Understanding short-term organic matter fluctuations to optimize drinking water treatment

A reliable and safe drinking water resource is the cornerstone of a healthy community and represents an indispensable asset for society. In Sweden, more than 50% of the drinking water is produced from surface water bodies (Svensk Vatten 2016), which is also the case for the larger part of Europe (European Environment Agency 2021). Human interactions with these water bodies and their catchments, such as irrigation, industrial activities, and recreation, have resulted in changes in water quality, including poorer ecological status (Peters et al. 2005; Zhang et al. 2020). However, natural changes in the environment also affect water quality, including climate variability and climate change.

In recent decades the process of brownification has gained more and more attention. Brownification results in a more yellow to brown-colored water, caused by increased natural organic matter (NOM) and iron concentrations in the water, mainly introduced through the catchment (Löfgren et al. 2003; Graneli 2012; Ekström 2013; Björnerås et al. 2017). The higher concentration of iron and organic matter, primarily organic carbon, diffracts light in a way that the water appears brown to the human eye. Brownification has mainly been observed in the northern hemisphere and knowledge about the working processes is still limited. The specific causes and associated drivers are many, since brownification occurs in a complex aquatic environment. Recent research consolidates around afforestation, climate change, and recovery from atmospheric sulfur deposition as main causes (Monteith et al. 2007; Björnerås et al. 2017; Kritzberg et al. 2019; Škerlep et al. 2020), but which of them that contributes the most is widely discussed. Kritzberg (2017) has shown that sulfur deposition may not be a main contributing factor as supposed by others, since the browning of water resources has continuously increased even after the peak period of sulfur deposition. Other researchers hypothesize that browning might be the return to a more natural state regarding the color of water, which was present in the northern hemisphere before industrialization (Meyer-Jacob et al. 2019). The relative importance of different causes is typically site specific, which makes it complicated to identify general mechanisms for all locations in the northern hemisphere (Eikebrokk et al. 2018).

Brown-colored water and the associated higher content of NOM has a manifold of non-desirable effects on drinking water production, but also on the recreational value of affected water bodies (Keeler et al. 2015; Eikebrokk et al. 2018). During drinking water production organic matter needs to be removed to avoid unwanted disinfection byproducts (Sillanpää 2015; Tak & Vellanki 2018). A higher NOM content negatively influences the process of disinfection with ultraviolet light and promotes microbiological regrowth in the distribution system (Hem et al. 2015). Moreover, brownification results in challenges for the ecosystem, for example, increased surface water temperatures due to higher energy dissipation of solar radiation induced by a darker water surface could disturb the fish community (Taipale et al. 2016; Hedström et al. 2017; Creed et al. 2018; Van Dorst 2020). Changing light transmission because of coloring may affect lake stratification, resulting in different oxygen levels in the water column, which in turn influence the food web structures of the ecosystem (Williamson et al. 2015; Pilla et al. 2018).

Various methods are used to remove NOM when treating water, including coagulation (chemical precipitation), filtration and adsorption to more complex procedures (Zhang et al. 2015). The most widely used practice is coagulation and flocculation (Eikebrokk et al. 2018; Health Canada 2019), in which metal salts are used to precipitate organic matter and enable gravity separation. Filtration is also a widely used alternative (Matilainen et al. 2010; Lidén 2016; Keucken et al. 2017). Higher organic content in the source water requires increased effort during treatment in all the methods (Sillanpää 2015), for example, increased coagulant dosages, increased backwashing frequency of filtration systems, as higher organic loads lead to faster clogging as well as higher maintenance costs (Ma & Hsiao 2016; Keucken et al. 2017; Hägg et al. 2020).

