patterns in pH occurred concurrently with daily peaks in QAzufre and QCaracarani.As a consequence, turbidity (surrogate for suspended solids formed downstream as a consequence of the precipitation of mineral phases) also experience daily peaks.In this study, hourly variations at a confluence affected by acid drainage due to daily freeze and thaw cycles were identified, showing that drastic changes in water quality are not limited to seasonal factors (i.e., snow melts, increase in rainfall) but could occur within intraday time scales.3. The formation of mineral phases downstream from the confluence was enhanced at distinct times of the day (dawn and mid-afternoon).Since the formation of suspended solids controlled the fate and transport of metals, attenuation of dissolved contaminants would improve within these time lapses, as particles have the potential to settle onto the streambed.
Hydrologic diurnal cycles constitute an important factor to consider when exploring sites affected by acid drainage.Diurnal cycles in flow rate, particularly in streams affected by acid

Conclusions
The objective of this paper was to show how daily variations in stream flow rates due to freeze and thaw cycles control metals' transport downstream in streams affected by acid drainage.We characterized the hydrologic and geochemical behavior of the Azufre River (pH < 2) and Caracarani River (pH ~8.6) confluence (northern Chile) by measuring stream flow rates, temperature, electric conductivity, pH, and turbidity and analyzing the concentration of metals in the aqueous phase.We showed how cyclic hydrological variations control the fate of contaminants in a watershed affected by acid drainage.Coincidental diurnal patterns were observed between flow rates, turbidity, and pH.Important drops in pH occurred when peaks in Q Azufre (Q Azufre /(Q Azufre + Q Caracarani )) occurred, with consequential drops in turbidity.Similarly, peaks in pH occurred within the same time range when the maximum Q Caracarani occurred, with consequential drops in turbidity.Changes in the flow rate of the Azufre River were caused by freeze and thaw cycles.The conclusions of this work are as follows: 1.
Due to the below-zero temperatures reached at night, Q Azufre was controlled by daily freeze-thaw processes.The daily pulse in flow rate in the acid drainage results in a pulse in the flux of dissolved metals downstream and in the dissolution of suspended solids.Although this process occurred for only a few minutes every day, it was sufficient to mobilize metals and degrade water quality downstream due to, for example, higher flux of dissolved arsenic.

2.
Diurnal behavior of water quality downstream from the confluence was controlled by diurnal changes in the upstream flow regime.Clear patterns in pH occurred concurrently with daily peaks in Q Azufre and Q Caracarani .As a consequence, turbidity (surrogate for suspended solids formed downstream as a consequence of the precipitation of mineral phases) also experience daily peaks.In this study, hourly variations at a confluence affected by acid drainage due to daily freeze and thaw cycles were identified, showing that drastic changes in water quality are not limited to seasonal factors (i.e., snow melts, increase in rainfall) but could occur within intraday time scales.

3.
The formation of mineral phases downstream from the confluence was enhanced at distinct times of the day (dawn and mid-afternoon).Since the formation of suspended solids controlled the fate and transport of metals, attenuation of dissolved contaminants would improve within these time lapses, as particles have the potential to settle onto the streambed.
Hydrologic diurnal cycles constitute an important factor to consider when exploring sites affected by acid drainage.Diurnal cycles in flow rate, particularly in streams affected by acid drainage, caused by freeze and thaw cycles are prone to occur at other sites where below-zero temperatures occur, such as mine sites located at high altitudes.The processes observed at this site impacted on the downstream water quality in a cyclical manner through daily "pulse-type" discharges.These diurnal changes may cause variability in the analysis of freshwater samples, critical in planning sampling and monitoring campaigns in these types of systems.The knowledge of when and where attenuation/mobilization of contaminants occurs is important to prevent toxic elements from reaching points of use for human consumption and agriculture.Appendix C: Preliminary geochemical model.
To build a preliminary geochemical model, the main reactions considered were the precipitation of Fe(OH) 3(am) and Al(OH) 3(a) .A chemical equilibrium model for downstream mixing was developed to simulate the formation of these as freshly precipitated phases as a function of the mixing ratio (V Azufre /V Caracarani ).The pH was also simulated.The water quality considered for simulations is summarized in Table C1.The calculations were performed using the PHREEQC code (Parkhurst and Appelo, 1999) and wateq4f.datthermodynamic database (Dzombak and Morel, 1990).Figure C1 describes the geochemical model.Other phases are relevant to systems affected by acid drainage (such as Schwertmannite and Basaluminite) and must be considered further to improve the current model.

