Removal of dissolved organic matter in road runoff with sludge-based filters from the drinking water treatment plant

Dissolved organic matter (DOM) comprises a class of organic mixtures characterized by a broad range of molecular weights, complex composition and structure, and physical heterogeneity (Wegley Kelly et al. 2022). DOM is commonly found in natural and human-impacted waters, such as surface water and wastewater (Chen et al. 2022). It influences the formation and dissolution of minerals, the stability of the colloidal state, and the cycling of trace metals, as well as the migration of organic and inorganic pollutants (Pan et al. 2023). During the preparation of drinking water, DOM can react with disinfectants, including chlorine-based ones, producing large amounts of highly toxic disinfection by-products such as polychlorinated biphenyls (PCBs), trihalomethanes (THMs), and other substances with carcinogenic, mutagenic, and teratogenic effects (Garrido Reyes et al. 2021; Wang et al. 2021). Additionally, DOM can interact with non-polar organic pollutants, such as pesticides and PCBs, via its hydrophobic benzene ring structure and lipophilic nature, resulting in the formation of composite pollutants that pose considerable adverse effects on the environment and human health (Artigas et al. 2020; Agboola & Benson 2021). Yuan et al. (2019) observed that DOM in runoff could form complexes with carbamazepine (CBZ), increasing the persistence and biotoxicity of CBZ in the aquatic environment. Therefore, although DOM is not inherently toxic, it is recognized as a precursor to toxic substances in water systems and should be removed during treatment processes.

Due to the randomness and scattered discharge of runoff, traditional methods, including coagulation, oxidation, and membrane filtration (Nguyen et al. 2020), are not suitable for the removal of DOM in runoff. In contrast, the adsorption method offers unparalleled advantages in the treatment of DOM in runoff. Cost-effective absorbent with efficient DOM removal capabilities should be developed to serve as filler in green infrastructures (e.g. wetlands and rain gardens) for removing DOM from runoff. This is valuable for effectively managing runoff pollution.

Due to the high demand for potable water, precipitated sludge is produced annually in water treatment facilities, reaching 1.5 billion m³. Generally, precipitated sludge is derived from the formation of hydrolysis products or hydroxide precipitation following the application of coagulants such as poly-aluminium chloride and polymerized ferric chloride to address colloidal particles, DOM, and other suspended materials in raw water. Hence, the water treatment plant coagulant–precipitated sludge contains a large number of active constituents, including Al₂O₃, SiO₂, and Fe₂O₃ that can effectively remove severely charged DOM through ion exchange or display strong adsorption abilities against various types of pollutants (Mouratib et al. 2020). As a result, the sludge can be employed as a filling material in runoff treatment facilities to reduce DOM from the runoff and eventually eliminate various pollutants, such as filters, constructed wetlands, and bioretention systems. This method makes it possible to remove DOM and other impurities from runoff while also making sludge from water treatment facilities more resource-efficient. So it opens a new way to utilize water treatment sludge as an effective adsorption medium for DOM removal and thus presents a relatively economical and eco-friendly alternative to the traditional filtration media.

In this investigation, water treatment facility sludge was employed as filler in a filter system to assess its capacity for the removal of DOM in the influent and effluent of the filter using analytical techniques such as gel permeation chromatography (GPC), UV–Vis spectroscopy, and excitation-emission matrix (EEM) fluorescence spectroscopy. This analysis aimed to identify the primary removal components of DOM in the runoff.

After one year of operation, six rainfall events were analyzed to evaluate the removal efficiency of runoff DOM in the filters. No significant difference in the characteristics of DOM in influent and effluent was found among them (Du et al. 2022). Subsequently, influent and effluent samples from the filters on June 19, 2022, were analyzed in this study. On that day, there was a rainfall event with a total depth of 20.9 mm, concentrated between 20:00 and 22:00, peaking at 21:00. The specific precipitation pattern is illustrated in Supplementary material, Figure S1. The selected rainfall event in June was chosen due to its occurrence after an unusually long dry period from March through May, lasting nearly three months with minimal precipitation. During this dry season, pollutants, including organic matter, dust, and oils, accumulated on urban surfaces. When the first significant rainfall occurred in early summer, these pollutants were flushed into the runoff, leading to higher pollutant loads.