NOM is a complex mix of organic material (Sillanpää 2015). It has a broad molecular weight distribution that makes removal complex (Matilainen et al. 2002). Matilainen et al. (2010) reviewed NOM removal through coagulation. The future challenges regarding increased NOM loads have been investigated, highlighting the necessity for a better understanding of NOM variations in the water coming to the treatment plant (Sillanpää 2015)⁠. Continuous adjustments according to the characteristics of the water being treated are necessary to allow for reliable and feasible treatment to meet governmental requirements and produce drinking water of high quality. Especially, surface waters are subject to changes in quality rather quickly, depending on changes within the catchment (e.g., meteorological and environmental factors). Predictions, as well as specific effects, of such changes are complicated and the uncertainty of lag time between changes happening inside the catchment and effects on the raw water quality is a challenge for the WTPs, especially when operated on a small scale (Scheili et al. 2016). Being able to make predictions of the expected raw water quality based on variations inside the catchment and upstream water bodies would enable the adjustment of the drinking water treatment process to a higher extent and even ahead of time. This procedure would help to better estimate the treatment costs, amounts of chemicals to be used, and accruing waste (sludge) (Persson 2011). In addition, it can help to increase the resilience of drinking WTPs, to allow for better maintenance planning or testing of new treatment practices and to test new treatment practices and technologies, as well as to assess possible expansion of the plant. Not only the drinking water producing companies would benefit from knowledge of incoming water quality and its variation, but also the consumers, since a more stable drinking water quality can be produced, possibly at a lower price. Source water quality predictions are usually available for long-term changes (Lidén 2016; Eikebrokk et al. 2018) however, the understanding of short-term changes is lacking and urgently needed.

Understanding short-term organic matter fluctuations in the incoming water to a drinking water plant facilitates optimal operation of treatment processes to remove such matter. The main objectives of this study are to understand and quantify short-term changes (days to weeks) resulting from meteorological and environmental factors, as well from the water transfer system. The focus is on changes in organic matter content leading to increased treatment effort for drinking water production and unwanted ecological effects. The present analysis was based on detailed water quality measurements at the major raw water resource of one of Sweden's largest drinking water production companies, Sydvatten AB.

Sydvatten AB produces potable water for almost one million consumers living in the southernmost county of Sweden. RWTP accounts for about 60% producing 1400 L/s on average (Sydvatten AB 2022). Besides the importance as a raw water resource, Lake Bolmen holds high ecological and recreational values.

The procedure employed was to analyze available data to detect variations in water quality, subsequently quantifying these variations and investigating their origin and controlling factors. In the water supply system studied, different elements contribute to the water quality acting on a wide range of time and space scales. Knowledge of these variations is useful in optimizing the treatment processes at the RWTP, minimizing costs and reducing waste products at the same time delivering drinking water of high quality. Although a specific case study is discussed here, general conclusions can be made regarding typical patterns of variation in different water quality parameters and their governing processes, as well as how such knowledge can be employed to streamline treatment processes.

The analysis here deals primarily with the two latter elements since they tend to produce short-term variations (i.e., days to weeks) that are of interest in the management of the RWTP. Changes in the main water body of Lake Bolmen produces variations on time scales from seasons to decades, which is of concern when discussing long-term trends such as land-use practices and climate change.

Lake Bolmen is a dimictic, oligotrophic lake with a catchment dominated by coniferous forests and wetlands. The lake is of significant value for commercial fishing, recreation, and the ecosystem services it provides. During the last decades the water in Lake Bolmen has been subject to brownification and the magnitude of color has doubled (Borgström 2020; Klante et al. 2021). The lake covers 173 km², it is Sweden's twelfth largest lake and the largest lake within the Lagan River Basin. Since 1987 it has been the main source water for RWTP.

RWTP was commissioned in 1963 and initially Lake Ringsjön was used as source water. The population increase in the Malmö metropolitan area during the 1960 and 1970s led to the decision to use Lake Bolmen, located about 100 km north of the RWTP, as a future source water, since Lake Ringsjö was not large enough to meet the demand. The transport of raw water is realized through a 80 km long tunnel (the Bolmen tunnel) and a 25 km long pipeline (south of Äktaboden) (Sydvatten AB 2020, 2021, 2022). The tunnel is located south of Lake Bolmen and it is separated from the major part of Lake Bolmen through the sub-basin Kafjorden in the south (see Figure 1).

Different data sets were used spanning the period 1st of November 2015 to 25th of October 2020. Unless otherwise specified the typical data resolution is daily or hourly measurements. Approximate sampling locations are shown in Figure 1.