Figure 1 .
Figure 1.Site of study.The Azufre River joins the Caracarani River in the higher Lluta River basin, northern Chile.The Lluta River is born when the Caracarani River joins the Colpitas River.Points "a" and "b" indicate where stream flow, temperature, EC, turbidity, and pH were measured.The arrow shows the stream segment along which turbidity, EC, and pH were measured downstream of the junction point.Point "c" indicates the location of the meteorological station.

Figure 2 .
Figure 2. Temperature data (a) and monthly accumulated precipitation (b) recorded within 8.5 km of the site of study (point "c" in Figure 1).In (a), the continuous horizontal line depicts monthly mean values while the vertical lines indicate minimum and maximum values within the month.

Figure 1 .
Figure 1.Site of study.The Azufre River joins the Caracarani River in the higher Lluta River basin, northern Chile.The Lluta River is born when the Caracarani River joins the Colpitas River.Points "a" and "b" indicate where stream flow, temperature, EC, turbidity, and pH were measured.The arrow shows the stream segment along which turbidity, EC, and pH were measured downstream of the junction point.Point "c" indicates the location of the meteorological station.

Figure 1 .
Figure 1.Site of study.The Azufre River joins the Caracarani River in the higher Lluta River basin, northern Chile.The Lluta River is born when the Caracarani River joins the Colpitas River.Points "a" and "b" indicate where stream flow, temperature, EC, turbidity, and pH were measured.The arrow shows the stream segment along which turbidity, EC, and pH were measured downstream of the junction point.Point "c" indicates the location of the meteorological station.

Figure 2 .
Figure 2. Temperature data (a) and monthly accumulated precipitation (b) recorded within 8.5 km of the site of study (point "c" in Figure 1).In (a), the continuous horizontal line depicts monthly mean values while the vertical lines indicate minimum and maximum values within the month.

Figure 2 .
Figure 2. Temperature data (a) and monthly accumulated precipitation (b) recorded within 8.5 km of the site of study (point "c" in Figure 1).In (a), the continuous horizontal line depicts monthly mean values while the vertical lines indicate minimum and maximum values within the month.

Figure 3 .
Figure 3. Mean monthly flow rate data for the Azufre River (a) and Caracarani River (b), measured during 2010-2011.The continuous horizontal line depicts mean values while the vertical lines indicate minimum and maximum daily flow rates.

Figure 3 .
Figure 3. Mean monthly flow rate data for the Azufre River (a) and Caracarani River (b), measured during 2010-2011.The continuous horizontal line depicts mean values while the vertical lines indicate minimum and maximum daily flow rates.

Figure 4 .
Figure 4. Hourly stream flow rates, temperatures and EC registered upstream of the confluence in August 2011.(a) QAzufre; (b) TAzufre; and (c) ECAzufre data.Autocorrelation analysis showed a 24 h periodicity for QCaracarani, QAzufre, TAzufre, and ECAzufre.The x-axis depicts the day of the month.

Figure 4 .
Figure 4. Hourly stream flow rates, temperatures and EC registered upstream of the confluence in August 2011.(a) Q Azufre ; (b) T Azufre ; and (c) EC Azufre data.Autocorrelation analysis showed a 24 h periodicity for Q Caracarani , Q Azufre , T Azufre , and EC Azufre .The x-axis depicts the day of the month.

Figure 5 .
Figure 5. Hourly data of the Caracarani River from August 2011.QCaracarani (a); TCaracarani (b); and ECCaracarani (c).The axis depicts the days of the month.

Figure 5 .
Figure 5. Hourly data of the Caracarani River from August 2011.Q Caracarani (a); T Caracarani (b); and EC Caracarani (c).The axis depicts the days of the month.

Figure 6 .
Figure 6.Turbidity and pH time series.Measurements were taken at 90 (a); 180 (b); 320 (c); and 1000 m (d) downstream from the confluence.A global peak (before 11 a.m.), a characteristic sink in the morning (between 11 a.m. and 12 p.m.), and a local peak towards the afternoon (between 1 and 2 p.m.) were observed.Turbidity and pH were highly correlated (correlation coefficients rk values are shown in each panel).

Figure 6 .
Figure 6.Turbidity and pH time series.Measurements were taken at 90 (a); 180 (b); 320 (c); and 1000 m (d) downstream from the confluence.A global peak (before 11 a.m.), a characteristic sink in the morning (between 11 a.m. and 12 p.m.), and a local peak towards the afternoon (between 1 and 2 p.m.) were observed.Turbidity and pH were highly correlated (correlation coefficients r k values are shown in each panel).

