Assessment of the physical characteristics and stormwater effluent quality of permeable pavement systems containing recycled materials

This paper evaluates the physical characteristics of two recycled materials and the pollutant removal efficiencies of four 0.2 m2 tanked permeable pavement rigs in the laboratory, that contained either natural aggregates or these recycled materials in the sub-base. The selected recycled materials were Crushed Concrete Aggregates (CCA) and Cement-bounded Expanded Polystyrene beads (C-EPS) whilst the natural aggregates were basalt and quartzite. Natural stormwater runoff was used as influent. Effluent was collected for analysis after 7–10 mins of discharge. Influent and effluent were analysed for pH, Chemical Oxygen Demand (COD), Dissolved Oxygen (DO), Electroconductivity (EC), turbidity, Total Suspended Solids (TSS), Total Dissolved Solids (TDS), Nitrate-Nitrogen (NO3-N), reactive phosphorous ( ) and sulphates ( ). Both CCA and C-EPS had suitable physical properties for use as sub-base materials in PPS. However, C-EPS is recommended for use in pavements with light to no traffic because of its relatively low compressive strength. In terms of pollutant removal efficiencies, significant differences (p < 0.01) were found in pH, EC, TDS, DO, and across all rigs whereas no significant differences (p > 0.05) were found with respect to TSS, turbidity, COD and NO3-N. Effluent from rigs containing CCA and C-EPS saw significant increases in pH, EC and TDS measurements whilst improvements in DO, TSS, turbidity, COD, and were observed. All mean values except pH were, however, within the Maximum Permissible Levels (MPLs) of water pollutants discharged into the environment according to the Trinidad and Tobago Environmental Management Authority (EMA) or the United States Environmental Protection Agency (US EPA). In this regard, the CCA and C-EPS performed satisfactorily as sub-base materials in the permeable pavement rigs. It is noted, however, that further analysis is recommended through leaching tests on the recycled materials.


Introduction
Sustainable Urban Drainage Systems (SUDS), Low Impact Development (LID), Water Sensitive Urban Design (WSUD) or Best Management Practices (BMPs) such as Permeable Pavement Systems (PPS), offer a viable solution to the stormwater management problems experienced within most Small Island Developing States (SIDS) across the Caribbean (Parkinson, 2002). Other SUDS such as detention/retention ponds and wetlands that provide vital stormwater *Corresponding author. Email: John2.Monrose@live.uwe.ac.uk pavement (RAP) to be suitable backfilling materials for stormwater and sewer pipes. RCA and CB have been used for unbound sub-base materials (Cameron et al., 2013;Garach et al., 2015;Poon & Chan, 2006). Blast furnace slag has been found suitable for use in road sub-bases (Nishigaki, 2000). Tatsuoka et al. (2013) reported that well-compacted crushed concrete aggregate (CCA) can be used as backfill material for civil engineering soil structures requiring high stability. Waste Glass (WG) has been used in pavements for several decades. Jamshidi, Kurumisawa, Nawa, and Igarashi (2016) presented a thorough review of the use of WG on the structural performance, durability and sustainability of asphalt, concrete and concrete block pavements. Murugan et al. (2016) used waste tyre crumb rubber to partially replace river sand in concrete block pavers in an attempt at improving the durability and sustainability of the blocks. Jindal and Ransinchung (2018) used RCA, industrial waste (fly ash) and agricultural wastes (rice husk ash and bagasse ash) in concrete pavements. Admixing industrial or agricultural wastes with pozzolanic properties enhances the strength and improves the durability of concrete thus making them as good as conventional concrete (Jindal & Ransinchung, 2018). Martinho et al. (2018) used RCA as substitutes of natural aggregates to produce asphalt mixtures.
Numerous researchers (Bhutta et al., 2013;Çetin, 2015;David, Chandrasegaran, & Nair, 2018;Jang, Ahn, Souri, & Lee, 2015;Khankhaje, Rafieizonooz, Salim, Mirza, & Hussin, 2017;Monrose, Tota-Maharaj, Mwasha, & Hills, 2019;Nishigaki, 2000;Rizvi, Tighe, Henderson, & Norris, 2010;Tavares & Kazmierczak, 2016;Yao, Wang, Liu, Yao, & Wu, 2019) have reported on the incorporation of recycled waste materials or recycled industrial by-products in permeable pavement surfaces, porous concrete in particular. Nishigaki (2000) used blast furnace slag in permeable paving blocks. Khankhaje et al. (2017) compared the effect of using two different sizes of oil palm kernel shell and cockle shell as partial replacement of natural coarse aggregate on the properties of pervious concrete pavement. Khankhaje et al. (2017) reported a decrease in compressive strength of the pervious concrete with increased shell contents but suggested that the values obtained satisfied the requirements for lightly traffic areas such as parking lots and footpaths. Rizvi et al. (2010) evaluated the use of RCA from old curb and gutter, sidewalks and sewers in pervious concrete pavement and found that up to 15% of natural course aggregate could be replaced with RCA without affecting the structural and hydraulic performance of the pervious concrete. Çetin (2015) incorporated recycled household plastics (low density polyethylene [LDPE]) as a supplemental aggregate (1, 3, and 6%) in porous asphalt to produce a permeable plastic pavement. Tavares and Kazmierczak (2016) used RCA from construction and demolition waste in pervious concrete and found that mean 28-day compressive strengths were within the expected range (2.8-28 MPa or 400-4000 PSI) for pervious concrete. David et al. (2018) found similar performance characteristics when comparisons of compressive strength, density and surface infiltration rate of pervious concrete containing RAP and waste tire rubber were made with pervious concrete that was made up of natural aggregate without additions. Monrose et al. (2019) used manufactured lightweight aggregates referred to as Carbon-Negative Aggregates in solid paving blocks and reported that CNA could replace natural course aggregates in these blocks by up to 100%.
