Quantification of Microplastics in Treated Drinking Water Using µ-FT-IR Spectroscopy: A Case Study from Northeast Italy

Abstract

Microplastics spread through the environment in various ways. Inland waters are an ideal medium for their dispersal, as they collect pollutants from various sources and transport them over long distances. From there, microplastics can enter the marine environment, break down into smaller particles or end up in drinking water treatment plants. However, the fate, transport and potential health effects of microplastics after ingestion of drinking water and water in food are not yet fully understood. It is therefore necessary to evaluate the quantification and identification of microplastics in drinking water by analysing real samples in order to assess the potential impact on human health. To this end, microplastic contamination in 32 treated drinking water samples from a surface water treatment plant in north-eastern Italy were analysed using micro-Fourier transform infrared spectroscopy (µ-FT-IR). The results indicated low levels of contamination, with all the samples containing less than 170 microplastics per litre, which is in line with European drinking water levels. Polyolefins with size 50–500 µm, such as polypropylene and polyethylene, were the predominant polymers detected (50.2%), while surprisingly polyethylene terephthalate was scarcely present (0.1%, size range 10–50 µm). Statistical analysis revealed a significant negative correlation between microplastic concentration and sampling volume, with larger volumes yielding fewer particles. This inconsistency likely results from the lack of bottle rinsing when only a fraction of the sampling volume is filtered. It was also found that rinsing the sampling bottles with ethanol alone prior to analysis was sufficient to ensure accurate quantification. These results highlight the challenges in standardising the detection of microplastics in drinking water and underline the need for optimised sampling protocols.

1. Introduction

Plastics have become ubiquitous in modern society owing to their low cost, durability, and versatility. Global production has increased exponentially since the 1950s, surpassing 390 million tonnes in 2021 [1] and is projected to grow at a compound annual rate of 4.2% between 2024 and 2030 [2]. Despite their extensive use, the end-of-life management of plastics remains inefficient: in 2019, only 15% of plastic waste was collected for recycling and 9% was effectively recycled, while approximately half was landfilled and nearly one-fifth incinerated. An estimated 22% of plastic waste was mismanaged, contributing substantially to environmental leakage [3]. As a consequence, approximately 9.5 million tonnes of plastic enter aquatic ecosystems annually [4], with significant impacts on biodiversity, ecosystem functionality, and socio-economic activities such as tourism [5]. Most of the global plastic pollution in the oceans is attributed to countries in South Asia, North America, and China [6].

Owing to their resistance to biodegradation, plastics fragment into progressively smaller particles, commonly referred to as microplastics (MPs; <5 mm) [7,8]. The European Chemicals Agency (ECHA) defines MPs as solid polymer-containing particles, with or without additives, that are persistent and capable of existing as small solid particles in the environment and resist biodegradation [9]. This category encompasses numerous synthetic polymers (e.g., polyethylene PE, polypropylene PP, polystyrene PS, polyamide PA, polyethylene terephthalate PET, poly vinyl chloride PVC, polyacrylonitrile PAN, polymethylmethacrylate PMA, polyurethane PU) in particle sizes ranging from 100 nm to 5 mm and fibres from 300 nm up to 15 mm in length and a length-to-diameter ratio > 3. MPs can originate from physical, chemical, photochemical, or biological degradation processes and occur in various morphologies, including fragments, fibres, films, foams, and spheres [10]. Natural polymers that have not been chemically modified are excluded because they are (bio)degradable or have a water solubility > 2 g/L [11].

MPs are now widely recognised as emerging contaminants of concern. They have been detected in seafood, honey, eggs, beer, salt, and, critically, drinking water [12]. Recent studies estimate that humans may ingest up to 5 g of MPs per week [13], primarily through food and water. A global review reported median concentrations of 2.2 × 10³ items/m³ in drinking water sources [14], predominantly with particle sizes > 50 µm [15]. Contamination of drinking water is of paramount importance as it is consumed in large quantities every day and contaminated water could be a major source of dietary intake of MPs [16]. In addition, MPs may contain additives, unreacted monomers, impurities (e.g., residues of catalysts/initiators or derivatives) or other substances such as pigments, lubricants, thickeners, antistatic agents, anti-fogging/clouding agents, plasticizers and flame retardants that can leach into water or absorb toxic chemicals from the environment, posing a toxicological risk to human health [17]. Hazards are also associated with environmental pollutants, such as Persistent Organic Pollutants (POPs) or metals, which can be transported by MPs in the environment and transferred to biota by ingestion. However, there is currently no scientific evidence for higher bioaccumulation of POPs on MPs compared to other types of particles in the environment [9].

Removal efficiency of MPs by drinking water purification facilities (DWPFs), ranges from 66.9% to 100% depending on their size, shape and the kind of physical or chemical treatment used [16].