To assess the filter's effectiveness, influent and effluent samples were collected at two key points: at the groove on the wall separating the sedimentation area and the filtration area and at the drainage pipe within the sampling chamber. Sampling was conducted at 10-min intervals, with each sample comprising 30 mL, starting once water flow was observed. This process was maintained consistently over a 2-h period to capture the temporal dynamics of the runoff event.

After collection, the water samples were immediately transported to the laboratory. To mitigate any concentration discrepancies arising from the sample settling, it was imperative to filter the samples using 0.22 μm membranes promptly. Following filtration, the samples were preserved in brown glass bottles at 4 °C. A multi-N/C 3100 TOC analyzer (Jena, Germany) assessed the dissolved organic carbon (DOC) concentration in the influent and effluent samples. To fractionate DOM, 20 mL of influent collected at different times were combined, as were 20 mL of effluent from the two filters at different sampling points.

The sludge sourced from the Third Water Purification Plant in Beijing, China, underwent a two-step process. Initially, it was dried in an oven at 105 °C until a constant weight was achieved. Subsequently, the dried sludge was ground into particles of approximately 0.15 mm and ignited in a muffle furnace (Furnace 5000, Thermolyne, IA, USA) at 500 °C for 4 h. The cooled sludge and the purchased quartz sand were introduced into the filters. The composition of the sludge was analyzed using an X-ray fluorescence spectrometer (ZSX primus, Rigaku, Japan), as presented in Table 1.

The molecular weight distribution of DOM in both mixed influent and mixed effluent filtered by a 0.22 μm membrane from both filters was analyzed through GPC (Bright LC, Lappertec, China). This GPC setup utilized a differential detector (S2020 A, Schambeck SFD GmbH, Germany) and an SB-803 HQ gel-filtration column (Shodex OHpak, Tokyo, Japan). The mobile phase used was a 0.05 mol·L⁻¹ sodium dihydrogen phosphate solution, and the flow rate was set at 2.0 mL·min⁻¹ with the column temperature maintained at 35 °C.

Following the resin isolation method (Pi et al. 2021), Amberlite XAD-8 resin (20–60 mesh), MSC-H cationic resin, and IRA-958 anionic resin were employed to categorize the DOM in the influent and effluent into six chemical fractions: hydrophilic acids (HiA), hydrophilic bases (HiB), hydrophilic neutrals (HiN), hydrophobic acids (HoA), hydrophobic bases (HoB), and hydrophobic neutrals (HoN). The initial step involved passing the filtered samples through the Amberlite XAD-8 resin (20–60 mesh). Backwashing the XAD-8 resin with 0.10 M HCl led to the separation of the HoB. The effluent from XAD-8 was acidified to a pH of 2.0 using 6 M HCl, after which it was sequentially passed through XAD-8 resin, MSC resin (Dowex Marathon MSC hydrogen, 20–50 mesh), and IRA-958 resin (free base). The HiN were exclusively present in the effluent. Subsequently, the XAD-8, MSC, and IRA-958 resins were back-eluted using 0.1 M NaOH to produce HoA, HiB, and HiA. Following elution or recovery, the pH of various DOM fractions was promptly readjusted to match the original pH of the source water. The XAD-8 resin was then subjected to a 15-h drying process at ambient temperature and Soxhlet-extracted with methanol for 8 h to obtain the HoN. The organic carbon mass balance was calculated after measuring the DOC content of various DOM components. This study determined that the organic carbon mass balance within the resin isolation system was 100 ± 10%.

The EEMs for both the influent and effluent, which had been filtered through a 0.22 μm membrane, were determined using a fluorescence spectrofluorometer (F-7000 FL spectrophotometer, Hitachi, Japan) with a 150-W Xenon arc lamp as the light source. The excitation (Ex) and emission (Em) wavelengths were adjusted to cover the 200–450 and 280–550 nm range, respectively. Both the excitation and emission monochromators had a 5 nm slit width, and the scan speed was set at 1,200 nm min⁻¹. To reduce the impact of Rayleigh light scattering, a 290 nm emission cutoff filter was employed. EEM Ex/Em contour maps were generated using Origin 8.5 software (Origin Lab Corp, MA, USA).

Synchronous fluorescence spectra for both the influent and effluent samples were obtained by scanning over the wavelength range of 200–550 nm with constant 30-nm steps. The scanning speed was configured at 240 nm·s⁻¹.

