Long-term monitoring of Mn and Fe removal in biofilters from a converted plant Open Access

Removal of iron (Fe) and manganese (Mn) from groundwater using biological filtration is considered suitable and cost-effective (Bruins 2016). This method has been commonly used in Europe, North America (Ramsay et al. 2018), and Argentina (Pacini et al. 2014). However, little is known about the biological removal of contaminants from water resources in developing countries (Abu Hasan et al. 2020). Recent studies have just confirmed the presence of autochthonous culturable manganese-oxidizing bacteria (MOB) and the feasibility of using biofiltration for Mn removal from groundwater sources under tropical conditions (Calderón et al. 2020; Araya-Obando et al. 2021). Biological filtration has some advantages over conventional treatment processes. Biofilters have minimal operation and maintenance requirements, do not require chemicals, and may reduce disinfection byproducts (DBP), for example (Evans et al. 2021). Thus, full-scale biofiltration conversion from conventional filtration systems has begun to receive increased attention (Bassett et al. 2018).

The first descriptions of the conversion of about 20 conventional treatment plants in France to biological processes date from the early 1990s (Mouchet 1992). The strategies used then involved media replacement, pH optimization, changes in the aeration conditions, and eliminating prechlorination or any reagents added at the head of the treatment line, representing an operational cost reduction of about 50–80%. Another study describing the conversion of an existing plant for Fe and Mn removal from physicochemical to biological processes was reported by Pacini et al. (2014). That study was carried out in the water treatment plant (WTP) at Las Toscas, Argentina, in October 2011 and included the following modifications: aeration of raw water using a perforated tray aerator with plastic rings, conversion of an existing circular settler in an up-flow roughing filter (URF) and eliminating dosing chemical reagents. Similar to Mouchet (1992) study, operating costs were reduced by 60%. Furthermore, conversion from conventional filtration to biofiltration and removing prechlorination is increasingly being used in surface water treatment plants throughout the U.S. to enhance the removal of organic and inorganic constituents (Brown 2020).

Short-term water quality deterioration (e.g., Mn release, turbidity breakthrough) and operational/hydraulic challenges can be present during the start-up of the converted plants (Brown 2020). Moreover, it is well-known that the long ripening time of virgin filter media to achieve very effective manganese removal is a major concern (Bruins 2016; Breda et al. 2019b). Therefore, it is critical to have proper planning and evaluation of the biofilter conversion strategies during the early stages of operation to avoid unintended consequences affecting the overall plant (Brown 2020). Recently, nutrients (phosphorus)/pH and substrate augmentation strategies have been demonstrated to be effective means to improve Mn control during the conversion to biofiltration (Lauderdale et al. 2016).

Once an acclimated biofilter has reached steady-state contaminant removal, many factors can impact its performance, such as variations in influent water quality, water temperatures, operational parameters, and biofouling (Brown 2020). Thus, long-term monitoring of converted biofilters is essential to evaluate their performance in case of variations in raw water quality and operational and hydraulic changes over time. Additionally, the monitoring can provide important evidence about seasonal impacts, for example, if cold groundwater temperatures have a negative impact on the Fe and Mn removal of full-scale converted biofilters. In general, a noticeable gap exists in the literature about the operational performance of converted water treatment plants for Fe and Mn removal over a long-term period.

This paper is a continuation of previous studies conducted in the Las Toscas WTP in Argentina. The operation of the converted plant began on October 31, 2011. As previously mentioned, Pacini et al. (2014) first studied the transformation from a physicochemical to a biofiltration process. That study included pilot-scale experiments, details about the design and construction of modifications, and monitoring of the start-up period of biofilters during the first 20 months of the converted plant operation. The following work performed by Piazza et al. (2019) confirmed the presence of several culture MOBs with Mn²⁺ oxidation and biofilm formation capacities in matured filter media collected from the biofilters of Las Toscas WTP after seven years of operation. The aim of this study was to demonstrate the long-term feasibility of biological filters for Fe and Mn removal in a converted physicochemical plant. For this purpose, monitoring data from November 2011 until April 2021 from the converted physicochemical plant Las Toscas located in Santa Fe, Argentina, was evaluated. Statistical comparisons were also included to determine seasonal variations in Fe and Mn concentrations at the influents and effluents of the biofiltration units.