Concerns regarding MPs contamination have prompted significant regulatory developments. Directive 98/83/EC [18] established the foundational requirements for drinking water quality in the European Union, but subsequent evaluations highlighted the need to update parametric values, strengthen risk-based approaches, improve consumer information, and harmonise approval systems for materials in contact with drinking water. These revisions align with broader environmental commitments, including the United Nations’ Sustainable Development Goal 6 [19], which also highlights the importance of tackling the sources of MPs pollution to prevent further water supply contamination.

In this context, the Commission Delegated Decision (EU) 2024/1441, supplementing Directive (EU) 2020/2184, introduced the first harmonised EU methodology for monitoring MPs in drinking water [20]. The Decision mandates the integration of standardised procedures into national monitoring programmes and requires laboratories to ensure adequate analytical capability. The protocol prescribes four cascade filters (100 µm and 20 µm for both sampling and procedural blanks) and a total sampling volume of 1000 L. Owing to the large volume and the requirement to avoid intermediate storage vessels, the methodology is best implemented on-site at drinking water treatment facilities. The Delegated Decision is prescriptive and legally binding, tailored specifically to drinking water intended for human consumption within the EU regulatory framework. However, as the Decision is recent, only a limited number of plants currently have the infrastructure necessary to comply. Until widespread implementation is achieved, laboratory-based approaches, although not yet harmonised, remain valuable for generating indicative data on MPs contamination.

Among the various protocols mentioned in the literature for the determination of MPs in ‘water intended for human consumption’, the Commission selected infrared (IR) or Raman optical microspectroscopy methods, as well as equivalent variations such as quantum cascade laser infrared (QCL-IR) microscopy. These techniques enable identification of the polymer type in individual particles and additionally provide information on its size and shape.

On a global scale, two recent international standards provide the most updated and harmonised technical frameworks for microplastic monitoring in water. ISO 16094:2025 [21] establishes analytical methodologies for the detection and identification of microplastics in waters with low suspended-solid content, such as drinking water, groundwater, and other clean matrices. The standard covers particles in the 1 µm to 5 mm size range and specifies the use of vibrational spectroscopy techniques (µ-FTIR or Raman microscopy) for polymer identification and particle quantification, together with guidance on contamination control and quality assurance. Complementing this, ISO 5667-27:2025 [22] provides detailed recommendations for sampling strategies across a wide range of water types, including drinking water, freshwater, marine water, and wastewater. It describes appropriate sampling approaches—such as grab sampling, cascade filtration, and net-based methods—along with protocols for sample handling, blanks, equipment selection, and documentation to minimise contamination and ensure reproducibility. Together, ISO 16094:2025 [21] and ISO 5667-27:2025 [22] offer a coordinated framework for standardised sampling and analysis of microplastics in water, supporting more reliable data generation and improved comparability across studies and monitoring programmes.

In this work, drinking water samples from a surface water treatment plant were sampled and analysed off-site to quantify the contamination with MPs. The treatment plant is located in the north-east of Italy. In this area, river water is used to supplement groundwater in times of high demand, when groundwater alone is not sufficient to meet the needs of the population. The plant can supply up to 30% of drinking water during the peak tourist season. The analytical method is based on micro-Fourier transform infrared spectroscopy (µ-FT-IR) and follows the guidelines and best practices in the literature for the analysis of drinking water and clean water samples [21,22,23]. Research on the management of MPs in drinking water is still in its early stages. Existing pretreatments that can reduce the concentration of MPs in water need to be improved to increase their efficiency [24]. To improve existing techniques, it is necessary to understand the characteristics of MPs found downstream of wastewater treatment plants in terms of composition, number and size.

The aim of the study was therefore to systematically investigate the effects of the relevant procedural steps of the proposed protocol in order to improve the quality and reliability of the results and to help laboratories plan their analysis to avoid false positives and negatives. All data were obtained in automatic µ-FT-IR mode. A statistical analysis of the data obtained was performed to identify statistically significant correlations between the variables considered, such as sampling (season, volume, bottle, etc.), water sample parameters (type of polymer, protein content, number of plastic particles, number of unknown particles, etc.) and analytical operating conditions (filtration system, order of sample analysis).

Furthermore, as wastewater management systems differ in terms of the type of processes (primary, secondary, tertiary and quaternary), more data on the efficiencies of treatments are needed to define the best strategies to tackle microplastic pollution. Finally, the mapping of local and global water pollution by MPs is still an open question. This work provides a picture of the microplastic pollution situation in one of the most densely populated and industrialised areas in Europe.