The 0.22 μm membrane-filtered influent and effluent samples were loaded into a 10 cm quartz window cuvette (210PC UV–Visible spectrophotometer, Düsseldorf, Germany). Wavelength scanning was carried out across the entire 200–800 nm range.

Filter B demonstrated superior DOC removal efficiency, with rates ranging from 70 to 80%. In contrast, filter A exhibited a much lower DOC removal rate, consistently in the 20–30% range throughout the entire rainfall event. Compared to traditional sand and gravel filters, which generally achieved DOC removal rates of 30–40%, and activated carbon filters, which can reach 70–85% under optimal conditions (Markiewicz et al. 2020), the sludge-based filter B showed relatively high performance, making it a promising alternative.

Unlike the inert quartz sand medium with a small specific surface area, the sludge from drinking water treatment plants exhibited an amorphous structure with well-defined pores, as observed in the scanning electron microscopy (SEM) and Brunauer-Emmett-Teller (BET) analysis (see Supplementary material, Figures S2 and S3). Before adsorption, the sludge particles displayed a rough, porous surface with larger pore spaces. The BET test results showed an initial specific surface area of 33.97 m²·g⁻¹ and a pore volume of 0.0006 cm³·g⁻¹, indicating abundant sites for adsorption. After adsorption, the sludge surface exhibited reduced porosity, with smaller particles visible within the pores. This change in surface morphology suggested that DOM molecules entered the sludge pores and occupied the available sites, effectively reducing pore volume and surface area. These observations confirmed that the sludge's porous structure played a crucial role in DOM removal by facilitating adsorption within the pores. Additionally, the sludge was characterized by its high aluminum and iron content, amorphous structure, and numerous reactive groups on its surface (see Supplementary material, Figure S4). These unique properties of the sludge facilitated the adsorption and subsequent removal of DOM from runoff, resulting in a higher removal rate in filter B.

In line with the separation principles of gel column chromatography, distinct peak positions indicate varying fractions of DOM with different molecular weights. Typically, high-molecular-weight DOM substances struggle to enter the micropores of the gel particles due to their large diameter. Consequently, they traverse the column more rapidly, requiring shorter elution times. In contrast, low-molecular-weight DOM substances can readily diffuse into the spaces between the gel particles and may even penetrate the micropores within them. As a result, their migration is comparatively slower, necessitating longer elution times. Additionally, the intensity of a gel chromatography peak reflects the DOM, with higher intensity indicating greater DOM content (Gao et al. 2022).

The influent sample displayed three distinct gel chromatography peaks at 6.06, 6.47, and 8.01 min, labeled P1, P2, and P3, respectively. In comparison to peak P3, the intensities of peaks P1 and P2 were higher, suggesting that the runoff DOM contained a higher concentration of larger molecular weight substances such as humic acid, polysaccharides, proteins, and similar compounds (Gbadegesin et al. 2022). Filtration resulted in a reduction in the intensities of peaks P1 and P2, making them somewhat more prominent compared to that of P3, particularly in filter B. This suggested that filter B was most effective at removing larger molecular-sized substances from runoff DOM. High-molecular-weight DOM contained abundant functional groups such as carboxyl, phenolic hydroxyl, and carbonyl, and was highly susceptible to adsorption through systems including hydroxyl substitution, surface complexation, and electrostatic attraction (Ren et al. 2018). Due to its amorphous porous structure, the sludge from drinking water plants showed remarkable adsorptive abilities. It also led to the enhancement of ligand exchange with high-molecular-weight DOM due to the presence of aluminum, iron, and calcium on the surface of the sludge, ultimately resulting in the immobilization of stable compounds (Zhao et al. 2023). Thus, sludge filter media in filter B were effective in removing more DOM from runoff compared to the inert quartz sand media used in filter A, especially high-molecular-weight DOM substances. It was reported that stronger complexation occurs between DOM substances with lower molecular weights and heavy metals (Dinu 2023). Therefore, the effluent from filter B is considered safer than that from filter A concerning the potential ecosystem toxicity of heavy metals.

As shown in Figure 4, in filter B, the removal of the DOM percentages was more successful than that in filter A. In filter B, the hydrophobic DOM percentages, especially HoA and HoN, had a higher treatment rate than the hydrophilic percentages, surpassing 90%. In general, the hydrophobic portion of DOM contained various aromatic carbons, phenolic structures, and compounds with conjugated double bonds (Qiu et al. 2019).