Meanwhile, only a handful of studies (Rahman et al., 2015b;Rodriguez-Hernandez, Andrés-Valeri, Ascorbe-Salcedo, & Castro-Fresno, 2016;Sañudo-Fontaneda, Charlesworth, Castro-Fresno, Andres-Valeri, & Rodriguez-Hernandez, 2014) have reported on the utilisation of recycled waste materials as reservoir aggregates in permeable pavements. Rahman, Imteaz, Arulrajah, Disfani, and Horpibulsuk (2015a), during a laboratory study, found RCA and CB to be suitable alternative sub-base material for permeable pavements. In the same study, Rahman et al. (2015b) investigated the hydraulic performance and pollutants removal efficiency of PPS using CB, RCA and RAP in combination with geotextile and reported that the geotechnical and hydraulic properties of these recycled waste materials in the pavement filter layers were consistent or superior to that of typical quarry aggregates. Rodriguez-Hernandez et al. (2016) reported that Recycled Aggregates (RA) from Construction and Demolition Waste (CDW) as sub-base materials in permeable pavements improved the hydrological output of the permeable pavements in terms of attenuation, retained rainfall, peak outflow and time to peak. Sañudo-Fontaneda et al. (2014) used basic oxygen furnace slag as sub-base materials in permeable pavements and reported water quality improvements in terms of Total Suspended Solids, turbidity and Dissolved Oxygen.
Recycled waste materials obtain their physical and chemical properties from their sources, processing methods, and handling techniques which in turn determine their suitability for use in construction with respect to structural (strength and durability) and environmental (leachability) requirements (Inyang, 2003). According to Arulrajah, Piratheepan, Disfani, and Bo (2013) different researchers have found that construction rubble possesses few negative environmental and social effects because leachate release and the presence of heavy metals are within acceptable limits for civil engineering applications.
From the lack of literature sources, it is evident that a research gap exists that involve the assessment of the performance of PPS wherein recycled waste materials are incorporated as filter materials in pavement storage reservoirs. This paper seeks, therefore, to assess the suitability of two recycled waste materials; Crushed Concrete Aggregate (CCA) and a novel material referred to as Cement-bounded Expanded Polystyrene (C-EPS) as filter materials for use in subbase reservoirs of permeable pavements. The evaluation has two objectives. Firstly, the physical characteristics of CCA are compared to that of natural quarried materials of basalt and quartzite aggregates. Secondly, the pollutant removal efficiencies of four 0.2 m 2 permeable pavement rigs containing either natural or recycled sub-base materials are determined and compared. The pollutant removal assessment involved the retrieval of "grab" samples of stormwater runoff from various urban settings across Trinidad. The samples were judiciously applied over the various pavements using a custom-built rainfall simulator. Outflow lasted between 7 and 10 min during each test run after which effluent samples were collected for water quality testing and analysis.
Expanded polystyrene (EPS) is a very low-density foam which contains discrete air voids in a polymer matrix (Cook, 1983). The EPS particles are manufactured by thermal expansion of polystyrene particles saturated with n-pentane. EPS particles are microbiologically stable, float in water, swell in petroleum and demonstrate resistance to mineral acids (with the exception of nitric acid) and bases. They are soluble in organic solvents such as ketones and aromatic hydrocarbons and are thermally stable between 0°C (32°F) and 90°C (194°F). Toxicity of polystyrene depends on the content of the monomer, which can be easily removed by washing with water (Sokolović, Sokolović, & Govedarica, 2009). EPS is usually deemed an environmental menace because it is not bio-degradable. However, it can easily be recycled into a product which can be utilised in practical sustainable construction (Mwasha, Armstrong-Richardson, & Wilson, 2013). EPS is 100% recyclable and thousands of tonnes are recycled each year in developed countries such as the UK (Ngugi, Kaluli, & Abiero-Gariy, 2017). The common characteristics, handling and uses of EPS has been discussed in detail by Mwasha et al. (2013).