2. Materials and Methods

2.1. Description of the Drinking Water Treatment Plant

The surface water from the river Sile is fed to the plant via a canal and the treatment plant is located 10 km downstream of the canal, so that in the event of an accidental release of pollutants into the river or canal, there is sufficient time to stop the flow of water to the treatment plant without contaminating the drinking water. A simplified schematic representation of the system can be found in Figure 1. After the largest objects have been removed through a grid, the water is treated with carbon dioxide (CO₂) to maintain the pH value at 7.2–7.4 and pre-oxidised/disinfected with gaseous chlorine dioxide ClO₂. Acidification allows the residual aluminium concentration (which is used in the subsequent flocculation process) to be kept within the legal limits. The suspended solids in the feed water are removed by flocculation with aluminium polychloride (PAC) as an agglomerating agent (from 25 to 40 mg/L, depending on the turbidity of the input water). After flocculation, the water undergoes an intermediate stage of disinfection. Activated carbon in powder form is then added to adsorb soluble impurities and the water is channelled into the sedimentation tanks. To remove particles that are too small to settle in the basin, the water is passed through a package of sand filters. Finally, adsorption takes place on granular activated carbon, which also removes excess chlorine, and the water is finally drinkable.

The flow rate of drinking water produced in the plant is in the range 0.1 m³/s–1.0 m³/s. All pipelines in the plant are made of high-density PE (HDPE), PVC or cast iron.

2.2. Standard Protocol for MPs Quantification

The proposed method is composed of the following sequential steps: (I) Sampling: a collection of water samples from the target sources to ensure representativeness and minimise contamination; (II) Pre-treatment: preparation of water samples to remove potential interferences; (III) Filtration: passing the pre-treated water through filters to separate and concentrate MPs for analysis; (IV) µ-FT-IR analysis: using µ-FT-IR to analyse the samples, allowing identification and counting of MPs based on their spectral characteristics in fully automatic mode.

A sketch of the sampling and analytical procedure is reported in Figure 2.

2.2.1. Water Sampling

The water samples were taken at two sampling points located near each other and close to the outlet of the water plant: at a fountain (Figure 3a), where water flows continuously at a low flow rate, and at an overflow point on the pipeline, where water flows at a high flow rate (Figure 3b). In the latter case, the spill valve is normally closed and is opened for a few minutes before sampling to remove any disturbances caused by the accumulation of stagnant water and particles in the sampling pipe, thereby ensuring a representative sample. A minimum sampling volume of 1 litre (L) was chosen, as this is representative of daily human consumption according to various scientific publications [24] and corresponds to WHO guidelines [25] and the recommendations of the ISO standards for water contaminants [22,26,27] and sampling volumes. At each sampling session, three water samples were collected consecutively and stored in three different bottles. They were filtered and analysed independently.

The water samples are then stored in dark glass bottles with capacities of 1 L or 2.5 L, equipped with a liner and screw cap. The sampling bottles are either new or, if previously used, undergo a cleaning procedure. Both used and new bottles were washed in a dishwasher (mod. PG8504, Miele Italia s.r.l, Appiano S.Michele (BZ), Italy) at 85 °C to remove organic material. The bottles were washed and rinsed with anhydrous ethanol. The effectiveness of this cleaning process was compared with treatment using nitric acid, which is commonly used for the degradation of organic material. Finally, the bottles were rinsed with demineralised water and anhydrous ethanol. After sampling, the bottles were capped, sealed with aluminium foil, stored in refrigerated PS (Styrofoam) boxes, and sent to the laboratory. Refrigeration prevents the proliferation of organisms and the degradation of MPs during the period from sampling to analysis. The refrigerated environment helps preserve the integrity of the samples and ensures their stability until subsequent laboratory analysis.

2.2.2. Water Sample Pre-Treatment

All samples were pretreated to minimise the presence of organic and inorganic components as much as possible. 10 mL/L 30% v/v H₂O₂ (hydrogen peroxide) was added to each sample immediately after sampling. Each sample was then homogenised and left to rest for at least 24 h at room temperature to allow the oxidative process to complete. According to Schrank et al. [28] and Al-Azzawi et al. [29], hydrogen peroxide ensures the removal of organic matter such as protein without degrading the most common classes of plastics: PA, PO, PET, PS, PU. In this way, all MPs contained in the samples remain unaffected by the pre-treatment and a much cleaner filter is obtained.

2.2.3. Filtration

1 L samples or a 1 L sub-sample taken from a 2.5 L bottle were filtered and analysed. For each water sample replicate, one filter was obtained, meaning that three filters were analysed for every sampling session. The filtration process was carried out in a vacuum filtration unit in which all parts were made of glass. To ensure the correct assembly of the unit, a leak test was carried out with a few drops of ethanol. This procedure allows the correct alignment and tightness of the components. A silicon filter (SmartMembranes GmbH, Halle (Saale), Germany) with a diameter of 13 mm, a pore size of 5–6 µm and a pitch of 12 µm was used for the procedure. The choice was dictated by the analysis technique, in which the transmission spectra of MPs are recorded directly on the filter in a µ-FT-IR device. The silicon filters were cleaned before using them to remove possible contaminations with the following procedure: (I) Sonication three times in ultra-pure water, for 10 min each time; (II) Cleaning in 10 mL of pure ethanol (99.5%). The cavitation generated by sonic waves at 45 kHz creates turbulence, effectively removing contamination from the packaging where the filters are stored. Finally, the absence of any residual fibres/particles is checked under a light microscope (Leica DMLP, Leica Microsystems GmbH, Wetzlar, Germany) in reflection mode with magnifications of 50×–100×.