Las Toscas WTP is located in Santa Fe, Argentina. The source water is groundwater from 10 wells. The treatment line (Figure 1) starts with a perforated tray aerator with plastic rings that was installed during the conversion in 2011, followed by the raw water tank of 250 m³. Subsequently, the aerated water is pumped to the biological filtration systems composed of an up-flow gravel roughing filter (URF) (old circular settler) and two pressure rapid sand filters (RSF) configured in parallel. The URF is composed of gravel (6–12 mm size) with a filter bed height of 1 m, whereas the RSF are composed of a 1 m layer of sand (1.10–2.20 mm size). Finally, the treated water is stored in a reservoir for its subsequent chlorination and distribution. The operational parameters of these filtration units are shown in Table 1. Note that total water production is typically reduced from 150 to 100 m³/h in winter. Subsequently, filtration rates and empty bed contact times (EBCT) are different during winter compared with the other seasons.

A summary of the upgrades, with three main events, in Las Toscas WTP during the period of study (November 2011 until April 2021) is shown in Table 2. The reduction of the aerated water reservoir (Event 1) was made to increase the volume of treated water. Meanwhile, the reasons why Events 2 and 3 were implemented will be discussed in the next chapter.

As shown in Figure 1, the sampling points considered in this study comprised the water samples collected monthly from the URF influent (SP1-URF_i), the URF effluent (SP2-URF_e), and the RSF effluent (SP3-RSF_e). In addition, the main operational records were reviewed by interviewing key employees of the WTP. Data included turbidity, and total Fe and Mn concentrations collected from 222 water analysis reports from November 2011 to April 2021. Mn and Fe were determined using a DR 2700 spectrophotometer (Hach, Loveland, CO, USA) following the phenanthroline and the 1-(2-Pyridylazo)-2-Naphthol PAN methods, respectively. Turbidity was measured using a 2100P portable turbidimeter (Hach, USA). The manufacturer's instructions were followed in all three tests.

Data analysis was performed in R (R Core Team 2021). Summary descriptive statistics for all variables were done using stat.desc from the pastecs library. Time series and boxplot of Fe and Mn concentrations were done using the R graphics package. Cumulative frequency distribution and density plots were made using the DescTools package to know the summary of data frequency below maximum contaminant levels (MCL) stipulated in local regulations (EnReSS 1994) and the distribution of the numeric variables, respectively. In addition, R was used to determine statistically significant differences between the seasonal variation of Fe and Mn concentrations in the URG and RSF. A statistical p-value less than 0.05 was considered to be significant. For this purpose, the Fligner–Killeen test (Stats Package) and Levene's test (DescTools package) were first performed to check the homogeneity of variance across groups (i.e., seasonal Fe and Mn concentrations). In addition, a test of normality of these data was done with the Shapiro–Wilk (n < 50) (Stats Package) and Lilliefors (Kolmogorov–Smirnov) normality tests (n > 50) (Nortest package). Considering the non-normality test results and that data were generally highly skewed to the right, a nonparametric approach was adopted using the Kruskal–Wallis rank sum test (Stats Package). Subsequently, median and interquartile range (IQR) were used for the description of non-normally distributed data (Habibzadeh 2017). Finally, as the groups had unequal numbers of observations, Dunn's test was used for post hoc pairwise multiple comparisons (DescTools package).