Data on precipitation and wind were collected by the Swedish Meteorological and Hydrological Institute (SMHI) on an hourly basis. Two meteorological stations are close to the study area and another is further away to the west (Figure 1). Bakarebo station is about 15 km north of the inlet to the Bolmen tunnel whereas Ljungby is approximately 20 km east. Precipitation data from Bakarebo station were used because, unlike Ljungby, it is in the Lake Bolmen catchment. Annual precipitation at Bakarebo and Ljungby differs significantly by up to about 25%. Mean annual precipitation is 857 mm in Bakarebo and 719 mm in Ljungby, a difference of 16%. Information regarding wind direction and velocity were acquired from Ljungby and Torup weather stations, including information about wind gusts. As no measurements are available closer to the study area, spatial similarity was assumed.

Monthly data on the Enhanced Vegetation Index (EVI) for the Lake Bolmen catchment were obtained from satellite images. EVI is a proxy for biomass, reflecting the amount of organic material available for transport from the catchment, which should be correlated with color (Didan et al. 2015; NASA EOSDIS Land Processes DAAC 2015).

Modeled flow data were available from SMHI. The simulated data are based on the S-HYPE model (Swedish Meteorological & Hydrological Institute 2022), which is the hydrological model developed by SMHI covering the whole of Sweden. In- and outflow of the Bolmen tunnel, as well as the total discharge from Lake Bolmen, are derived from measurements.

Information about ice cover (freeze-up and break-up) on Lake Bolmen was recorded by SMHI until 2006. Similar data from Lake Vidöstern, approximately 20 km north-east of Bolmen, was used to reconstruct an ice cover dataset for Lake Bolmen for the study period.

Water quality measurements carried out by Sydvatten AB were used in this study. They were derived from online monitoring equipment that records at fixed intervals. Data with daily and hourly time steps were used, with the focus on the latter.

Two sampling locations were selected, the first at the Bolmen tunnel inlet in the southern part of Lake Bolmen, and the other at RWTP source water inlet after the water has passed through both the 80 km long tunnel and the 25 km pipeline. The parameters measured at the first location were absorbance of ultraviolet light at 254 nm (A254), water level, turbidity and flow into the tunnel. The same measurements were taken at the second location and in addition, nitrate, total organic carbon (TOC), dissolved organic carbon (DOC), and dissolved oxygen were recorded. However, in this investigation the analysis has been limited to parameters that are available at both locations. The A254 sensor was set to a maximum of 70 m⁻¹ until the end of 2020. Thus, at very high peaks the data level out and the actual peak magnitude were not recorded. After 2020 the limit was increased to 120 m⁻¹ and only once this limit was reached. For both cases there are plateaus in the data series, and the actual magnitude may have been higher than that recorded. Variation above the levels mentioned are not represented.

Monthly samples were also taken approximately 4 km upstream of the tunnel inlet at Piksborg (Figure 1). These data were used to determine the effect on water quality during transport through Kafjorden. Data from the local water association were also used to analyze the situation in Kafjorden with regard to the inflow at Piksborg, coming from the main part of Lake Bolmen, and the contribution from Murån, a tributary discharging downstream of Piksborg.

Note that A254 is not commonly used to measure the color of water regarding brownification, but typically an absorbance above 400 nm is employed. This is because brownification can be caused by increased levels of iron as well as NOM. The combination of those is best represented by a measurement above 400 nm (Bennett & Drikas 1993; Löfgren et al. 2003; Köhler et al. 2013; Björnerås et al. 2017; Munar et al. 2018). However, the high correlation between A254 and the NOM fraction, which is supposed to be removed by coagulation in water treatment plants (WTPs), makes it a suitable measure for adjusting coagulant dosages in real time (Köhler et al. 2016; Cascone et al. 2022). Thus, color in this study always refers to A254 unless specified otherwise.