After assembling the filtration system, all samples were filtered into aliquots of about 100–150 mL. By adding small aliquots of the sample into the measuring funnel each time, only the lower part of the device is contaminated, which simplifies cleaning during the recovery phase. Before filling each aliquot, the sample bottle was shaken vigorously to minimise the adhesion of the MPs to the inner surfaces and to facilitate suspension. Each empty bottle was rinsed three times with 50 mL of absolute anhydrous ethanol by shaking vigorously. After each rinse, the contents of the bottle were poured into the funnel to collect the remaining MPs. In contrast, the 2.5 L bottle was not rinsed with ethanol due to sub-sampling.

After rinsing the bottle, the particles were collected from the funnel. Finally, ethanol was carefully dripped into the inside of the funnel using a glass pipette to ensure that the particles were collected on the silicon filter at the bottom of the funnel.

Before proceeding with the next sample, the surface of the funnel was treated with nitric acid and then rinsed with grade 3 demineralised water [30] followed by ethanol to ensure a clean and uncontaminated environment. Considering the low estimated particle counts in the samples, even low levels of contamination could significantly affect the quantification of MPs. Considering the various sources of MPs, such as clothing, laboratory materials and atmospheric debris, as highlighted by Dris et al. [31], a careful procedure to maintain a clean working environment free from MPs and other forms of contamination was strictly followed.

2.2.4. µ-FT-IR Analysis

As described in the literature [20,32,33], the coupling of FT-IR with optical microscopy allows particles down to 10 µm in size to be analysed. The analyses were performed with the aim of developing an automated method for the assessment of drinking water samples that minimises the need for operator intervention and makes it into a routine laboratory analysis. Prior to µ-FT-IR analysis, the filters were kept overnight in a vacuum oven at 60 °C to remove moisture from the sample, as water causes interference in the resulting absorption spectrum. The vibrational-rotational peaks of water are identified as two broad spectral regions between 4000 and 3000 cm⁻¹ and 2300 and 1300 cm⁻¹ [34]. All filters were placed in small Petri dishes covered with aluminium foil during the drying phase to avoid contamination.

The filter surface was analysed using a liquid nitrogen MCT detector µ-FT-IR Nicolet™ iN™10 infrared microscope (Thermo Fisher Scientific, Madison, WI, USA). The surface was analysed entirely without sub-sampling, following the total model of ISO 16094:2025 [21]. The first step was to create a map of the entire filter surface, which was then processed with the proprietary image analysis to detect the particles on the surface. All collected particles were selected using the Particle Wizard section of the Thermo Scientific™ OMNIC™ Picta™ Software, version 1.8.0.240. 24 co-scans (aperture 100 µm × 100 µm, spectral range 4000 cm⁻¹–650 cm⁻¹, spectral resolution 4 cm⁻¹) were acquired in transmission mode so that the entire cross-section of the filter and the MPs deposited on it were analysed. The image analysis software subtracts the signal from the silicon filter, which records the background spectrum, and only displays the spectra of the MPs. The number of scans was set to 24 to minimise the time required for analysis without compromising spectral quality.

All particles on the filters were analysed. The number of particles detected ranged from a minimum of 260 to a maximum of 1828, with an average of 994 ± 376 particles.

In addition, contrast and brightness were adjusted to exclude semi-transparent spots and holes of the filter to improve the appearance of the video image. The spectra were compared with those in the standard libraries provided by the manufacturer, using an algorithm built into the software to distinguish the plastic particles from the non-plastic particles such as cellulose and protein material. The software assigns each spectrum a Hit Quality Index (HQI), which ranges from 0 to 1 and represents the correlation between the collected spectra and the spectra in the libraries. A higher HQI means a more reliable identification. Each spectrum was compared to specific polymer reference libraries such as: (I) HR Polymer additive and plasticizers; (II) HR Sprouse Polymer additives; (III) HR Sprouse Polymers by transmission; (IV) MPs Reflection; (V) Organic by Raman Sample Library Polymers; (VI) Bibliotheque Particules; (VII) WizardPoly supplied by Termofisher; (VIII) Fibre standard, supplied by CNR-STIIMA, which was created using reference standard fibre samples.