The literature surrounding the use of EPS in civil engineering applications is mostly devoted to lightweight EPS concrete wherein the mechanical properties of these materials are assessed. Quarried aggregate is typically replaced with varying percentages of EPS beads depending upon structural requirements (density and strength) (Babu, Babu, & Wee, 2005;Bouvard et al., 2007;Chen & Liu, 2004;Kan & Demirboga, 2009;Kaya & Kar, 2016;Xu, Jiang, Xu, & Li, 2012). EPS blocks have been used worldwide as a type of geofoam lightweight fill material, typically used in embankments where the long-term applied stresses do not exceed circa 100 kPa (2000 lbs/ft 2 ) (Ngugi et al., 2017). EPS has further been used a recycled material in asphalt. Baker, Abendeh, Abu-Salem, and Khedaywi (2016) blended asphalt (bitumen) with recycled packaging waste polystyrene instead of common polymer. They suggested that the modified polystyrene-asphalt binder, had the potential of performing satisfactorily under hot climatic conditions and could be used for lightly-loaded areas such as playgrounds, training areas, parking lots and sidewalks.
Few studies (Orlov, Martynov, & Kunytskiy, 2016;Osuagwu, Nwakwasi, & Nwachukwu, 2018;Schöntag, Pizzolatti, Jangada, de Souza, & Sens, 2015;Sokolović et al., 2009) have reported on the use of EPS as a filter material for water/wastewater treatment and no studies, to the authors' best knowledge, have reported on the use of EPS in permeable pavements. Sokolović et al. (2009) investigated, using a laboratory pilot filter, the efficiency of separation of iron hydroxide flocks from water through an EPS filtration bed. Schöntag et al. (2015) compared the pollutant removal efficiencies of rapid filters made either of sand and anthracite coal or polystyrene (PS) granules. The authors reported that the two filter types achieved similar performances. Additionally, the PS granules did not release detectable amounts of styrene in the water and can therefore be used as filter media. The authors did note, however, that monitoring for long term degradation and the possibility of leaching of styrene into the water was recommended. Orlov et al. (2016) argued that the cost of potable water treatment schemes is largely dependent on the cost of the filtration system and suggested that in Ukraine, filters made up of EPS as opposed to sand could provide savings in capital investment by 40-50%, in operating costs by 30-40% and in electricity by 7-9%. Osuagwu et al. (2018) used EPS as a viable adsorbent for the removal of iron was raw water. They reported 36% iron removal efficiencies within a contact time of 5 mins.

Design and construction of permeable pavement rigs
Permeable pavement rigs were designed in accordance with technical guidance from literature sources (Bentarzi et al., 2016;Drake, Bradford, & Marsalek, 2013;Scholz & Grabowiecki, 2007). Moreover, numerous institutions worldwide have provided general guidance relating to the design and construction of permeable pavements. In the UK, British Standard BS 7533-13 (BSI, 2009a) offers guidance on the design of permeable pavements. Likewise, the Interlocking Concrete Pavement Institute ICPI (2016) has also provided industry guidance for PICP in the U.S.A. and Canada. Similarly, in Australia, the Concrete Masonry Association of Australia (CMAA) has freely made available several design guidance manuals on permeable paving on their website (CMAA, 2016). Each of these standards provides similar recommendations relating to site boundary conditions, pavement structure (layer thickness, aggregate gradations), pavement usage and so on. Nevertheless, the use of permeable pavements as a stormwater management option is scarce across Caribbean SIDS. Consequently, only certain recommendations from these industry guidelines along with previous studies relating to aggregate gradations and pavement layer thicknesses were considered in this research.
Four 450 × 420 × 610 mm (18 × 16.5 × 24 in.) permeable pavement rigs were constructed from 19 mm (3/4 in.) construction plywood as a tanked system. The rigs were made watertight by inserting a 2 mm thick layer of commercially available PVC based pond liner on the inside. Three 12.5 mm (1/2 in.) PVC outflow pipes were inserted through one face of each rig at varying heads of 50, 250 and 480 mm (2, 10 and 19 in.), above the base of the rigs. Three of the four rigs are shown in Figure 1.