Water temperature in SP1-URF_i (which represents the aerated water pumped from the reservoir), ranged from around 15 °C (winter) up to 30 °C (summer), and the pH varied between 6.6 and 7.3 and remained withing the same range in the filtration units. Pacini et al. (2014) reported that the dissolved oxygen concentration in this sampling point fluctuated in a range from 2.6 up to 5.3 mg/L depending on the water level in the reservoir. The authors also reported that the redox potential varied between 230 and 260 mV. Table 3 shows the Fe, Mn, and turbidity results for the 10 years of the operation monitoring period. SP1-URF_i showed a lower variation of Fe and Mn concentrations. Moreover, the median Fe concentration is higher than that of Mn. It can be also seen that a substantial reduction of Fe concentrations occurred between SP1-URF_i and SP2-URF_e, indicating that Fe was mainly removed in the URF. Clearly, the URF played a role as pretreatment reducing Fe and Mn concentrations in SP2-URF_e(which represents the RSF influent), thus facilitating the performance of the RSF. In addition, results in SP3-RSF_e did not exceed the maximum contaminant levels (MCL) stipulated in the local regulation (EnReSS 1994). Thus, the converted biological filtration system of Las Toscas WTP effectively removed these parameters during the period of study. Specifically, around 98% and 95% of all Fe and Mn concentrations, respectively, were below the MCL. Similar to the metals, turbidity was partially reduced in the URF and subsequently in the RSF. Moreover, turbidity in the treated water never exceeded the MCL of 2.0 NTU and just 6% of the data did not meet the recommended value of 0.5 NTU called for in the local regulation (EnReSS 1994).

Figure 2 shows a comparison of how efficiently the entire treatment process was working for the two metals. To identify the contribution of each filter, removal efficiencies of Fe and Mn in the URF and RSF were estimated considering the influent as the values at the head of the treatment line (SP1-URF_i). As shown in Figure 2, around 68% of the Fe was removed in the URF, whereas the remaining 29% was removed in the RSF. In comparison, Mn removal efficiencies were quite similar through the filtration units. About 46% and 47% of Mn were removed in the URF and RSF, respectively. Moreover, the combination of URF with RSF produced total Fe and Mn removal efficiencies of nearly 96% and 88%, respectively. Similar removal percentages have been reported in different double filtration systems. For example, Pacini et al. (2005) reported in a pilot system that combined URF with RSF total removal efficiencies of 94% for Fe and 92% for Mn. In the same study but combining URF with slow sand filtration (SSF), the total removal efficiencies were around 95% for Fe and 88% for Mn. Sánchez & Burbano (2006) reported in a full-scale URF combined with SSF, total removal efficiencies of 92% and 89% for Fe and Mn, respectively. Breda et al. (2016) reported two full-scale rapid filters in series removed all Fe and 90% of the Mn in the first filter after the start-up period. Different mechanisms (physicochemical and biological) may have contributed to Fe and Mn removal in mature biofilters. Piazza et al. (2019) found that the URF and RSF in Las Toscas WTP were still biologically active after seven years of operation. According to Mouchet (1992), the biocatalytic process performed by bacteria provokes Fe and Mn biogenic oxides. Subsequently, its autocatalytic properties contribute to maintaining the functionality of the mature media over time (Bruins et al. 2015; Breda et al. 2019a).

Figure 3 shows the variation of Fe and Mn concentrations over time in the sampling points. Vertical dotted lines denote dates of upgrades that were made in the converted water treatment plant during the period of study. During the first two years of operation, concentrations detected in SP1-URF_i ranged around 0.65–1.5 mg/L and 0.25–0.50 mg/L for Fe and Mn, respectively. Additionally, SP2-URF_e reached steady-state conditions for Fe and Mn removal after ∼11 months of operation. Concentrations of Fe in treated water (SP3-RSF_e) were below the MCL 24 h after conversion and Mn required just 15 days to reach the same result. This rapid start-up of RSF was probably due to the original aged coated sand not being replaced during the conversion (Pacini et al. 2014). Moreover, Mn concentrations in SP3-RSF_e did not exceed Mn concentrations (Figure 3(b)) during the first months of operation. Therefore, there was no evidence of Mn release from the original Mn-coated filtration media. It is well-known that the removal of a preoxidant can result in the release of Mn²⁺ (Bassett et al. 2018; Brown 2020).