The minimum threshold was set to a match result ≥ 70% and only optimally identified particles were counted. Thanks to the imaging function of the software, the particle sizes (length and width) could be determined with a limit of detection (LOD) of 10 µm.

2.3. Recovery Standard

Before proceeding with the analysis of the drinking water samples, a verification test of the entire filtration process was carried out. A standard suspension containing a known number of MPs was used to verify that MPs were correctly recovered during the individual steps. This ensures that the operator acquires proper manual dexterity and precision during the sample preparation steps that precede the analysis in the μ-FT-IR. The preparation of the four suspensions was performed according to the standard procedure ISO 1833-4:2023 [11,35]. Each standard suspension of fibrous PET MPs was prepared starting from a thread of 256 filaments (2970 dtex) supplied by Aquafil S.p.A (Arco (Trento), Italy)Before preparing the standard suspensions, the number of filaments was verified according to UNI EN 1049-2:1996 [36] and they were divided into groups of ten (Figure 4).

To prepare the PET standard suspensions, the synthetic thread was cut using a microtome with a 200 µm slot, following the UNI 5423:1964 [37] and UNI EN ISO 137:2015 [38] standards and wool fibres were used to secure it. After that, all the fibres were treated with 10 mL of demineralised water and 5–7 mL of sodium hypochlorite (NaClO), which completely degrades the wool. Afterwards, the suspension was transferred to a larger Erlenmeyer flask. To ensure complete fibre recovery, the flask was rinsed and filled with 50 mL aliquots of ultrapure water. until the volume of 960 mL was reached. Moreover, the flask was rinsed and filled with 10 mL of a 1:1 solution of demineralised water and ethanol to collect any remaining fibres adhering to the walls, which were then transferred to the Erlenmeyer flask. The procedure was repeated four times, and the resulting suspensions were filtered. All the fibres collected on the filters were counted using aNicolet™ iN™10 infrared microscope (Thermo Fisher Scientific, Madison, WI, USA) (Figure 5), mapping the entire filter area Before μ-FT-IR analysis the silicon filters with the collected fibres were dried in an oven at 60 °C for a night in order to completely eliminate the water absorbed by the fibres that can interfere with the absorption bands of the polymers and reduce spectral identification. All the suspensions were filtered by using two silicon filters. The sum of the collected PET microfilaments, the average value, the standard deviation and recovery rate were calculated. All these data are shown in

Table 1. The recovery rate for PET standard suspension, average, standard deviations, theoretical filaments value and percentage.

Volume (L)	Replicate	PET (Theoretical Value Number of Fibres: 256)
		N° Fibres	% Recovered Fibres
1	R1	225	88
1	R2	207	80
1	R3	211	82
1	R4	239	93
	Average	221	85.8
	St.dev	13	5.9

. Recovery rates of more than 80% were achieved.

2.4. Quality Control and Analysis (QA/QC)

A careful procedure was followed to keep the working environment clean from MPs and other forms of contamination. The laboratory bench was cleaned with paper soaked in pure ethanol before analysing each sample. The funnel was then treated with nitric acid and rinsed with grade III demineralised water to remove any residue remaining after washing. At least three rinses with pure ethanol were carried out before the funnel was used. The tweezers used to handle the filters were washed with pure ethanol. To maintain cleanliness, all containers, the filtration device and the tools used for filtration were wrapped or covered with aluminium foil to reduce the risk of deposition of particles or fibres. Filtration was performed with bare hands (to avoid contamination by gloves), which were thoroughly washed to remove possible contamination. The lab coat worn during the analysis was made of cotton. In addition, cotton clothing was worn under the lab coat during both sampling and laboratory analysis. As described in the literature, it is advisable to use a controlled airflow to maximise air purity and minimise airborne contamination, e.g., in a “clean air laboratory” or in a laminar flow cabinet [23]. Despite all these procedures, for the analysis, it is crucial to quantify the possible contamination by analysing the blanks (procedural and reagent). This step is essential to ensure the integrity of the samples and prevent the presence of any contaminants. This is necessary to verify that the MPs counted in the analysis are derived from the sample itself and not from external sources of contamination [33]. Both procedural and reagent blanks were used to assess contamination from the environment and the procedure as well as from the substances used. In this procedure, the reagent blank was ethanol: 10 blanks containing 150 mL of ethanol were filtered through a silicon filter and subsequently analysed. The results obtained did not indicate the presence of plastic materials. This procedure was carried out subsequent to a preliminary investigation of the ethanol contained in high density polyethylene bottles (HDPE). As a few particles of PE were collected and identified, likely released from the plastic bottle itself, glass bottles were used instead. Concerning the procedural blank, a silicon filter was subjected to the same cleaning process (sonication) as the operative filters. Then, the filter was used for filtration of the reagent blanks and subsequently analysed in a micro-FT-IR to quantify any contamination with MPs. Each time the filtration equipment was cleaned, a procedural blank was performed. Therefore, one procedural blank was conducted per filtration session. No contamination was detected in the procedural blanks.

