Water Quality Changes during Riverbank Filtration in Budapest , Hungary

The paper gives an overview on the changes in water quality during riverbank filtration (RBF) in Budapest. As water from the Danube River is of high quality, no problems occur during regular operation of RBF systems. Additionally, water quality improved through the past three decades due to the implementation of communal wastewater treatment plants and the decline of extensive use of artificial fertilizers in agriculture. Algae counts are used as tracer indicators to identify input of surface water into wells and to make decisions regarding shutdowns during floods. RBF systems have a high buffering capacity and resistance against accidental spills of contaminants in the river, which was proven during the red mud spill in October 2010. The removal rate of microorganisms was between 1.5 log and 3.5 log efficiency and is in the same order as for other RBF sites worldwide.


Introduction
Riverbank filtration (RBF) is a widely used natural water treatment process where, by definition, at least 50% of treated water must originate from surface water.It has been observed that the surface water source, the hydrogeological characteristics of the aquifer, the protected watershed area and the particularities of production play an important role on the quality of the produced water [1][2][3].
RBF offers many advantages concerning improvement of water quality.This type of ecosystem service is used in many watersheds globally, including India, China, USA and Germany.RBF along the Danube River has been used for water supply in Budapest for over 150 years [4].Due to the high quality of the Danube River water and favorable hydrogeological conditions at the Szentendre Island upstream of the city of Budapest, no post-treatment except disinfection is required after RBF processes.This unique situation enables us to study long-term trends in the characteristics of water quality parameters.
The focus is set on basic water quality parameters to describe microbiologically mediated reactions, physical sorption and mixing processes during RBF and the resulting attenuation of pollutants.Spatial changes in redox potential conditions were studied in RBF processes in Berlin, where it has been found that temperature variation strongly influenced the efficacy of microbial removal processes [5].
The vulnerability of RBF processes to climate change has been discussed in prior studies concentrating on oxic/aerated and anoxic conditions of the aquifer layers [6].Redox conditions are profoundly affected by both microbiologically mediated pathways of nitrogen transformations (nitrification and denitrification) and physicochemical sorption processes and phase equilibria [7].Local redox conditions, however, can only be indirectly controlled in the aquifer by the operator (i.e., pumping rates).In spite of the improvements in traditional water quality parameters, concerns arise regarding the microbial parameters of the Danube both upstream [8] and downstream of Budapest [9] related to the increasing incidence and severity of extremities.
The aim of this study is to give an overview regarding the efficiency of RBF processes.The basic concept is to analyze physical, chemical, microbiological and biological parameters and highlight existing connections.Challenges include seasonal variations in river water quality, floods, droughts, industrial and agricultural pollutant input variations.Therefore, it is important to consider water quality parameters which can be determined at a high number, high frequency and at low cost.Also, it is important to determine how these measurements can improve the level of service by faster and established interventions, lower disinfectant concentration and effective operational strategies.

Site Description
As the efficiency of RBF is site specific and the water quality changes are affected by many other factors besides source water quality, e.g., water level changes, travel times of bank filtrate, pumping regime of wells, etc., a large dataset is required to be able to determine reliable operational methodology.In this paper, data from the period 2006 to 2017 from a total of up to 756 wells were overviewed to assess changes in water quality.The maximum capacity of the RBF systems of Budapest Waterworks is 1.0 million m 3 /day; the recent average supply is about 456,000 m 3 /day.Compared to the average discharge of the Danube River in Budapest, which stands at 200 million m 3 /day, only 0.23% of the water is extracted from the river discharge via bank filtration.A unique situation occurs in Budapest whereby there is no riverbed clogging observed [10] and no distinct clogging layer exists in the riverbed affecting water quality.This may be due to the high river flow velocity of 0.8-1.6 m/s, the depth of the river and the related shear forces.At such levels of flow velocity, fine particles do not settle, only coarse sand and gravel do [11].At many other RBF sites worldwide, clogging profoundly affects the infiltration rates of river water and results in a highly active biological layer in the riverbed which often notably contributes to water quality changes, especially considering oxygen consumption and attenuation of organic compounds [11][12][13][14][15].
Budapest Waterworks operates 756 RBF wells to supply water to 1.89 million inhabitants.The wells are predominantly located on Szentendre Island and Csepel Island (Figure 1).A detailed description of well types and operation procedures are to be found in Nagy-Kovács et al. [16].

