The type of MP used in this study was polyethylene (PE). The density of the PE is between 0.88 and 0.96 (Rillig et al., 2017; Hüffer et al., 2019; O’Connor et al., 2019). The PE grains between 200–500 µm were set apart by using the sieves. The grain sizes of this PE were measured by Zeiss Primostar optical microscopy and ZEN 2.6 Primostar software. The grain size statistics and the histogram of the used PE particles are given in Table 1 and Figure 1, respectively. Approximately, 7.5% of the grains were smaller and 3.5% of the grains were higher than the used sieve gap sizes which were a minimum of 200 µm and a maximum of 500 µm. Scanning electron microscopy (SEM) was used to identify the shape, and the surface morphology of the PE particles was irregular and rough, so they can be called fragments (Figure 2). If the pore size and the grain size of the saturated porous medium are too small for the movement of the plastic particles, straining will occur. According to the studies on pore straining, the diameter of the particles should not exceed 5–10% of the porous medium grain diameter to avoid the pore straining (DeNovio et al., 2004).

In order to test different grain sizes and porosities, large silica balls, small silica balls, sand, and coarse gravel materials were used in the experiments (Figure 3). For the pore size and grain size measurements of large silica, small silica, and sand, the materials were mixed with resin in a plastic holder and left to dry. Then the dried samples were cut horizontally near the bottom. The bottom sides of the samples were examined with an optical microscope to measure the pore and grain sizes. The particle size of the coarse gravel was directly measured with a tape measure. For the pore size measurements of the coarse gravel, pore sizes were directly measured on the column filled with the material. Between 40 and 90 measurements were performed for each material to obtain reliable pore/grain size distribution of each used material.

Depending on the magnification of the optical microscopy, image resolution changes. The smallest measurable pixel sizes for the images of the large silica, small silica, and sand were 213 μm, 206 μm, and 211 μm, respectively. Approximately, 58% of the pores were higher than 800 μm and 97% of the pores were higher than 400 µm for the large silica. 93% of the pores were higher than 300 µm for the small silica. Some 78.6% of the pores were higher than 300 µm for the sand. Coarse gravel pore size varied between 1 and 20 mm (Table 2 and Supplementary Figure S1).

According to the grain size distributions of the porous medium, the most dominant grain size intervals for the large silica, small silica, sand, and coarse gravel are 1700–2000 μm, 1000–1100 μm, 400–800 μm, and 23–28 mm, respectively (Table 3 and Supplementary Figure S2).

The fluorescent tracer experiments were performed by vertically placed columns under a steady flow rate within the saturated medium to investigate the hydrodynamic parameters of the porous medium materials used in this study. The experimental setup is given in Figure 4. The X indicates the distance from the injection point to the top of the material. L₁ indicates the height of the material. L₂ is for the length of the small outlet tube, which indicates the distance from the top of the material to the water sampling point. R₁ and R₂ are the diameter of the column and the diameter of the small outlet tube, respectively. The experimental system was cleaned after each experiment. The parameters of the experimental setup are given in Supplementary Table S1.

The experiment started with calculating the density of the materials and their total porosities to determine the mean transit time of the water in the column. A volumetric flask was used for the density measurements. The material of known volume was filled with water and the volume of the used water was measured, then the ratio between these volumes was calculated for the total porosity measurements. The mean transit time was calculated by Equations 1, 2 in order to decide the start time and the intervals of the water sampling/fluorescent measurement (Fetter, 2014). T = X u (1) where T, mean transit time (time); X, distance from injection point (distance); and u, mean velocity (distance/time). u = Q A*n (2) where u, mean velocity (distance/time); Q, flowrate (volume/time); A, cross-sectional area (area); and n, total porosity (dimensionless, ratio).

AquaFluor^® Handheld Fluorometer (Turner Designs) was used for the fluorescent measurements of the samples. At the beginning of each experiment day, the fluorometer was re-calibrated and the fluorescein (uranine) injection solutions were prepared. For the calibrations, 300 μg/L solutions were used and equaled 300 units on the fluorometer. The injection solution was prepared at 0.2 g/L concentration. Minimum three injections/experiments were done for each material. The experiments were implemented by different flow rates and volumes of the injected tracer. The flow rates (Q) and injected volumes (V_inj) are given in Table 4.