2.5. Graphic Analysis and Statistics

Excel was used to process the data obtained from the µ-FT-IR analysis, while the programming language R [39] was used to create the diagrams and perform the statistical analyses. The main packages used were: (I) readxl [40]; (II) ggplot2 [41]; (III) mass [42]. A non-parametric regression model was used to find correlations between variables. Non-parametric models involve finding some balance between fitting the observed sample of data (model fit) and “smoothing” the function estimate (model parsimony). Typically, this balance is determined using some form of cross-validation, which attempts to find a function estimate that does well for predicting new data. As a result, non-parametric regression models can be useful for discovering relationships between variables, as well as for developing generalizable prediction rules. Both linear and robust regression were applied in this study. Sampling parameters (variables), such as the season (the samples were taken from February 2023 to January 2024), the volume of water sampled, the type and pre-treatment of the bottle used, were analysed to assess the impact on the result of the analyses performed.

3. Results

Recognising the importance of previous research, our analytical procedure was developed taking into account the key aspects highlighted by [43,44] for drinking water and the most recent international standards [21,22]. To ensure the robustness and reliability of the results obtained regarding the presence and characterisation of MPs in drinking water, it is imperative to take these indications into account during the analytical procedure by using a statistical method to assess the above-mentioned aspects.

In this study, 32 drinking water samples from two different sampling points of the same plant were analysed to identify and quantify the MPs present. All samples analysed had a turbidity value of less than 1 NTU (Nephelometric Turbidity Unit). The samples were collected in glass containers and then filtered offline. The automatic identification of the particles collected on the filters was facilitated by µ-FT-IR software, (Particle Wizard section of the Thermo Scientific™ OMNIC™ Picta™). However, in certain cases, manual controls were performed or analytical approaches were used to break down the signal originating from several substances and identify the individual components.

The microplastic particles identified were predominantly fragments with a circular morphology and a size distribution that mainly comprised three size classes: 500 µm to 100 µm (27%), 100 µm to 50 µm (70%) and 50 µm to 10 µm (3%). It is noteworthy that particles smaller than 50 µm were the least common size class within the microplastic distribution. As the process of chlorination has been shown to lead to fragmentation of MPs [45], medium-sized MPs were the most prevalent in the analysed samples.

The polymers identified in the samples included polypropylene (PP), polyethylene (PE), ethylene-propylene-diene monomer (EPDM), polyamide (PA), polytetrafluoroethylene (PTFE), polyurethane (PU), polystyrene (PS) and polyethylene terephthalate (PET), which correspond to the most common synthetic polymers found in drinking water. All spectra of the identified polymers are listed in Figure A1 (Appendix A).

Figure 6 shows the number of MPs, categorised by polymer type, detected in drinking water samples by µ-FT-IR analysis. The LOD (Limit of Detection) of the method used allowed the identification of MPs particles with a minimum size of 10 µm.

The most common MPs in the samples were from the PO group (PP, PE, EPDM). PP and PE are plastics produced worldwide and are commonly used in food packaging materials, construction, agriculture, piping, containers, toys, automotive, etc. [1]. Moreover, their presence is consistent with the origin of the analysed water, which comes from surface water where these polymers often accumulate due to human activities [46]. Additionally, although EPDM is defined as a polyolefin rubber, it is not included in the list of the most common MPs, according to ISO 16094:2025 [21]. However, the use of this polymer is increasing dramatically in the synthetic rubber sector as it has excellent ozone and ageing resistance compared to natural rubber and its properties make it suitable for sealing applications [47]. Recently, in drinking water this polymer has increased and probably this could be due to high use of artificial turf (AT). It is a surfacing material that simulates natural grass by using mainly fibres, which are widespread pollutants in river and sea surface waters. Moreover, other plastic components of AT are composed of PP (draining sheet and piping), a filling made of styrene-butadiene rubber (SBR) and pad layer made of Polyurethane resin (PUR). In this case study, as the inlet data of the DWTP are missing, the evaluation can only be considered a hypothesis.

The small amount of PET found indicates effective removal by coagulation treatment with PAC, despite its high presence and worldwide distribution, typically in packaging and beverage bottles, and in the textile industry as a fibre material [48]. In particular, fibres are more easily absorbed on the surface of the flock and because of this, the coagulation-sedimentation process might have had a removal efficiency of 50.7–60.6% as reported by other authors [48]. This behaviour could explain the low content of PET in particulate form, which is probably due to fragmentation and the absence of the polymer in fibre form.