After the reduction of the aerated water reservoir from 250 to 30 m³ (Event 1), Fe concentrations in SP1-URF_i (Figure 3(a)) increased from around 1.0 up to 2.5 mg/L. This result may be explained by the reduction of the hydraulic residence time of the aerated water reservoir by about 88%, with, therefore, more iron sludge accumulating in the tank and, due to inefficient cleaning processes, precipitates passing to the filters. Similarly, Mn concentrations increased from around 0.3 up to 0.5 mg/L in SP1-URF_i (Figure 3(b)). Probably, the increase of Mn was less noticeable than Fe because the abiotic oxidation of Mn²⁺ by oxygen is very slow at pH values below 9 (Buamah et al. 2009). Subsequently, the higher Fe loading seems to have a negative effect on Fe and Mn removal in the URF. Notably, Mn concentrations in SP2-URF_e gradually increased from around 0.1 up to 0.4 mg/L. It has been reported that Fe concentrations in the feed water and Fe loading were found to be inversely proportional to Mn removal (Bruins et al. 2014). Despite this, URF is appropriate as a pretreatment given its high solid-retention capability (Pacini et al. 2005). Thus, this phenomenon did not affect the Fe and Mn concentrations in treated water (SP3-RSF_e) and the MCL were met. The main exception was of Mn at the end of 2017; however, no explanation was found for this in the historic operational reports.

As was described in Table 2, the aerated water reservoir was restored to its original capacity of 250 m³ in April 2017 (Event 2). It can be seen from Figure 3 that after that, Fe and Mn concentrations in SP1-URF_i decreased, as was expected. Fe concentrations in SP2-URF_e (Figure 3(a)) did not show variations from their behavior before Event 2. Moreover, SP2-URF_e showed more stable behavior (Figure 3(b)) even though it exhibited less Mn capacity uptake compared with the first two years of operation. The upgraded strategy in 2017 did not consider that excess biomass accumulation and inorganic particles in biofilters provoke long-term fouling of media and filter underdrains (Brown 2020). Thus, the backwash system was not improved. Backwashing with water alone is a weak cleaning process due to the limited abrasion and collisions between fluidized particles (Amirtharajah 1993). If the backwashing is ineffective, excessive inorganic particles increase head loss and affect water quality (Amirtharajah 1993). To overcome these shortcomings, in Event 3, a new backwashing strategy using a compressed air scour cycle was implemented together with manifold upgrades to improve fluidization of the filter media. The discharge was reduced by up to 75% after the upgrade. In fact, from Figure 3, it can be seen that Fe concentrations in SP2-URF_e decreased in the subsequent months. Moreover, some values of Mn in SP2-URF_e were around MCL. Remarkably, treated water in RSF (SP3-RSF_e) showed steady-state operation for Fe and Mn removal below MCL during the period of study. Results confirmed the relevance of URF as pretreatment and RSF as a polishing step during the simultaneous removal of Fe and Mn in Las Toscas WTP.

The location of Las Toscas WTP results in typical seasonal temperature variations of 21–32 °C (summer), 17–26 °C (autumn), 11–21 °C (winter), and 16–27 °C (spring) (WeatherSpark.com 2020). According to data collected from 2012, these variations resulted in water temperatures in SP1-URFi of 25–30 °C (summer), 22–25 °C (autumn), 15–19 °C (winter), and 19–26 °C (spring).

Figure 4 shows seasonal Fe and Mn concentrations at the sampling points. Data include ripening and steady-state conditions of the URF and RSF. Although the groundwater source at Las Toscas WTP was from 10 wells and despite the changes in water temperatures, no seasonal variations of Fe and Mn concentrations were observed in the raw water (before aeration). Similar behavior was shown by URF influent SP1-URF_i, suggesting a constant performance of the aeration step. In some cases, groundwater sources have shown important seasonal variations in Fe and Mn concentrations (Li et al. 2005). Seasonal long-term variations in Mn concentration are common at surface water sources (Hoyland et al. 2014). Figure 4 shows similar median values between seasons, in all sampling points of the influents and effluents of the filters, indicating that the seasonal effect was probably negligible (for the operational parameters shown in Table 1). Further statistical tests (section 2.3) confirmed no significant differences between the seasons and Fe and Mn concentrations. Moreover, it can be seen from Figure 4 that the performance of Las Toscas WTP was not affected during winter where the water total production is typically reduced and the operational parameters were different (Table 1).