The longitudinal dispersivities and the kinematic porosities of the porous medium materials were calculated using the inverse solution method. One-dimensional advection–dispersion equation (ADE) was used to have analytic breakthrough curves (BTC) (concentration/time). Then these analytically calculated curves were fitted to the BTCs that were created by the fluorescent measurements of the outlet samples. The fitting process was applied by Solver Add-in of Excel application. This application allowed to modify the dispersivity, kinematic porosity, and mean velocity to get the best fit by an iterative process. The dispersivity and kinematic porosity were achieved by calculating the average of the results of the multiple experiments for each material and condition.

Under steady-state flow and saturated conditions in a homogeneous porous medium, the transport of suspended particles can be described by the advection and dispersion equation (Eq. 3) (Bear and Cheng, 2010). ∂ C ∂ t = D L ∂ 2 C ∂ x 2 − u ∂ C ∂ x (3) where ∂C/∂t, solute concentration change over time (concentration/time); C, solute concentration (mass/volume); D_L, longitudinal dispersion coefficient (area/time); X, distance (distance); u, average linear water velocity (distance/time).

The longitudinal dispersion coefficient is a linear function of the average velocity and dispersivity of the porous medium. It was defined as under saturated conditions by Bear (1988) and is given in Eq. 4. D L = D e +   α L v (4) where D_L, longitudinal dispersion coefficient of the medium (area/time); D_e, effective dispersion coefficient of the medium (area/time); α_L, longitudinal dispersivity (distance); and v, average flow velocity (distance/time).

The velocity of the groundwater ranges between 1 cm/day to 1 m/day. The effect of diffusion in the transport of solutes is negligible for the flow in a saturated porous medium for most of the studies (Raaijmakers, 2018). The tracer injection was implemented instantaneously. The length of the small outlet tube from the top of the material to the outlet where the samples were taken was measured and considered as only advective transport was valid there.

The recovered mass was calculated with the assumption of zero-mass loss. The recovered mass was used in the analytical solution of the 1D advection–dispersion equation as initially injected mass. Eq. 5 was used to calculate the recovered mass (Goldscheider and Drew, 2007). m R = Q ∫ 0 ∞ c ( t ) ∂ t = Q × ∑ [ c ( t ) × Δ t ] , (5) where Q, flow rate (volume/time); c(t), tracer concentration by time (mass/volume); Δt, timestep (time).

The ADE for instantaneous injection under the assumptions of the 1D, semi-infinite, isotropic, and homogeneous porous medium was used to create analytical BTCs (Kreft and Zuber, 1978) (Eq. 6). C i ( x , t ) =   M 0  x A n  u x 4 π D x t 3   exp ( − ( x − u x t ) 2 4 D x t ) , (6) where C_i, concentration at the sampling point (mass/volume); M₀, recovered mass (mass); x, distance between the sampling point and the injection point (distance); u, mean velocity (distance/time); D, dispersion coefficient (area/time); t, time (time); A, cross-sectional area (area); n, total porosity (dimensionless, ratio).

A new Plexiglass column and experimental setup were built for this study. The length and the inner diameter (R₂) of the column were 100 and 8 cm, respectively (Figure 5). The column was filled with the materials and then saturated. The PE particles were injected for a finite duration with a constant flow rate. The porous mediums were assumed as isotropic and homogenous. When the column is filled, X indicates the length of the porous medium. L₁ is the distance between the sampling point and the outlet of the column. L₂ is the distance between the PE suspension reservoir and the inlet of the column. Only advective transport was considered for L₁ and L₂.

The density of the PE particles is lower than the distilled water. A suspension was needed for the homogenous injection of the MP. Ethanol and distilled water were used to create a lower density liquid. The ethanol and distilled water ratio were determined as 56 and 44%, respectively. The density of this liquid mixture was around 0.87 g/cm³. PE particles were successfully suspended within this liquid matrix. The column and the circulation system were built considering to minimize the evaporation of the ethanol.

The column was filled with the material and the ethanol–distilled water mix (Reservoir 1) was circulated for a while to saturate the porous medium. After the steady-state flow conditions were achieved, the valve of Reservoir 1 was closed, and the valve of Reservoir 2 was opened at the same time to start the finite-duration MP injection. The experimental system was cleaned after each experiment.

The PE concentration of the injection suspension injected volume and the mean flow rate for each experiment are given in Table 5.