The rigs were made up of an 80 mm (3.1 in.) deep I-Paver interlocking concrete block surface, a 50 mm (2.0 in.) deep bedding layer, a geotextile layer, a 100 mm (4.0 in.) deep base course and a 250 mm (10.0 in.) deep sub-base layer. The I-Pavers were supplied by a concrete block manufacturer in Trinidad. Each block paver unit measured 80 × 197 × 143 mm (3.1 × 7.8 × 5.6 in.) and weighed 4.35 kg (9.6 lbs).  The bedding layer consisted of 5 mm (sieve no. 4) ASTM No.8 washed aggregate. The base course layer was formed of 12.5 mm (1/2 in. sieve) ASTM No. 57 washed aggregate. The subbase layer comprised of either 19 mm (3/4 in. sieve) ASTM No.5 aggregates or C-EPS. The varying sub-base materials used during the course of the study are listed in Table 1. CCA was obtained by crushing and sieving aged precast concrete cylinders in the laboratory. A 450 × 420 × 250 mm C-EPS rigid filter block ( Figure 2) was prepared in the laboratory. The materials used were Premium Plus Cement (PPC) which contains approximately 30% pozzolanic material and 70% Ordinary Portland Cement (OPC), tap water and EPS beads. The PPC was sourced from a local manufacturing plant (Trinidad Cement Limited, Claxton Bay, Trinidad, W.I). PPC is manufactured in accordance with international standards (ASTM International, 2017;EN 197-1, 2011). According to the manufacturer, PPC is an eco-friendly cement option whose production has significantly reduced the manufacturer's carbon footprint. The EPS beads were also sourced from a local supplier (Mecalfab Limited, O'meara Industrial Estate Arima, Trinidad, W.I.). Prior to the production of the final C-EPS block, repetitious and numerous trial mixes were done which lead to an optimum mix design. Mixes were evaluated based on hydraulic conductivity and compressive strength characteristics. The optimum mix is presented in Table 2. The mixes ensured that excessive amounts of cement were not utilised. This was necessary to produce a filter material with adequate hydraulic conductivity.
For each mix, three 50 × 70 mm cylindrical samples were prepared in PVC sample holders. These samples were used for hydraulic conductivity (coefficient of permeability) testing using the falling head method (Das, 2010). Plumbers putty was used to seal around the top edge of the cylinder to prevent water from bypassing the sample. The C-EPS sample preparation for   Figure 3. For compressive strength testing, two 100 × 200 mm and six (6) 150 × 300 mm cylinder samples and six (6) 100 mm cube samples were prepared in accordance with BS EN 12390-3:2009 (BSI, 2009b) and ASTM C936 (ASTM International, 2018a) respectively. All samples were secured, de-moulded after 24 h, labelled and cured in water at a standard temperature of 20 ± 1°C for at least 28 days prior to testing. Low compressive strengths were expected, hence a Tinius Olsen tension and compression testing machine (Tinius Olsen TMC, Pennsylvania USA) was used as opposed to a traditional concrete compressive strength testing apparatus. Prepared samples and compressive strength testing are shown in Figure 4. All compressive strength results were less than 1.0 MPa. The chemical compositions of the CCA and C-EPS were determined using an X-ray fluorescence spectrometer (XRF) model SRS 3400 (BRUKER AXS, Inc., Madison, Wisconsin, USA). An ELTRA CS2000 Carbon/Sulphur Determinator (ELTRA GmbH, Haan, Germany) was used to determine the percentage of Carbon and Sulphur present. Compound identification was carried out using a Bruker-Axs X-Ray Diffractometer (XRD) Model D8 Advance (BRUKER AXS, Inc., Madison, Wisconsin, USA). For each material, approximately 500 g of fines passing the No. 12 (1.7 mm) sieve was used for analysis. Results from the XRF are presented in Table 3. CCA contained mostly CaO and SiO 2 at 29% and 62% respectively whereas C-EPS consisted predominantly of CaO, SiO 2 and C at 52%, 16% and 20% respectively.
For all rigs, a nonwoven geotextile layer was placed between the bedding layer and the aggregate base course layer. The properties of the geotextile layer are listed in Table 4. The Minimum Average Roll Value (MARV), as defined in ASTM D4433 (ASTM International, 2018b), is a manufacturing quality control tool used to provide purchasers/users a 97.7% degree of confidence that any samples will exceed reported values. Numerous researchers have reported on the

Geotechnical and physical testing
The geotechnical and physical tests conducted on the pavement aggregates are listed in Table 5. In addition to these tests, particle size distribution tests of the aggregates were performed by sieve analysis in accordance with ASTM C136 (ASTM International, 2014b). To this end, a manual "grab" method of collecting samples of stormwater runoff was used. Grab samples are discrete samples of fixed volume taken to represent local conditions in the flow (Butler & Davies, 2011). Attempts were often made to capture representative "first flush" samples. These "first flush" samples typically contain the largest percentage of the total contaminant loadings especially in small catchment areas with predominantly impervious surfaces and which experience high intensity storms (Geosyntec Consultants and Wright Water Engineers, 2009). These high intensity rainfall events are commonplace in Trinidad. First flush samples were not always obtained due to difficulties in estimating the timing of rainfall events. This made weather forecasting a crucial aspect of the sampling efforts. A total of thirty (30) individual 100 L grab samples were collected from rainfall events between December 2016 and August 2018. Five (5) 20 L polyethylene buckets were used in this regard. Trinidad has an annual rainfall depth of approximately 2000 mm. There is a strong seasonal variation in rainfall whereby 75-80% of rainfall is received during the wet season (June to December). The remaining 20-25% is received during the dry season (January to May) (Monrose & Tota-Maharaj, 2018).