Several studies have demonstrated that cold winter temperatures (∼3–17 °C) cause a long start-up of biofilters (Cai et al. 2014; Pacini et al. 2014; Lauderdale et al. 2016; Ciancio et al. 2020; Evans et al. 2021). Furthermore, Pacini et al. (2014) found during the first 20 months of operation of Las Toscas WTP a noticeable decrease in the removal efficiency of Fe and Mn in the URF during winter (∼7–17 °C). However, evidence of the present report suggested that this effect was gradually dissipated during the years of operation, probably by the development of new bacterial communities in the filter media capable of surviving at the winter water temperatures (15 °C). In the case of the RSF, seasonal temperature variation did not affect its performance. This is consistent with earlier studies that reported good performance or no seasonal effect on the Mn removal efficiency after the start-up of biofilters. Tekerlekopoulou et al. (2012) found that seasonal variations did not affect the performance of full-scale biofilters for ammonium, Fe, and Mn removal. It is known that low-temperature water will change the growth rate of functional oxidizing bacteria (Cai et al. 2014). In addition, previous studies showed that bacteria growth and biological Mn oxidation could be inhibited at water temperatures below 14 °C (Ratkowsky et al. 1983; Berbenni et al. 2000). Thus, in this study, it seems that seasonal water temperature variations (from 25–30 °C in summer to 15–19 °C during winter) do not affect biofilter performance.

Iron and manganese full-scale biofiltration conversion from conventional filtration systems has received increased attention due to the potential benefits compared with physicochemical systems. However, few studies have addressed their performance during a long-term period, considering the impact of water quality and seasonal variation in the removal efficiencies. The main goal of this study was to evaluate Fe and Mn removal in the converted Las Toscas water treatment plant (WTP) located in Santa Fe, Argentina, for long-term monitoring (from November 2011 to April 2021). The results showed that WTP conversion, consisting of an up-flow roughing filter (URF) and two parallel rapid sand filters (RSF), was effective. The converted WTP effectively removed up to 95% and 88% of Fe and Mn from groundwater for 10 years without the use of chemical reagents or nutrients substrate augmentation strategies. Moreover, significant Fe and Mn removal in the URF (68% and 46%, respectively) and subsequent removal in the RSF confirmed the relevance of the URF as pretreatment and the RSF as a final polishing step. Probably, Fe and Mn biogenic oxides and their autocatalytic properties have contributed to maintaining the full functionality of the mature media over time. No significant differences between the seasonal variation of Fe and Mn concentrations in the URG and RSF influents and effluents were found despite the seasonal fluctuation of the water temperature. Most likely, minimal water temperatures (∼15 °C) did not affect the growth or the Fe-Mn oxidizing capacities of functional bacteria. On the other hand, results emerging from this study indicate that the performance of Las Toscas WTP was mainly influenced by other factors such as the modifications in the aerated water reservoir and the subsequent Fe-loading fluctuations in the feed water to the filters. Finally, this study showed that converting a conventional filtration system to a biological one is stable and efficient through the years and it is, therefore, a suitable strategy.

The authors thank Mr Norberto Zanier and Ing. Gastón Van de Velde of the Cooperativa de Servicios Públicos (CODESELT, Las Toscas, Santa Fe, Argentina) and Lic. Paola Tomadin from the Laboratorio de Análisis Agroindustriales (Quimi-Lab, Villa Ocampo, Santa Fe, Argentina) for sharing valuable data from Las Toscas WTP plant and for the support received during the study. Thanks are also extended to Ing. Paolo Giuliano and Ing. Javier Dallacasa from Ingeniería de Aguas Rosario (IDEAR, Santa Fe, Argentina) for their technical support.

This research was supported by the Vice rectory of Research and Extension (VIE) of the Instituto Tecnológico de Costa Rica (project number 1460-066). The authors would like to acknowledge the Scholarship Committee of the Instituto Tecnológico de Costa Rica for the doctoral scholarship awarded to José A. Araya-Obando and the Center of Sanitary Engineering (CIS) from the Faculty of Sciences and Engineering of the National University of Rosario (Santa Fé, Argentina) for the support received during the internship of José A. Araya-Obando in Santa Fe, Argentina.

There are no conflicts to declare.

All relevant data are included in the paper or its Supplementary Information.