TD-700 Laboratory Fluorometer (Turner Design) with a blue channel was used to measure the plastic concentrations of the samples. A fluorometer was used because it was determined that the PE particles have a certain level of reflectance for the blue channel. Yet, this was an indirect measurement. So, the inversion of the fluorometer measurements was performed. The inversion equations were found to convert the device measurements to the plastic concentrations. Six PE suspensions of known concentrations were prepared. These suspensions were measured by the fluorometer. The PE concentrations were plotted versus device measurements. A trendline with an equation was defined (Figure 6). These equations were used to convert the device measurements to the PE concentrations. The samples were collected in tubes with a cap to avoid the evaporation of the ethanol. PE concentrations of the samples were measured after the end of the experiment.

The same principles with the solution method of the fluorescent tracer experiment were used also for the tests with PE (Eqs 3–5) (Bear, 1988; Goldscheider and Drew, 2007; Bear and Cheng, 2010). Only the longitudinal dispersion was considered. The porous media were assumed as isotropic and homogeneous. The 1D ADE equation for finite-duration injection by Ogata and Banks (1961) with the assumptions of the 1D, semi-infinite, isotropic, and homogeneous porous medium was used (Eq. 7). A similar procedure–the fluorescent tracer tests solution for the inverse modeling by using Solver Add-in on Excel–was applied for the PE tests with the additional modifiable parameters (concentration injected and injection time). C =   C 0 2   [ erfc  ( L −   v x t 2 D L t ) + exp ( v x L D L )  erfc ( L + v x t 2 D L t ) ] , (7) where C, solute concentration (mass/volume); C₀, initial solute concentration (mass/volume); L, flow path length (distance); v_x, average linear water flow velocity (distance/time); t, time since the injection of the solute (time); D_L, longitudinal dispersion coefficient (area/time).

The fitted experimental and the analytical fluorescent tracer concentration BTCs for each medium and each column test were given in Supplementary Figure S3. The recovered mass was calculated with Eq. 5 and was used as an injected mass while doing analytical solution (Table 6). The results of the fluorescent tracer column tests, kinematic porosity (n_k), and dispersivity (α) values for each medium are given in Table 7.

Concentration measurements were initiated at the beginning of the first sample collection. However, the sampling was faster than the measurements. This caused a waiting time for most of the samples up to 10 min. It is expected that a certain level of fluorescent activity might have diminished during this period depending on the sensitivity of the fluorescent to the light.

The injected mass and the recovered mass were accepted as equal, based on the assumption that there was not any mass lost. The recovered mass was calculated using the sample measurements. The above-mentioned error in the sample measurements might have been valid for the calculations made by using recovered masses also.

The PE particles did not transport up to the outlet of the column in the materials of small silica and sand. The MPs settled down at the very beginning of the inlet of the column and they clogged the inlet. The MPs were transported successfully through the column in the materials of large silica and coarse gravel.

For experiment P1 for large silica, the pumping flow rate was 70 cm³/min, the injection duration was 16 min, the injection concentration was 0.024 g/cm³, and the injection volume was 1120 cm³ which was lower than the pore volume (1996.04 cm³). The injected PE particles clogged the column entrance. There were not any PE particles in the samples and on the filter. The PE particles did not transport throughout the column.

In experiment P2 for large silica, the pumping flow rate was 104.8 cm³/min, the injection duration was 5 min, and the injection concentration was reduced from 0.024 g/cm³ to 0.01 g/cm³, and the injection volume was 524 cm³. The PE particles were transported up to the first 10 cm from the bottom of the column and then stopped there. There were a very low number of particles found on the filter and in the samples.

In experiment P3 for the large silica, the pumping flow rate was 206.28 cm³/min, the injection duration was 10 min, the injection concentration was 0.01 g/cm³ same as experiment P2, and the injection volume was 2062.8 cm³. The PE particles were transported up to the first 10 cm from the bottom of the column and then stopped there same as experiment P2. The amounts of the observed PE particles on the filter and in the samples were higher than those for experiment P2. Yet, the increase in the pumping flow rate did not create a significant change in the transport of the PE particles.