Captured stormwater runoff samples were applied uniformly over the rigs using a purposebuilt rainfall simulator at intensities not exceeding 2.0 L/min. Numerous studies (Alsubih, Arthur, Wright, & Allen, 2017;Nnadi, Newman, Coupe, & Mbanaso, 2015;Sounthararajah et al., 2017) have successfully used rainfall simulation techniques in their research. Outflow (effluent) exited at the outlet of the permeable pavement rigs and was permitted to flow for several min (7-10 min) prior to collection in 300 ml sampling bottles for analysis. Throughout the stormwater application events, the influent (raw stormwater) was continuously stirred to ensure particles remained in suspension. The collected outflow samples were analysed immediately or refrigerated at 4°C to minimise any changes in the physio-chemical properties of the samples prior to analyses. A schematic of the experimental set up is shown in Figure 5. It is made up of a 100 L storage mixing container equipped with a submersible pump and a variable-speed heavy-duty mixer, valves to control flow rates, flow metres to measure flow rates and perforated PVC pipes to distribute the stormwater. Outflow exited the permeable pavement rigs through a 12.5 mm outflow pipe located at the base of each rig.

Water quality analyses
The pollutant removal efficiencies of the four (4) rigs were compared through analysis of various influent and effluent water quality parameters -pH, Chemical Oxygen Demand (COD), Dissolved Oxygen (DO), Electroconductivity (EC), Total Suspended Solids (TSS), Total Dissolved Solids (TDS), Nitrate-Nitrogen (NO 3 -N), reactive phosphorous (PO 3− 4 ), sulphates (SO 2− 4 ) and turbidity (Nephelometric Turbidity Unit [NTU]). All water quality sample analyses were in accordance with the American Public Health Association standard methods for the examination of water and wastewater (APHA, 1998). Table 6 identifies the standard methods used in this paper and the Minimum Detectable Levels (MDL) for each parameter. pH was measured using an Orion 3 Star benchtop metre (Thermo Scientific, Beverly, MA, USA). DO was measured using a YSI 5000 Benchtop DO Meter and probe (YSI, Yellow Springs, OH, USA). A Jenway 4520 Conductivity metre (Jenway, Staffordshire, UK) was used to measure electroconductivity (μS/cm). A Hach Colorimeter (DR/820) (Hach, Loveland, CO, USA) was used to measure NO 3 -N (mg/L), SO 2− 4 (mg/L), PO 3− 4 (mg/L), and turbidity (Nephelometric Turbidity Units [NTU]). The concentration of NO 3 -N (mg/L) was determined by a cadmium reduction method using Hach NitraVer5 Nitrate reagent powder pillows. SO 2− 4 (mg/L) was measured by the SulfaVer 4 Method using Hach SulfaVer 4 sulphate reagent powder pillows. PO 3− 4 (mg/L) was measured by the Amino Acid method using Hach Molybdate and Amino acid reagents. The Absorptometric method was used to measure turbidity. COD was analysed using a Hach DRB200 Reactor block (Hach, Loveland, Colorado, USA) (for sample digestion) and a Hach Spectrometer (DR/5000) TNT 822 COD vial (Hach, Loveland, CO, USA). COD was measured rather than BOD due to ease of measurement and the correlation between the two parameters. A similar approach was taken by Pilon, Tyner, Yoder, and Buchanan (2019) for evaluating the effect of porous concrete on water quality parameters in Alcoa, TN, USA. TDS (mg/L) and TSS (mg/L) were measured using a 0.45 μm Whatman filter paper.

Data analysis
Statistical analyses of the results were performed using the IBM statistical software Statistical Package for the Social Sciences (SPSS) version 20 (IBM, 2011). Descriptive statistics, tests for normality using goodness-of-fit statistics, one-way analysis of variation (ANOVA), Pearson's correlations and the Mann-Whitney U two-independent samples tests were used for the analysis all water quality parameters. A 95% confidence interval was used for all statistical analyses. Boxplots and bar charts were used to examine and present variations and/or similarities in mean pollution concentration results. Mean pollutant removal efficiencies were calculated from Equation (1).
where C in is the inflow (influent) concentration and C out is the outflow (effluent) concentration for individual samples.