The magnitude of NOM and yellowish-brown colored water in Lake Bolmen is subject to variations in time and space at different scales. Drivers for these variations are manifold, but commonly related to changes occurring inside the catchment and to internal processes in the water body (Björnerås et al. 2017; Kritzberg et al. 2019; Klante et al. 2021). Given the temporal resolution of the analyzed data, this study mainly focuses on changes on the short-term scale, including weekly, daily, or hourly changes within rather close proximity to the area of interest, see Borgström (2020) and Klante et al. (2021).

Standard statistical parameters (e.g., mean, standard deviation, percentiles) were calculated initially to characterize the data sets. Lag-time corrected Pearson correlation was used to determine co-variation of parameters at different locations, including the travel time of water in the Bolmen tunnel. It was also employed to relate changes in A254 to variations in precipitation, runoff, and EVI, as well as to determine the time difference between rainfall events and the effects on runoff. Regression analysis and more advanced statistical techniques, for example, spectral methods and Singular Spectrum Analysis (SSA), were also employed to quantify patterns in the data. Mass balance calculations were used to estimate the possible contribution to NOM from different sources. Residence times in the water bodies were calculated using a simple plug-flow model. Given the time scales analyzed, it was assumed that biotic interactions between NOM and the environment are limited because organic material is nearly inert on such a time scale (Tranvik 1998).

For various reasons, recordings from measuring equipment can be associated with errors that may impact the analysis. Thus, data quality control is needed to handle such values without introducing a bias. Data quality control and censoring were performed through a combination of procedures. The standard deviation was calculated for a sliding window of size X (usually between 7 days and one year) and all data points that fell outside ± 4 times the standard deviations were removed. Negative and positive peaks that fell outside the standard deviation of a rolling window of size X (usually between 1 day and four months) were also removed. Remaining outliers outside the upper and lower 0.1 percentile were neglected. Any data points with a larger difference than one between the actual data and the median of a rolling window of size X (usually 24 hours) were removed.

The order of the procedures and the window sizes (X) were adjusted depending on the input characteristics of the data. However, when the sliding window approach was used, it was always the first step. This procedure resulted in an average loss of data below 1%, whereas the highest loss was 2.5%. After the removal of unreliable data, missing data were interpolated using a time-step respecting approach. These procedures were only necessary for data from Sydvatten AB, as data from SMHI have already been subject to their standard quality control procedures.

The following section discusses the water quality variations in the source water from smaller to larger scales, both with reference to space and time.

The Bolmen tunnel is drilled into solid bedrock and its average depth is 50 m below ground surface (Stanfors 1987). Thus, the tunnel is located below the groundwater table, which is typically encountered at a depth of 2.75–4.3 m below the ground surface along the line of the tunnel (Geological Survey of Sweden 2021). The bedrock along the tunnel is mainly gneiss of varying composition with minor amounts of granite. There are several zones of fractured and crushed rock (Stanfors 1987) with groundwater within the fractures (Fetter 2014). As a result of the tunnel design, groundwater enters the tunnel, which may cause groundwater drainage and leakage into the water from Lake Bolmen. However, under normal circumstances there is leakage into the tunnel (WSP Environmental 2011b). Due to variations in the bedrock and the depth of the tunnel the leakage varies along the tunnel. Considering the entire tunnel, the inflow is relatively evenly distributed and a leakage flow of about 0.001 L/s/m has been estimated (WSP Environmental 2011a). At an average flow rate of 1.4 m³/s, the groundwater inflow would contribute about 0.5% of the flow leaving the tunnel. The flow measurements at the tunnel confirm the intrusion of groundwater, although the measured mean flow difference for the period 2015–2019 indicates a higher contribution of groundwater estimated to be 0.85%.

The reduction in color and turbidity during the transport through the tunnel is most likely due to sedimentation. The tunnel contains stretches with rather still-standing water where settling of particles can occur. A simple, steady-state mass balance model for the tunnel results in an exponential decrease in the concentration along the tunnel. With the applicable data, the observed reduction in concentration experienced in the tunnel corresponds to a sediment fraction that settles in the range of very fine silt to clay (about 0.004 mm). Although the shear stress along a plane-sloping tunnel would exceed what is required to initiate transport for such fine fractions, the actual shape of the tunnel and the pools of water present probably allows for sedimentation. There is an apparent risk that increased extraction through the tunnel, implying increased flow and water velocity, could resuspend some material and cause less reduction in the tunnel regarding the transport of certain water quality parameters.