Aromatic carbon components in DOM may engage in hydrophobic interactions and π–π interactions, as well as form hydrogen bonds with the carbon-containing functional groups in sludge, particularly carbon–carbon double bonds (Ahmed et al. 2018), thereby facilitating the removal of hydrophobic organic compounds in filter B. Additionally, hydrophilic functional groups in DOM like –COOH and –OH play a significant role in removing hydrophilic components from runoff DOM through π–π interactions, hydrogen bond formation, and electrical interactions between functional groups in sludge. Also, hydrophobic organic compounds that are sludge-adsorbed serve as carbon sources for microorganisms and aid in the further breakdown of hydrophilic DOM fractions into smaller molecular proteins and hydrophobic organic acids (Huang et al. 2018). Given that hydrophilic DOM fractions were known to strongly complex heavy metals reported (Lian et al. 2022), filter B, which was filled with sludge from drinking water treatment plants, was advantageous for the simultaneous removal of heavy metals.

The peak intensities of DOM in the effluent spectra were all reduced compared to the influent spectra, indicating an increase in fulvic-like, humic acid-like, and protein-like substances in the effluent. The intensities of the four peaks in the three-dimensional fluorescence spectrum of DOM in filter B's effluent had significantly lower levels than those in filter A, which further demonstrated that filter B had a more potent DOM treatment process. The fulvic-like and humic acid-like substances, regarded as non-biodegradable organics (Yin et al. 2018), could be ingested more easily as a result of their adsorption onto the filter medium. Compared to filter A, filter B's sludge likely had many more functional groups, which helped remove substances similar to fulvic acid and humic acid more quickly. Besides, humic and fulvic acid-like substances were typically hydrophobic because of their longer chain fatty acid constituents and higher levels of aromatic carbon (Qiu et al. 2019). Thus, filter B's sludge contained a higher concentration of hydrophobic organic compounds. This finding corresponds to Section 3.3's conclusions. The fluorescence spectra intensity of protein substances, marked as peaks T3 and T4, which represented biodegradable DOM (Shen et al. 2024), was decreased in filter B's effluent compared to filter A. This suggested that a larger community of microorganisms resided in filter B due to the adsorption of hydrophobic organics by the sludge, giving carbon sources for bacterial growth. It also implied that the removal of more hydrophilic organics in filter B was due to a higher occurrence of microorganisms. The decay of aromatic groups and the collapse of macromolecules may be responsible for the change in position peak T1 toward a shorter wavelength in filter B's effluent. Functional groups may be eliminated along with the decline of aromatic rings and conjugated bonds within the chain structure (Ma et al. 2018).

Table 3 presents the UV–Vis absorbance spectra of DOM, including parameters such as SUVA₂₅₄, SUVA₂₆₀, A₂₅₃/A₂₀₃, and A₃₀₀/A₄₀₀. SUVA₂₅₄ indicates the humification level of DOM, where a higher SUVA₂₅₄ value corresponds to a more highly humified DOM with high aromatic and macromolecular weight (Pearce et al. 2023). A₃₀₀/A₄₀₀ is used to assess the degree of aromaticity and molecular weight distribution of DOM (Tan et al. 2022). A higher A₃₀₀/A₄₀₀ value corresponds to a lower degree of aromaticity, smaller molecular weight, and higher fulvic acid content in DOM. It is noteworthy that the SUVA₂₅₄ value in the effluent of filter A was higher than that in filter B, while the A₃₀₀/A₄₀₀ value for the DOM in the effluent of filter B was significantly higher than that of filter A. This variation can be attributed to more DOM with high aromatic and macromolecular weight remaining in the runoff during quartz sand filtration, resulting in a relatively higher degree of humification. In contrast, DOM in the effluent of filter B, which contained sludge, showed less aromaticity and mainly consisted of smaller fulvic acid molecules. This aligns with the results discussed in Sections 3.2 and 3.4.