This same trend was observed in drinking samples taken from other DWTP outlets [49] where the presence of PET and PTFE is absent in comparison with PP, polyolefin material. In detail, the data show that the number of PO particles is 680 times larger (680 vs. 1) than PET, in line with other studies [50], where it was observed that coagulation with PAC was more efficient in the removal of PET than PO because of PET greatest binding affinity with PAC (three times larger) due to the C=O group in its structure. Moreover, the effectiveness of the coagulation-flocculation process was found to depend on water pH, with the removal of PE being higher when the pH value is lower than 6, as alu-based salts form smaller floc with higher surface area. The low efficiency of PO removal by coagulation-flocculation was also reported by Ma et al. [51]. It was observed that removal efficiency depends on the size of the PO particles, ranging from only 8.3% for 500 μm particles and falling to 1% for larger particles (2–5 mm).

With regard to the presence of PA and PTFE, these compounds were detected in lower quantities compared to PO, yet higher than PET (112 and 34 particles, respectively, across all samples). This discrepancy can be attributed to the extensive utilisation of PA in diverse industrial sectors, including the textile industry (specifically in hosiery and sportswear), the fishing one, the automotive one, the electrical and electronics one (for mechanical parts such as gears and bearings), and packaging one. The existence of PA in rivers is widely documented in the literature, both in sediment and water [52,53]. The presence of PA found in this study is consistent with the surface origin of the water analysed. Conversely, PTFE has numerous applications due to its properties such as high temperature resistance, non-reactivity, low coefficient of friction, and electrical insulation capability. Consequently, it is primarily used as a coating for pipes, pumps, impellers, and bearings, as well as in the production of gaskets and O-rings. Therefore, it is possible that some MPs detected may originate from the contamination of pipes, valves, and gaskets inside the water treatment plant. The micro-FTIR analysis of the 32 water samples detected and identified 1 polyurethane (PU) particle and 4 polystyrene (PS) ones. Moreover, carbonate salt was detected in all the samples, at a level of about 6%.

The results found show significantly higher quantities of MPs characterised by a density lower than that of water, confirming the difficulties in removing such polymers as reported by [54]. In particular, polymers with higher density are more effectively removed by means of sedimentation tanks due to their greater sedimentation capacity. Conversely, less dense polymers, such as polyolefins, present greater challenges in water treatment processes because they tend to remain suspended. The distribution of the number of MPs/L, with a subdivision into classes of amplitude 5 MPs/L, shows a non-normal distribution of the acquired data as shown in Figure 7. The highest percentage (81%) is found between 0 and 35 MPs/L detected per drinking water sample, with a peak in the range between 0 and 5 MPs/L in accordance with levels of MPs data from European studies that are in the lower range concentration, as indicated in Directive 2020/2184 of European Parliament and Council [55] for drinking (0.0001 to 440 MPs/L). Pivokonsky [56] found that raw water MPs concentrations (≥1 μm) in two DWTPs near the Uchlava River in the Czech Republic varied widely, with mean concentrations of 23 and 1296 MPs/L. However, in general, in Europe the quantity of MPs (0.000–0.6 particles) is lower than Asia (0.7–440 MPs/L) or America (12–316 MPs/L) and depending on volume and type of sampling. The data obtained highlight that when the source water in DWTP is a contribution of surface water and groundwater the level of MPs contamination is more variable depending on various factors as reported in the literature [57].

4. Discussion

The drinking samples do not show variability depending on the seasonality. Probably, this effect as shown by other authors is more observable in samples taken from the inlet of the plant in comparison with outlets, because of the effect of weather conditions, which affect the quality of the water depending on pollutant concentration in aquatic environments and influence their physicochemical properties [49].

A statistical analysis was performed to identify possible correlations between the number of MPs detected and the type of polymers, considering variables related to the drinking water samples and the methods of analysis. After the non-normality of the data had been checked using the Shapiro–Wilk test (p-value = 2.20 × 10⁻¹⁶), a non-parametric regression test was selected to determine statistically significant relationships between the factors under consideration. When examining linear regression and calculating studentised residuals, i.e., the dimensionless ratios resulting from dividing a residual by an estimate of its standard deviation, six samples were found to exceed the usual threshold of 3. These outliers, which represent the natural variability in the number of MPs from the drinking water samples, do not result from the method itself, as evidenced by the recovery rates and blanks, and must therefore be kept in the analysis. Therefore, a robust regression was used to minimise the influence of outliers and deviations from a normal data distribution [58]. To confirm the accuracy of the method, a comparison was also made between the standard residual errors (S) of the linear regression (S = 10.1) and the robust regression (S = 0.8). A lower value of S indicates a more accurate regression model that provides more accurate predictions.