However, water quality measurements of higher resolution are needed to allow for better estimation of the interacting processes determining the raw water quality, including more details about the processes inside the tunnel, its hydraulics and interactions with the surrounding bedrock. To develop the knowledge about the connection with the intruding groundwater, further analysis, involving other water quality parameters, is required.

Not all surface WPTs get their raw water from a lake, but rather from streams. Typically, water quality in a stream is subject to changes at a higher frequency, responding to changes within the catchment more quickly than a lake. A detailed analysis of long-term changes within a stream located in southeastern Sweden and the effects on the water color was made by Škerlep et al. (2020). In comparison, a lake acts as a buffer regarding quality changes, assuming that the raw water is not taken in close to the main tributaries of the lake. Temnerud et al. (2014) analyzed the drivers of color in different streams and clearly demonstrated that the buffer capacity of lakes in a stream catchment had a significant impact on the lake water quality.

Conditions in the southern part of Lake Bolmen are expected to influence Kafjorden markedly, since most of the inflow originates from the main lake.

However, A254 measured at the tunnel inlet is always higher than the upstream measurements at the Kafjorden inlet (Figure 7). At the tunnel, similar values to those for the inflow to the sub-basin are expected, possibly even lower because of the increasing water volume downstream and associated dilution. The most likely explanation is that a major contribution of brown-colored water to Kafjorden occurs downstream of the inlet. There are two main possible sources: (1) cross-boundary transport of material to Kafjorden (e.g., from tributaries, direct surface runoff, and surface drainage through ditches), and (2) internal processes, mainly resuspension. which is the most likely source.

To determine whether Murån is a major contributor to color in Kafjorden, a steady-state mass balance approach was attempted based on measurements at the three locations, Kafjorden inlet, Murån, and Bolmån (outflow from Kafjorden). However, the measurement frequency at Murån is lower than at the other sites; samples are taken here every other month. TOC, a well-known proxy for water color (Škerlep et al. 2020) was used for calculation. Because of the limited data available and the uncertainties involved, a robust and reliable estimate of the relative contribution from Murån could not be made, even though it is clear that color values at Bolmån are always higher than at the Kafjorden inlet. Viewing time series of color measurements, events when Murån are expected to contribute to the increase in color in Kafjorden may be identified (Figure 8), but in other cases the reason for this increase may be related to other factors. Overall, the resulting color in Kafjorden is a function of a multitude of factors, interacting in a complex manner. To allow for a more detailed analysis of the contribution to Kafjorden from the different sources, more measurements at a higher frequency are necessary.

Another factor possibly affecting the color in Kafjorden is resuspension of bottom material by surface waves during severe wind conditions. The possibility of resuspension in Lake Bolmen itself was discussed by Klante et al. (2021). As Kafjorden is relatively shallow, on average 2.5 m, with fetch lengths of up to 2000 m, the potential for resuspension is rather large. However, the availability of bottom material to be resuspended depends on the depositional conditions during calmer conditions (e.g., velocity, turbulence, available material), that is, whether it is possible for NOM to settle (Bengtsson & Hellström 1992). Waves have much greater ability to resuspend bottom material than large-scale, unidirectional currents, since an oscillatory boundary layer flow develops beneath waves, implying larger shear stresses. Mobilization of bottom material is related to the shear stress on the bed.

For the area around Lake Bolmen the prevailing winds are southwesterly and westerly. Average wind speed at Ljungby is 2.3 m/s and maximum gust speed was 32.8 m/s on the 8th of January 2005. At Torup, approximately 40 km northwest of the study area the wind climate is similar. Average wind speed is 2.9 m/s and the maximum measured gust speed was 28.5 m/s, measured on 28th of October 2013. Because of its similar topography, implying higher agreement with the study area, the analysis in this study was performed with data from the Ljungby station. The wind measurements are referenced to an elevation of 10 m above ground. The data were not corrected for the local influences of topography, vegetation, and structures.