The injection concentration could not be decreased more because the lower values than 0.01 g/cm³ cannot be detected by the fluorometer used in this study. When the pore sizes of the porous media are evaluated, it is seen that the distribution graphs are not symmetrical. Therefore, when talking about pore sizes, they should be evaluated by considering maximum, minimum, and standard deviation values instead of average values. According to the pore size measurements of the large silica, it was seen that 97% of the pore sizes were higher than 800 µm while the PE particle size was between 200 and 500 µm. It is determined that the PE particle size and the pore size of the large silica were not suitable for the transport of the particles. The pore size measurement method might have caused a certain level of error. Also, there were not enough amount of PE particles smaller than 200–500 µm to perform further experiments. When column experiments are evaluated, Dong et al. (2021) and Chu et al. (2019) used a combination of 0.25–5 µm diameter MPs for 0.35–0.45 mm porous media and 1 µm diameter MPs with 0.25–0.30 mm grain size, respectively. Considering the study of Hou et al. (2020) as the experiment using larger size MPs, it was seen that MPs with 40–48 µm diameter and two different porous media with 1–2 and 2–4 mm grain sizes were used. In this study, MPs with a larger diameter were used. 200 and 500 μm MP were tested with 1.7–2.0 mm and 23–27 mm porous media.

Considering the previous results, the experiments were continued with the coarse gravel, which had larger particle and pore sizes than the other materials. In experiment P1 for coarse gravel, the pumping flow rate was 40.59 cm³/min, the injection duration was 50 min, the injection concentration was 0.02 g/cm³, and the injection volume was 2029.5 cm³ which was slightly higher than the pore volume (1910.09 cm³). The particle movement in the column was very slow. The particles mostly stopped at the bottom half of the column. The pumping flow rate was too low, thus the mean velocity of the injection liquid in the column was too slow. The mean velocity of the injection liquid in the column was 2.02 cm/min. It can be said that the 2.02 cm/min and lower velocities are not enough for the transport of the PE particles.

In experiment P2 for coarse gravel, the pumping flow rate was increased to 153.84 cm³/min, the injection duration was 15 min, the injection concentration was 0.02 g/cm³, and the injection volume was 2307.6 cm³ which was higher than the pore volume (1910.09 cm³). The PE particles were successfully transported throughout the experimental system.

Eq. 7 was used for the analytical solution of the P2 experiment to estimate the transport parameters. The analytical BTC was fitted to the experimental BTC by the manipulation of the injected concentration (C₀), injection duration (T_inj), dispersivity (α), and mean velocity (u). These four parameters had to be changed by the Solver Add-in in order to get a reasonable fit. The fitted BTC are shown in Figure 7. The results of the parameters are given in Table 8.

The PE experiments resulted in 3.02 cm for the dispersivity of the coarse gravel while the fluorescent tracer experiments resulted in 2.58 cm for the same medium. These values are in the same decimal value, and they are quite similar in terms of assessing the dispersivity of a medium.

The ethanol–water-PE suspension was prepared for the injection in order to avoid the sedimentation of the PE particles in reservoir 2 and the column. The experimental setup was built as close as possible to minimize the evaporation of the ethanol to sustain the stable suspension of the injection. Yet, the experimental setup was open to the atmosphere during the sampling. This might have caused a certain level of evaporation of the ethanol which might change the suspension stability.

For the P2 experiment, the recovered mass was calculated by Equation 5 using the measured concentration of the samples. The recovered mass was 26.7 g. According to the Solver Add-in fitting, the injected PE amount was calculated as 23.1 g. It was known that the injected/pumped PE mass was 46.2 g. The coherence between the recovered mass and the Solver calculated mass indicates the consistency between the results. The difference between the known injected mass (46.2 g) and the calculated masses (26.7 and 23.1 g) points out that 58–50% of the PE particles must be retained in the porous medium. Also, the relative method for the measurements of the PE samples might have a certain level of error margin on the PE mass calculations even though the recession coefficient of the calibration was found to be 0.97.

The ADE (Eq. 6) employed for the PE tracer test results is often used for the investigation of the transport of very small particles like fluorescents. In the present study, the PE particle sizes were much bigger than those. The BTCs don’t fully/well overlapped/fitted may lead to a discussion of the appropriateness of the ADE for the PE transport.

The transport of the PE particles might have also been affected by the zeta potential (electric charge) between both PE–PE and PE-media. If the particle and porous media have the same electric charge (positive–positive or negative–negative), that would make the PE particles easier to move through the porous medium. The change in ionic strength and pH and the presence of humic acid in the environment are the factors that directly affect the zeta potential (Chu et al., 2019; Dong et al., 2021). The zeta potential of the particles and porous media was not examined within the scope of this research.