For the statistical analysis, the data were indexed, i.e., numbers were assigned in place of the levels present in the variables under consideration so that the data could be analysed and presented in numerical form. With regard to sampling, the following variables were considered: (I) sampling period, four levels from February 2023 to January 2024, covering all seasons; (II) sampling location, two possible sampling points, from a constantly running fountain or from a tap along the WTP; (III) volume of the bottle used for sampling, two levels (1.0 L or 2.5 L); (IV) type of bottle, new or previously used; (V) bottle treatment, two pre-treatments of the bottles before sampling, ethanol only or nitric acid followed by ethanol. In addition, two variables were considered in the context of the MP analysis: (I) number of unidentified particles, total number of particles that could not be identified divided into four levels (0–500, 501–1000, 1001–1500 and 1501–2000); (II) total number of identified proteins divided into four levels (0–250, 251–500, 501–750 and 751–1000). Fragments of proteins and cellulose are commonly found in surface water samples; however, in this study, no cellulose was detected, although some protein contamination was observed.

In this study, an alpha value of 0.05 is used as the cut-off value for significance: at a p-value ≤ 0.05, there are statistically significant differences in the observations made; at a p-value > 0.05, the differences are not significant. The robust regression in relation to the number of MPs found shows only two statistically significant correlations for the number of MPs: with the polymer type (p = 3.28 × 10⁻¹⁸) and the sampling volume (p = 9.88 × 10⁻³). The frequency of PP (median = 3.5, interquartile range (IQR) = 11.0, sum = 305), PE (median = 1.0, IQR = 2.0, sum = 113) and EPDM (median = 3.0, IQR = 6.3, sum = 262)) is in line with the literature [59], which indicates high levels of polyolefins in surface waters, such as those entering the studied facility, due to human activities in lakes and rivers. Polyolefins, whose density is lower than that of water, tend to remain buoyant and therefore appear to be less readily captured by the treatment processes.

The second significant difference that emerged was between the number of particles found and the sampling volume, i.e., the larger the sampling volume (2.5 L), the fewer MPs. This correlation was attributed to the method of analysis, in which only 1 L of water was filtered from a 2.5 L container, and the bottle was not rinsed to collect adhering MPs, unlike the 1 L samples.

The other parameters analysed showed no significant correlation with the number of MPs in the sample. The p-values derived from the robust regression between the number of MPs detected and the following variables are in detail: (I) season (p-value = 0.24); (II) sampling location (p-value = 0.31); (III) bottle type (p-value = 0.28); (IV) bottle treatment (p-value = 0.31); (V) number of unidentified particles (p-value = 0.70); (VI) number of proteins (p-value = 0.86). The presence of protein particles is due to their incomplete oxidative degradation. However, this does not invalidate the analysis, as the assumed operating conditions, such as pre-drying of the filter, help to reduce water signals that could interfere with the characteristic peaks of the synthetic polymers, allowing them to be distinguished from other organic materials present.

Contrary to other findings [49], the statistics have proven that the season has no influence on the presence of MPs, just as the sampling point has no influence on the number of MPs present, so sampling is possible all year round, either from a continuous fountain or from a normally closed spill valve at an overflow point on the pipeline that is left open for a few minutes before collection. Furthermore, there is no difference between new and used bottles, provided both are treated appropriately prior to sampling. The bottle treatment can be performed effectively with ethanol alone, so no nitric acid is needed, reducing both the cost and the environmental footprint of the process.

Furthermore, the count of MPs was not affected by the amount of protein present in the analysed sample, as all water samples were treated with hydrogen peroxide, nor by the number of unknown particles where the algorithm identified a noisy spectrum, such as the effect of baseline variation.

5. Conclusions

In this study, the presence of microplastic particles in drinking water samples was analysed using μ-FTIR spectroscopy. Thirty-two samples from a water treatment plant that treats the surface water of a river in northeast Italy were analysed. The concentration of microplastics varied considerably between the samples, ranging from 1 to 170 MPs/L, which are in line with the reported levels of microplastics in drinking water [20]. The types of polymers contained also varied considerably. The most frequently detected polymers were polyolefins, which is consistent with the occurrence of these polymers in surface waters influenced by anthropogenic activities. Polyamide and polytetrafluoroethylene were also detected, which is probably due to contamination by the incoming water or possible components in the water treatment plant.

The presence of polyethylene terephthalate, despite its extensive use, indicates that it was effectively removed by coagulation treatment with aluminium polychloride. Polystyrene and polyurethane were also detected in trace amounts. To summarise, the results indicate significantly higher quantities of microplastics, whose density is lower than that of water, which underlines the difficulty in removing these polymers. The statistical approach used includes the assessment of the non-normality of the data distribution and the implementation of a regression model, deciding between linear and robust regression depending on the impact of outliers on the final assessment. This methodological framework allowed the identification of variables, both in sampling and analysis, which have an impact on the quantification of microplastics.

The main conclusions of this study include the need to filter and analyse all the sampled water, the feasibility of using bottles that have already been used for sampling and the possibility of treating these bottles exclusively with ethanol before use.