Physical and geotechnical characteristics
The physical and geotechnical characteristics of the various construction aggregates used in the rigs are present in Table 7. The specific gravity of the CCA was as expected, lower than that of basalt and quartzite aggregates but greater than the typical requirement of 2.0 kg/m 3 (Rahman et al., 2015b). Conversely, water absorption of the CCA was significantly higher than that of basalt and quartzite aggregates but less than the typical requirement of 10% (Rahman et al., 2015b). Nevertheless, it fell at the low end of the 6-14% range of acceptable water absorption values for recycled materials in civil engineering applications (Poon & Chan, 2006). The CCA performed remarkably well under the LA abrasion and impact tests with results better than the quartzite aggregates and below 50%. It must be noted that a significant portion of the abrasion was due to the disintegration of the cementitious paste (mortar) which surrounded the natural aggregates. The pH of the CCA was notable, being significantly higher than the basalt and quartzite aggregates. This high pH value indicates high alkalinity which can be attributed to the chemical composition of the cementitious paste which is rich in calcium hydroxide Ca(OH) 2 and other compounds (Table 3). This is consistent with results provided by Rahman et al. (2015a) who reported slightly less alkaline values for RCA (10.5) and CB (9.5) and stated that these values were within expected limits. Based on these physical measurements, CCA demonstrated the potential to be a suitable construction material to substitute or add to traditional quarried materials in permeable pavement applications. The particle size distribution (PSD) or gradation curves of the aggregates used in each test rig are shown in Figure 6. The bedding course, base course and sub-base course aggregates fall into ASTM grading classifications No.8,No.57 and No.5 respectively.
With regards to the C-EPS, average hydraulic conductivity values ranged from 1330 to 1764 mm/h which were satisfactory for this application. Compressive strength results were, however, relatively low; below 1.0 MPa. Hence, from a structural view point, C-EPS is recommended for use in permeable pavements non-traffic areas such as building aprons, sidewalks, footpaths, landscapes, pedestrian access and bicycle lanes.

Test for normality
The results of the test for normality using the one-sample Kolmogorov-Smirnov and Shapiro-Wilk goodness-of-fit measures in SPSS are presented in Table 9. Water quality parameters where the null hypothesis is true follow a normal distribution if p > 0.05 and are formatted as bold italics. The distribution required transformation where p < 0.05. Since the bulk of the data violated conditions for data normality, which limited the use of standard parametric testing, non-parametric statistical tests were utilised.

Analysis of variance
The one-way analysis of variance (ANOVA) in SPSS was used to determine whether there were any statistically significant differences between the means of the water quality parameters from at least two of the rigs. The results are presented in Table 10. The null hypothesis suggests that the means of each water quality parameter is the same across all rigs. The null hypothesis was rejected (p < 0.05) for all parameters except COD, NO 3 -N, turbidity and TSS.

Correlation analysis
The bivariate correlation function in SPSS was used to evaluate the strengths of the relationships between the varying pavement rigs (sub-base variations) and the outflow pollutant concentrations. The nonparametric Pearson's correlation coefficient was used to determine the strength of the correlations, if present, whilst the p-values determined the significance of the relationships. The results of the bivariate correlation analysis using Pearson's coefficients are presented in Table 11. A 95% confidence level was used. The results showed that all pavement  200 0.195 0.200 0.533 0.200 0.234 0.200 0.126 0.200   rigs were significantly correlated (p < 0.01) with the water quality parameters in all cases except COD, NO 3 -N, turbidity and TSS (p > 0.05). These results support the ANOVA results presented in Table 10.
To further analyse the significance of the distribution of water quality parameter results between each pair of rigs, Mann-Whitney two-independent samples tests at a confidence level of 95% were used. The results of these analyses are listed in Table 12. It is noticeable that for all parameter distributions, there were no significant differences (p < 0.05) between the distributions for Rigs 1 and 2, both of which contained natural aggregates in their sub-base layer. The distributions are therefore statistically equal between Rigs 1 and 2 for all parameters. There was significant evidence (p < 0.05) to show that the distribution of 60% of the water quality parameters are different between rig groups 1 vs. 3; 1 vs. 4; 2 vs. 3 and 2 vs. 4. In other words, the distribution of all water quality parameters except COD, NO 3 -N, turbidity and TSS were significantly different (p < 0.05) between a rig which contained natural aggregates in the subbase to a rig containing recycled material. This was not surprising given the composition of the recycled materials used (Table 3). Comparisons between Rigs 3 and 4 which contained CCA and C-EPS respectively, showed that the distribution of 50% of the water quality parameters were significantly different (p < 0.05).