Figure 7 .Figure 7 .
Figure 7. Representative daily cycles of the streams for 15-16 August 2011: Azufre River data (a); and Caracarani River data (b).The peaks in QAzufre occurred when the TAzufre exceeded 0 °C with an approximate lag of 7 h.QAzufre peaks occurred at the same time peaks in mixing ratio (QAzufre/(QAzufre + QCaracarani)) While the Azufre River remained partially frozen, EC tended to reach higher values with an approximate lag of 1 h.QCaracarani reached peak values towards the midafternoon.

Figure 8 .
Figure 8. Turbidity versus pH shows a positive linear regression from pH 2 to 5.5 (circles), and decreased from pH 5.5 and upward (squares).The increase in turbidity was attributable to iron and aluminum phases that form within pH 3-5.

Figure 8 .
Figure 8. Turbidity versus pH shows a positive linear regression from pH 2 to 5.5 (circles), and decreased from pH 5.5 and upward (squares).The increase in turbidity was attributable to iron and aluminum phases that form within pH 3-5.

Figure 10 .
Figure 10.Arsenic, iron and aluminum in the solid phase.Relation between arsenic concentration in the solid phase and iron and aluminum concentration in the solid phase (a).Mass rate of metals versus total suspended solids (b).

Figure 10 .
Figure 10.Arsenic, iron and aluminum in the solid phase.Relation between arsenic concentration in the solid phase and iron and aluminum concentration in the solid phase (a).Mass rate of metals versus total suspended solids (b).

Figure 11 .
Figure 11.Conceptual model of the hydrologic and chemical processes that occur downstream from the Azufre River-Caracarani River confluence.

Figure 11 .
Figure 11.Conceptual model of the hydrologic and chemical processes that occur downstream from the Azufre River-Caracarani River confluence.

Figure A2 .
Figure A2.Hourly hydrological data from February 2011.(a) TAzufre; (b) QAzufre.The axis depicts the days of the month.QAzufre was extrapolated for values greater than 250 L/s.

Figure A3 .
Figure A3.Hourly hydrological data from March 2011.(a) TAzufre; (b) QAzufre.The axis depicts the days of the month.

Figure A2 . 27 Figure A2 .
Figure A2.Hourly hydrological data from February 2011.(a) T Azu f re ; (b) Q Azufre .The axis depicts the days of the month.Q Azufre was extrapolated for values greater than 250 L/s.

Figure A3 .
Figure A3.Hourly hydrological data from March 2011.(a) TAzufre; (b) QAzufre.The axis depicts the days of the month.Figure A3.Hourly hydrological data from March 2011.(a) T Azu f re ; (b) Q Azufre .The axis depicts the days of the month.

Figure A3 .
Figure A3.Hourly hydrological data from March 2011.(a) TAzufre; (b) QAzufre.The axis depicts the days of the month.Figure A3.Hourly hydrological data from March 2011.(a) T Azu f re ; (b) Q Azufre .The axis depicts the days of the month.

Figure A4 .
Figure A4.Hourly hydrological data from April 2011 (a) T Azu f re ; (b) Q Azu f re and Q Caracarani ; (c) Mixing ratio R; (d) Azufre River electric conductivity (EC Azu f re ).The axis depicts the days of the month.Q Caracarani was extrapolated for values > 350 L/s.

Figure A5 .
Figure A5.Hourly hydrological data from May 2011.(a) TAzufre; (b) QAzufre and QCaracarani; (c) Mixing ratio R; (d) ECAzufre.The axis depicts the days of the month.QCaracarani was extrapolated for values > 350 L/s.Figure A5.Hourly hydrological data from May 2011.(a) T Azu f re ; (b) Q Azu f re and Q Caracarani ; (c) Mixing ratio R; (d) EC Azu f re .The axis depicts the days of the month.Q Caracarani was extrapolated for values > 350 L/s.

Figure A5 .
Figure A5.Hourly hydrological data from May 2011.(a) TAzufre; (b) QAzufre and QCaracarani; (c) Mixing ratio R; (d) ECAzufre.The axis depicts the days of the month.QCaracarani was extrapolated for values > 350 L/s.Figure A5.Hourly hydrological data from May 2011.(a) T Azu f re ; (b) Q Azu f re and Q Caracarani ; (c) Mixing ratio R; (d) EC Azu f re .The axis depicts the days of the month.Q Caracarani was extrapolated for values > 350 L/s.