SUVA₂₆₀ typically reflects DOM's presence of hydrophobic elements. The greater the SUVA₂₆₀ value, the greater the concentration of hydrophobic components (Zhang et al. 2022). In contrast, the value of SUVA₂₆₀ in the influent was larger than that in the effluent, which demonstrated that more hydrophobic organic compounds were contained in the road runoff. In addition, the SUVA₂₆₀ value of filter B effluent was significantly lower than that of filter A, indicating that substantial amounts of hydrophobic components had been removed from runoff DOM after being adsorbed by the sludge in filter B. This result matches the findings, as shown in Figure 4.

A₂₅₃/A₂₀₃ describes the nature of the substituents on the aromatic ring. A high A₂₅₃/A₂₀₃ value may suggest a high content of carbon, carboxyl, ester, hydroxyl, and other similar functional groups, while a low value may indicate that most of the substituents are aliphatic chains (Qiao et al. 2018). The higher A₂₅₃/A₂₀₃ in the effluent from filter A, relative to filter B, indicates a greater presence of macromolecules, larger molecular weights, and more complex aromatic ring substituents in the effluent of filter A. In contrast, the lower A₂₅₃/A₂₀₃ value in the effluent of filter B indicates that aromatic rings were likely heavily substituted by aliphatic chains with smaller molecular weights.

The filter has been operating stably for over two years, demonstrating effective removal of runoff DOM. The utilization of sludge from drinking water treatment plants as a filter medium for an extended period has raised worries about possible secondary contamination. However, despite the high rate of heavy metals such as iron and aluminum, their stability during the application process of the sludge ensures that there is very minimal environmental impact (Nguyen et al. 2022). Muisa et al. (2020) studied constructed wetlands using sludge as a base material and found the minimal release of aluminum into the water, ranging from 0.02 to 0.06 mg·L⁻¹. Ecological risk assessments have shown that sludge has negligible adverse effects on planktonic plants and animals in lake water, likely due to its relatively large specific surface area that promotes microbial growth (Kamaruzaman et al. 2023). Therefore, it can be classified as recyclable waste with low ecological risk. Based on the benefits observed in this study, using sludge as a filler in green infrastructures for runoff DOM treatment is feasible.

The study investigated the removal of DOM from road runoff in Beijing using various filter media. Methods such as resin fractionation, gel permeation chromatography, three-dimensional excitation–emission matrix fluorescence spectroscopy, and UV–Visible spectroscopy were employed to assess the DOM removal efficiency of these filters. The filter filled with sludge from drinking water treatment plants demonstrated exceptional performance in removing runoff DOM, achieving a DOC removal rate of 70–80% throughout the entire rainfall event. This filter effectively targeted macromolecular DOM components, hydrophobic organic compounds, as well as acidic and neutral organic substances from road runoff. Fluorescence spectrum analysis indicated that the sludge from drinking water treatment plants could efficiently adsorb fulvic acid-like and humic acid-like substances present in runoff DOM. Furthermore, it facilitated the proliferation of microorganisms, thereby enhancing the degradation of protein-like substances. Additionally, filtration resulted in decreased aromatization and humification in runoff DOM by removing aromatic compounds and macromolecular substances. The degree of humification and molecular weight of DOM in the effluent notably decreased in the sludge-filled filter due to effective removal processes applied to aromatic compounds and macromolecular substances from road runoff. These findings highlighted the potential of sludge-based filters as a sustainable and effective solution for urban runoff treatment, particularly in areas with high traffic pollution. However, further research is needed to evaluate the long-term performance of these filters, including potential clogging or adsorption saturation and their impact on microbial ecosystems. Pilot studies in traffic-intensive urban areas and different climatic conditions are recommended to optimize the applicability of this technology. Additionally, integrating sludge-based filters into comprehensive urban water management systems alongside existing treatment methods could enhance treatment efficiency and protect downstream aquatic environments.

The authors would like to express their gratitude to EditSprings (https://www.edits prings.cn/) for the expert linguistic services provided.

This work was supported by the National Key R&D Program of China (grant no. 2021YFC3200700), the Pyramid Talent Training Project of the Beijing University of Civil Engineering and Architecture (grant no. JDJQ20200302), and the Project of Construction and Support for high-level Innovative Teams of Beijing Municipal Institutions (grant no. BPHR20220108).

All authors contributed to the study's conception and design. Material preparation, data collection, and analysis were performed by X.D., W.J., and R.J. The first draft of the manuscript was written by X.D. and W.J., and all authors commented on previous versions. All authors read and approved the final manuscript.

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

The authors declare there is no conflict.