Change in pH pH determines the acidity or alkalinity of a water/wastewater sample by measuring the fraction of hydrogen (H + ) and hydroxyl (OH − ) ions present in the sample (Tota-Maharaj, 2010). Figure 7 shows box and whiskers plots for influent and effluent pH values for the rigs. Influent values ranged from 6.7 to 10.4 with a mean of 8.1 ± 1.1 (Table 8). Notable differences in mean effluent values were observed between the rigs with natural materials to those with recycled materials. The ANOVA (Table 10) showed significant variations of pH (p < 0.01) across the rigs. This confirms the pattern of pH distribution observed in Figure 7. Mean effluent pH values from Rig 1 (7.8 ± 0.1) and Rig 2 (7.9 ± 0.1) were neutral as compared to the alkaline mean effluent pH values from Rig 3 (12.0 ± 0.1) and Rig 4 (12.3 ± 0.1). The high pH values from Rigs 3 and 4 can be attributed to the dissolving of calcium hydroxide, Ca(OH) 2 from the hardened cement paste as the stormwater percolated through the sub-base materials. The Ca(OH) 2 was produced from cement hydration and the soluble metal alkalis present in cement (Dhir & Jackson, 1996). A similar explanation was provided by Zhang et al. (2018) for the reported high pH values obtained from permeable pavements with porous concrete and cement brick surfaces. However, the effluent pH from porous concrete surfaces tend to decrease over time due to the carbonation of the porous concrete. Guidelines for reuse of stormwater for domestic use or irrigation set a pH range of 6-9 (US EPA, 2012). Outflow from Rigs 3 and 4 would therefore most likely not be suitable for these types of reuse without further treatment. However, permeable pavements with high pH effluents can be beneficial and behave like a buffer for acidic rainfall events (Collins, Hunt, & Hathaway, 2010;Kazemi & Hill, 2015;Razzaghmanesh & Borst, 2019). Effluent pH from permeable pavements is highly reflective of the composition of the materials used within the pavement structure. Numerous studies have reported increases in pH effluents from permeable pavements consisting of porous concrete (Collins et al., 2010;Crookes, Drake, & Green, 2017;Drake, Bradford, & Van Seters, 2014b;Pilon et al., 2019;Vadas, Smith, & Luan, 2017; (Reddy, Xie, & Dastgheibi, 2014).

Change in chemical Oxygen demand (COD)
COD measures the organic strength of wastewaters in terms of the total quantity of oxygen required for oxidation to CO 2 and H 2 O (Sawyer, McCarty, & Parkin, 2003). The COD is widely used as a measure of the susceptibility to oxidation of the organic and inorganic materials present in water bodies (Chapman & Kimstach, 2002). Box and whiskers plots for influent and effluent COD concentrations along with average removal efficiencies are presented in Figure 8. Influent COD concentrations ranged from 35.1 to 119.0 mg/L with a mean of 70.9 ± 21.8 mg/L. The ANOVA (Table 10), Pearson's correlation tests (Table 11) and the Mann-Whitney U twoindependent samples tests (Table 12) showed no statistically significant difference (p > 0.05) between rigs with respect to COD effluent concentrations. COD removal was similar amongst all rigs, ranging from 4.2% (Rig 2) to 7.2% (Rig 1). Rigs 3 and 4 reduced COD by 4.5% and 6.5% respectively. The decrease in COD can most likely be attributed to the aerobic conditions and the subsequent oxidation of pollutants and organic material within permeable pavement rigs (Pilon et al., 2019). Zhang et al. (2018) found that after 48 h residence time, a permeable pavement containing shale bricks at the surface removed COD by approximately 46%. Balades, Legret, and Madiec (1995) reported that COD was reduced by a porous pavement by 80-90%. Pilon et al. (2019) reported 36% reduction in COD from a porous concrete section of a parking lot in Alcoa, TN, USA.

Change in dissolved Oxygen (DO)
Dissolved Oxygen (DO) determination measures the amount of dissolved (or free) oxygen present in a water or wastewater sample (Hammer & Hammer, 2007). In liquid wastes, DO determines whether the biological changes result from aerobic or anaerobic organisms (Sawyer  , 2003). Oxygen is essential to all forms of aquatic life, including those organisms responsible for the self-purification processes in natural waters. In freshwaters, DO at sea level ranges from 15 mg/L at 0°C to 8 mg/L at 25°C. Concentrations less than 5 mg/L may threaten the functioning and survival of biological communities and below 2 mg/L may lead to the death of most fish (Chapman & Kimstach, 2002). Mean DO influent and effluent concentrations and removal efficiencies are presented in Figure 9. Influent DO values had a mean of 7.19 ± 0.75 mg/L ranging from 5.67 to 8.46 mg/L. The ANOVA (Table 10) showed significant differences in DO (p < 0.01) across the rigs. DO was slightly reduced by 2% in Rigs 1 and 2. Rigs 3 and 4 on the contrary, produced a slight increase in DO by 3% and 5% respectively. This result was not expected and required further research. Further DO analysis was conducted on Rig 4 which sought to determine the most likely cause of increased DO concentrations. A modified approach was taken whereby six different influent samples of varying DO concentrations were poured into and stored in Rig 4 for a residence time ranging from 4 to 26 days for each sample. Results of the effluent DO measurements made are presented in Figure 10. All DO concentrations increased (except for the case of distilled water) from their initial values to an average constant value of circa 7.5 mg/L. DO values increased by 200% in some samples. It is common for the DO of stored water in permeable pavements  to deplete as residence time increases (Kazemi & Hill, 2015) because of microbiological activity. However, no odours signifying microbiological decay were detected during sampling events from Rig 4. Additionally, microbiological activity was not expected due to the high pH environment (Selvakumar & O'Connor, 2018). The increased DO concentration results are once again inconclusive and require further research.