Groundwater Flow Modeling
Travel times have been determined by ground water flow modeling using the MODFLOW software (USGS, Reston, VA, USA).The modeling served to determine the travel time of water particles arriving in the well.The ratio between river bank filtrate and ground water was investigated.The original proportion of the produced water is primarily controlled by the actual level of the Danube and the rate of drawdown.Calculations were carried out based on a 2-m average Danube level and a drawdown of 2 m.Later, separate well capacities for different Danube water levels were also determined in 2012 [17].Table 1 gives an overview of the distances and travel times between the Danube River and the production wells.

Groundwater Flow Modeling
Travel times have been determined by ground water flow modeling using the MODFLOW software (USGS, Reston, VA, USA).The modeling served to determine the travel time of water particles arriving in the well.The ratio between river bank filtrate and ground water was investigated.The original proportion of the produced water is primarily controlled by the actual level of the Danube and the rate of drawdown.Calculations were carried out based on a 2-m average Danube level and a drawdown of 2 m.Later, separate well capacities for different Danube water levels were also determined in 2012 [17].Table 1 gives an overview of the distances and travel times between the Danube River and the production wells.

RBF Monitoring Network and Samples
Water sampling from the river and the wells was carried out following the Hungarian guidelines and standards [18].Analyses are carried out systematically and adjusted to changing circumstances to ensure safe and secure water supply and to gain data for optimal operation of the RBF systems.The Danube River was sampled at least weekly either on Szentendre Island or Csepel Island.Every well in operation was sampled regularly at least twice a year from its sampling tap.Some siphon systems were sampled at their collecting pipes.
All analytical methods for the determination of discussed parameters are provided in the Supplementary Material.
The minimum, median and maximum values were prepared.Due to the large number of data, short events such as floods or spills would not affect the median values which are used to determine removal rates for different groups of RBF wells.
For physical and chemical parameters, mean removal rates have been determined for the whole time period.Lowest and highest removal rates were calculated using the mean concentration in the river water and the maximum and the minimum concentration in the bank filtrate, respectively.As the sampled well water is a mixture of bank filtrate of different age depending on the location of infiltration in the riverbed, the depth of the flow path in the aquifer and the pumping rate of the well, pairing of data is not useful.In no case, data from the same date of sampling can be compared as the water sampled from the well is days to weeks old and has nothing to do with the river water quality at the sampling date.As RBF acts as a buffer for water quality, it is feasible to use the mean concentration in river water and to compare with minimum and maximum concentrations observed in the well water.For microbiological parameters, the mean logarithmic removal rate was calculated as the difference between the logarithm of the average cell count in Danube River water and the logarithm of the average cell count in the bank filtrate.
The available dataset has been discussed in five different Chapters (Sections 3.1-3.5).For all chapters, a table has been prepared to better demonstrate the results that have been analyzed during this study.Median values are in bold presented in the first row, the range of each parameter with minimum and maximum values is given in the second row, and n represents the total number of samples for the parameter in italics.The dataset is formed by culminating results from the monitoring plan determined by Hungarian regulations, operation-related experiments, sampling during extreme hydrological events and a major accidental pollution event in the Danube River basin.Due to this fact, the number of samples (n) varies for each parameter, as authors chose to present all reliable data from the period 2006-2017.

Physical Parameters and Selected Cations and Anions
This chapter summarizes the results of physical parameters, cations and anions measured on a regular basis.The temperature of the river water ranges from −1.4 to 26.3 • C with a median of 13.2 • C (Table 2).Due to the short distance between the riverbank and the wells and the heat capacity of the aquifer material, only a low buffering effect was observed-the temperature of bank filtrate ranges from 0.7 to 21.0 • C with a median of 11.9 • C. As the aquifer thickness is only 5-17 m (Table 1), the buffering effect is lower in Budapest compared to other RBF sites with larger aquifer thickness, e.g., in Torgau with 50-60 m thickness [14,19].The electrical conductivity (EC) of river water and bank filtrate varies from 283 to 652 µS/cm and from 303 to 1809 µS/cm, respectively.The mean EC is very similar for the Danube River water and well groups on Szentendre Island, indicating a high portion of bank filtrate, which has been calculated from groundwater flow modeling as 60-80% [17].The maximum EC values in bank filtrate, which are significantly higher than those of the river water are only observed on Csepel Island, where industrial and agricultural activities even outside of the well head protection zones are still affecting the water quality in some wells.The mean turbidity of bank filtrate is very low Water 2019, 11, 302 5 of 14 (0.05-0.07 NTU) compared to river water and the removal of particles is not a function of travel time.Slightly increased turbidity values on Csepel Island are related to iron and manganese precipitates.Despite the long-term operation of most of the wells, there is still a dissolution of carbonates in the aquifer, resulting in an increase in hardness (Ca and Mg) compared to river water.Sodium and potassium concentrations in river water and bank filtrate are within the same range.
As for sodium, chloride concentrations in river water and bank filtrate are within the same range, indicating a high portion of bank filtrate.Fluoride, boron, silicon and cyanide concentrations are at low levels both in river water and bank filtrate and do not pose any risk for the water supply.Phosphate levels in river water are below the LOD (limit of determination), but are surprisingly also found in the bank filtrate.
Seasonal temperature changes in river water and bank filtrate and flow-related changes in EC can be used to estimate travel times of the bank filtrate [14,20].Figure 2 shows the temperature data for the Danube River water and well group Balpart II with a short travel time in the range of 2-5 days.It can be seen how the seasonal temperature variation of the river influences the river bank filtrate as the buffering effect of the aquifer is considered low.Also, a change in the water temperature variation can be seen starting from August of 2012 that is linked to the fact that the production rate decreased by 30% at the site.