Figure A6 .
Figure A6.Hourly hydrological data from June 2011.(a) TAzufre; (b) QAzufre; and QCaracarani; (c) Mixing ratio R; (d) ECAzufre.The axis depicts the days of the month.QCaracarani was extrapolated for values > 350 L/s.Figure A6.Hourly hydrological data from June 2011.(a) T Azu f re ; (b) Q Azufre ; and Q Caracarani ; (c) Mixing ratio R; (d) EC Azu f re .The axis depicts the days of the month.Q Caracarani was extrapolated for values > 350 L/s.

Figure A6 .Figure A7 .
Figure A6.Hourly hydrological data from June 2011.(a) TAzufre; (b) QAzufre; and QCaracarani; (c) Mixing ratio R; (d) ECAzufre.The axis depicts the days of the month.QCaracarani was extrapolated for values > 350 L/s.Figure A6.Hourly hydrological data from June 2011.(a) T Azu f re ; (b) Q Azufre ; and Q Caracarani ; (c) Mixing ratio R; (d) EC Azu f re .The axis depicts the days of the month.Q Caracarani was extrapolated for values > 350 L/s.

Figure A7 .Figure A8 .
Figure A7.Hourly hydrological data from July 2011.(a) T Azu f re ; (b) Q Azufre and Q Caracarani ; (c) Mixing ratio R; (d) EC Azu f re .The axis depicts the days of the month.Q Caracarani was extrapolated for values > 350 L/s.

Figure A8 .Figure A9 .
Figure A8.Hourly hydrological data from September 2011.(a) TAzufre; (b) QAzufre and QCaracarani; (c) Mixing ratio R; (d) ECAzufre.The axis depicts the days of the month.Figure A8.Hourly hydrological data from September 2011.(a) T Azu f re ; (b) Q Azufre and Q Caracarani ; (c) Mixing ratio R; (d) EC Azu f re .The axis depicts the days of the month.

Figure A9 .
Figure A9.Hourly hydrological data from October 2011.(a) TAzufre; (b) QAzufre and QCaracarani; (c) Mixing ratio R; (d) ECAzufre.The axis depicts the days of the month.Figure A9.Hourly hydrological data from October 2011.(a) T Azu f re ; (b) Q Azufre and Q Caracarani ; (c) Mixing ratio R; (d) EC Azu f re .The axis depicts the days of the month.

Figure A10 .
Figure A10.Hourly data from November 2011.(a) TAzufre; (b) QAzufre and QCaracarani; (c) Mixing ratio R; (d) ECAzufre.The axis depicts the days of the month.QCaracarani was extrapolated for values > 350 L/s.Figure A10.Hourly data from November 2011.(a) T Azu f re ; (b) Q Azu f re and Q Caracarani ; (c) Mixing ratio R; (d) EC Azu f re .The axis depicts the days of the month.Q Caracarani was extrapolated for values > 350 L/s.

Figure A10 .
Figure A10.Hourly data from November 2011.(a) TAzufre; (b) QAzufre and QCaracarani; (c) Mixing ratio R; (d) ECAzufre.The axis depicts the days of the month.QCaracarani was extrapolated for values > 350 L/s.Figure A10.Hourly data from November 2011.(a) T Azu f re ; (b) Q Azu f re and Q Caracarani ; (c) Mixing ratio R; (d) EC Azu f re .The axis depicts the days of the month.Q Caracarani was extrapolated for values > 350 L/s.

Figure B1 .
Figure B1.Daily turbidity and pH time series measured upstream of the confluence at the Azufre River (a) and Caracarani River (b). pH remained relatively constant in both cases, whereas turbidity reached a peak at around 12 p.m. and 11 a.m., respectively.

Figure C1 .
Figure C1.Conceptual model of the precipitation of Fe(OH)3(a) and Al(OH)3(a) as a function of the mixing ratio.pH was estimated as a function of coupled reactions that include precipitation and other acid-base reactions (i.e., carbonate system).

Table 1 .
Representative water quality data for the Azufre River and Caracarani River (sampling period: 2010-2012).

Table 1 .
Representative water quality data for the Azufre River and Caracarani River (sampling period: 2010-2012).

Table 2 .
Concentration of total suspended solids (TSS) and corresponding concentrations of arsenic (As), iron (Fe), and aluminum (Al).Samples were taken randomly, downstream from the confluence, through several field campaigns.

Table C1 .
Representative water quality of the Azufre River and Caracarani River used to simulate the pH concentration of Fe(OH)3(am) + Al(OH)3(am) using the PHREEQC code.

Table C1 .
Representative water quality of the Azufre River and Caracarani River used to simulate the pH concentration of Fe(OH) 3(am) + Al(OH) 3(am) using the PHREEQC code.