Measurements at Lake Bolmen (Borgström 2020) and studies at other locations (Tranvik 1998; Meyer-Jacob et al. 2019) show that the concentration of organic matter, and as a consequence the color, is higher at the bottom of the lake compared to at the surface, indicating sedimentation and increased concentration with depth. Thus, events involving strong winds and resulting surface waves will have the possibility to resuspend settled material. Such resuspension can lead to a higher concentration of colored organic matter within a limited period when the hydrodynamic disturbance of the lake system is large. No wave measurements are available for Lake Bolmen and the wave climate must be obtained using mathematical models, as done by Klante et al. (2021). Wind direction is decisive in determining the emergence of large waves because of the fetch lengths. In Kafjorden, fetch lengths are longest northeast-southwest (about 2000 m), while the northwest-southeast fetch length is about 1000 m. Other topographical features, e.g., forested areas around the lake and smaller islands, will also influence wave build-up. The wind-generated waves are typically fetch-limited in Kafjorden, implying that the fetch length determines the wave properties, not wind duration. For most wind speeds equilibrium is reached for less than an hour for the longest fetches. The water surface needs to be free of ice cover for wind forces to generate waves.

The results presented indicate the possible presence and importance of resuspension. Currently, available high-frequency data are taken at a single location (the tunnel inlet), which is not exactly where resuspension occurs, although upstream effects due to this phenomenon should be observed here. With regard to Lake Bolmen itself, the measurements at the tunnel inlet are not able to closely represent the prevailing water quality conditions of the more exposed areas, where resuspension also may take place. Measurements at higher frequency for different locations in the lake could help to investigate the general prevalence of resuspension. It is also important to stress that the wind climate used for the analysis is located about 20 km east of the actual area of interest and outside the actual catchment of Lake Bolmen. Due to topographical influences the wind climate at Lake Bolmen may be slightly different in time, magnitude, and direction.

Water quality in lakes correlate strongly with catchment properties, although internal processes are sometimes important depending on upstream transport conditions. Significant changes inside the catchment, such as changed runoff or vegetation cover will affect the water quality. Source water from Lake Bolmen is used to produce drinking water, which requires awareness of such changes that may lower the source water quality. It should be pointed out that changes are often not of natural origin but are caused by human activities. When the latter is the case, the awareness typically needs to be even higher, since chemical and biological contaminants from such activities may have long-lasting effects on the water quality.

RWTP has an advantage because of the transport time through the tunnel and pipeline that other treatment plants typically do not have, a lead time of about 8 days is available to adjust the treatment processes. Nevertheless, knowledge of short-term fluctuations in water quality of the incoming water to a treatment plant facilitate better management.

Since the processes affecting water quality are complex, as well as their interactions, with marked variation in time and space, it is difficult to derive simple relationships that connect driving factors with observed parameters (e.g., color). This was also the case in the present study where such relationships could only be established for selected events, but hard to demonstrate for entire data series. Although the present analysis for the water supply system at Lake Bolmen did not in general connect observed peaks in color to severe weather events, such connections were established in several cases. The importance of Murån, as well as resuspension of bottom material due to surface waves, as sources for color were established.

More measurements in space and time would facilitate the development of quantitative relationships between color and the governing factors. In the case of RWTP, it is recommended to more carefully monitor the small tributary catchment of Murån that probably contributes significant material transport to Kafjorden, which would allow for a reliable quantification of this contribution. Measurement of the water color in Kafjorden in connection with strong winds together with mapping of the bottom sediment would resolve the importance of resuspension.

For treatment plants that are not able to monitor the incoming water quality ahead of time like RWTP, increased preparedness in connection with significant precipitation events as well as storms should be considered, because higher NOM content is much more likely to occur after such events. If the aim is to optimize the treatment processes and adjust specific chemical dosages, it is important to know how the transport through the water supply system to the treatment plant works, including traveling times and mixing conditions. In general, all treatment plants can benefit from knowledge on the relationship between severe weather events and NOM content in the water, even without the advantage of monitoring ahead of time. One example may be the scheduling of maintenance work of specific parts in the treatment plant; another example could be plant expansion to accommodate increasing population. When constructing new treatment plants, it is suggested that source water resources are analyzed in advance to detect point sources of organic matter and to allow for strategic placement of the water intake to mitigate the influence of such sources.