Redox-Related Parameters
The removal of organic compounds (mainly natural humic and fulvic acids) is of high relevance for the required post-treatment (to supply microbiologically stable drinking water) and especially for disinfection with regards to the potential formation of disinfection by-products.The total organic carbon (TOC) concentration in the Danube River water ranges from 1.6 to 10.0 mg/l (Table 3) and has a median similar to that of the Rhine River at Düsseldorf [21] and about half of the concentration in the Elbe River [14,19].Despite the low input concentration, the removal rate for TOC is relatively high, ranging from 11 to 75%.The TOC removal is higher if the travel time is longer.The removal of organic aromatic compounds causing UV absorption ranges between 0 and 92%.The median specific UV-absorbance (calculated as UV254/TOC) is 2.61 l/(m•mg) for Danube River water and 1.93-2.45l/(m•mg) for bank filtrate.At the RBF site Torgau, Elbe River, Germany, the specific UV-absorbance of the river water was 2.94 l/(m•mg), increasing to 3.17 l/(m•mg) in the riverbed, where easily biodegradable and less UV-active organic compounds were removed, and decreased along the >200 m long flow path to 2.54 l/(m•mg) due to further attenuation of UV-active compounds [14].In Budapest, a similar removal indicates high removal of UV-active compounds.
The chemical oxygen demand (COD) removal rates during RBF were about 7-93% but cannot be used as the TOC to assess the removal of organic compounds as it cumulatively removes all oxidizable compounds present in the water, including inorganic constituents such as ammonium and iron.

Redox-Related Parameters
The removal of organic compounds (mainly natural humic and fulvic acids) is of high relevance for the required post-treatment (to supply microbiologically stable drinking water) and especially for disinfection with regards to the potential formation of disinfection by-products.The total organic carbon (TOC) concentration in the Danube River water ranges from 1.6 to 10.0 mg/L (Table 3) and has a median similar to that of the Rhine River at Düsseldorf [21] and about half of the concentration in the Elbe River [14,19].Despite the low input concentration, the removal rate for TOC is relatively high, ranging from 11 to 75%.The TOC removal is higher if the travel time is longer.The removal of organic aromatic compounds causing UV absorption ranges between 0 and 92%.The median specific UV-absorbance (calculated as UV 254 /TOC) is 2.61 L/(m•mg) for Danube River water and 1.93-2.45L/(m•mg) for bank filtrate.At the RBF site Torgau, Elbe River, Germany, the specific UV-absorbance of the river water was 2.94 L/(m•mg), increasing to 3.17 L/(m•mg) in the riverbed,

Figure 1 .
Figure 1.Location and names of well groups of the Budapest Waterworks at Szentendre Island (upstream the capital) and Csepel Island (downstream the capital).

Figure 1 .
Figure 1.Location and names of well groups of the Budapest Waterworks at Szentendre Island (upstream the capital) and Csepel Island (downstream the capital).
Water 2018, 10, x FOR PEER REVIEW 7 of 15

Table 1 .
Distance and travel time between the Danube River and RBF wells in Budapest.

Table 1 .
Distance and travel time between the Danube River and RBF wells in Budapest.

/Well Group Type of Wells Distance between the Riverbank and Wells (m) Thickness of Aquifer (m) Travel Time of Bank Filtrate (days)
HW-horizontal (collector) well, VW-drilled (vertical) well, SW-shaft well.

Table 2 .
Physical parameters and selected cations and anions, median (min-max) values of Danube River water and bank filtrate with different travel time (t), Budapest, 2006-2017.