In several studies, e.g., Persson (2011), Björnerås et al. (2017), Borgström (2020) and Klante et al. (2021), an increasing trend in water browning was observed for Lake Bolmen, which will affect the water treatment in the future, leading to raised treatment efforts, increased costs, and more waste material. Most climate change scenarios predict a rise in temperature and higher probabilities for extreme weather events (Swedish Meteorological & Hydrological Institute 2021). Rising annual mean temperature results in milder winters (for Bolmen above freezing temperatures) and may subsequently lead to the lack of ice cover and an exposed water surface increasing the potential of bottom sediments being resuspended by wave-induced currents. In addition, the annual average temperature of Lake Bolmen will increase, which may influence the stratification, affecting the sedimentation process of NOM. and resulting in a higher color magnitude. Furthermore, negative impacts on the ecosystem are expected, since stratification is an important feature in lake ecology in the boreal region and higher water temperatures during summer will affect the food web system within lakes as well as negatively influence the water quality at the bottom.

The purpose of the study was to understand and quantify short-term changes and their governing factors in a water supply system, focusing on the NOM content (i.e., color). The system studied comprised Lake Bolmen and the downstream located treatment plant, RWTP. Two system elements influence short-term water quality changes at RWTP: (1) a tunnel and pipeline, and (2) Kafjorden, a sub-basin of Lake Bolmen. The former tends to produce changes over hours and days, the latter over days to weeks. Longer time scale changes (seasons to decades) are typically related to Lake Bolmen and its catchment.

In the tunnel, besides hydrodynamic effects such as advection, diffusion, and mixing, sedimentation seems to decisively influence the water quality. The outgoing water has a significantly lower color and turbidity. With good approximation the concentration ratios between the tunnel outlet at RWTP and the tunnel entrance were 0.86:1 for A254 and 0.46:1 for turbidity with a time delay of about 8 days. Short-term changes in Kafjorden were a function of local inflow to the basin, mainly from a small tributary, Murån, and internal basin processes such as sedimentation and resuspension. Although relationships between water quality and the governing factors are complex, it was shown that during certain periods Murån contributes to the color in Kafjorden. Resuspension due to oscillatory bottom currents from wind-induced waves is also a color increasing mechanism during high winds from certain directions. More generally the relationship between color, surface runoff, and EVI (taken as a proxy for organic material available for transport).

To further support the observations from the study and its results, more detailed measurement should be conducted. For the tunnel and pipeline, the temporal resolution is sufficient, but the spatial information is restricted to the in- and outflow section. To confirm the significance of sedimentation and its variation in time and space, more measurement locations are needed along the tunnel. Also, to resolve the importance of groundwater flow for water quality, more measurements are required that focus on this flow. Reliable and robust estimates of the contribution from both Murån and resuspension to the color in Kafjorden need more measurements of flow and a variety of other environmental parameters.

Beyond the specific case in this study, general conclusions can be drawn about typical variation patterns in different water quality parameters and their governing processes, as well as how to use this knowledge to optimize treatment processes. Focus was on color, but advance information about other parameters may also help to streamline water treatment. Overall, better monitoring of the source water will enhance the possibility for advanced analysis, which will eventually expand knowledge of possible interactions and help in optimizing treatment processes for color and other pollutants removal.

Greatly appreciated is the financial support from Sydvatten AB and Sweden Water Research AB which made this study possible. We are thankful for the conduction of long-term monitoring and the provision of data by Lagans Vattenråd and the Swedish Institute for Hydrology and Meteorology. We thank H. Hosseini, K. M. Persson, O. Söderman, J. Rankinen, T. Persson and A.-T. Klante for discussion and feedback as well as help in making specific data available to us. Also, the critical comments and suggestions by the anonymous reviewers were very helpful.

Data cannot be made publicly available; readers should contact the corresponding author for details.

The authors declare there is no conflict.