Change in Nitrate-Nitrogen (NO 3 -N)
Nitrogen, which exists in four main forms (organic, ammonia, nitrite and nitrate) is of historical environmental concern in water which has led to the regulation of its concentration in surface waters for decades. Excessive levels of nitrogen, when discharged into receiving waters, can promote the growth of undesirable aquatic plants such as algae and floating macrophytes (Butler & Davies, 2011;Sawyer et al., 2003). Box and whiskers plots for influent and effluent NO 3 -N concentrations along with average removal efficiencies are presented in Figure 11. Influent NO 3 -N concentrations ranged from 0 to 6.50 mg/L with a mean of 1.49 ± 1.52 mg/L. NO 3 -N concentrations from Rig 4 were slightly higher than the other three rigs. The ANOVA (Table 10), Pearson's correlation tests (Table 11) and the Mann-Whitney two-independent samples tests (Table 12) showed no statistically significant difference (p > 0.05) between rigs with respect to NO 3 -N effluent concentrations. No rigs removed NO 3 -N except for Rig 3. Rig 1 increased NO 3 -N by 27%, Rig 2 by 21% and Rig 4 by 98%. Rig 3 reduced NO 3 -N by 20%. Water infiltrating through permeable pavements tend to cause increases in NO 3 -N (James & Shahin, 1998) which could be attributed to aerobic conditions that were likely present throughout the pavement allowing ammonium-nitrogen (NH 4 -N) to be nitrified to NO 3 -N (Collins et al., 2010;Drake et al., 2014b;Razzaghmanesh & Borst, 2019;Tota-Maharaj & Scholz, 2010). Denitrification of nitrate (NO 3 -) into nitrogen gas [N 2(g) ] requires anoxic conditions which are unlikely to be present in a permeable pavement structure given that these pavements are designed to be free draining (Drake et al., 2014b). Numerous studies (Bean et al., 2007;Braswell, Winston, & Hunt, 2018;Collins et al., 2010;Drake et al., 2014b;James & Shahin, 1998;Razzaghmanesh & Borst, 2019), which have compared runoff from conventional asphalt pavements to effluent discharges from permeable pavements, have reported increased NO 3 -N concentrations. However, a few studies (Gilbert & Clausen, 2006;Pagotto et al., 2000) have shown a reduction in NO 3 -N concentrations. Change in reactive phosphorous (PO 3− 4 ) Phosphorous is an essential nutrient for living organisms and is present as both dissolved and particulate species in water. High levels of phosphorous in water may lead to eutrophication and algal growth (Chapman & Kimstach, 2002). Box and whiskers plots for influent and effluent PO 3− 4 concentrations along with average removal efficiencies are presented in Figure 12. Influent PO 3− 4 concentrations ranged from 0.60 to 4.9 mg/L with a mean of 1.90 ± 1.09 mg/L. Mean effluent PO 3− 4 concentrations ranged from 1.39 ± 0.82 (Rig 4) to 2.20 ± 0.0.79 (Rig 1). The ANOVA (Table 10) showed significant variations of PO 3− 4 (p < 0.01) across the four rigs. Rigs 1 and 2 recorded slight increases in PO 3− 4 of 22% and 9% respectively, while Rigs 3 and 4 had reductions of 18% and 33% respectively. The increases in PO 3− 4 from rigs 1 and 2 can most likely be ascribed to the decomposition of organic matter present in the rigs. The removal of PO 3− 4 from Rigs 3 and 4 can be attributed to the formation of phosphate salts from the reaction of PO 3− 4 ions and the cementitious sub-base materials which lead to adsorption by the cementitious components. Wang, Chen, Kong, and Shi (2014) used cementitious materials for the sequestration of phosphorous from wastewater and reported removal rates of 80% for phosphorus concentrations ranging from 20 to 1000 mg/L. Agyei, Strydom, and Potgieter (2002) removed PO 3− 4 ions from aqueous solutions using fly ash, slag and Ordinary Portland Cement (OPC) and found that the rate and removal efficiency of PO 3− 4 was linked to increasing CaO and/or Ca 2+ ions in the adsorbents released into solution via hydration and dissolution. This agrees with the results presented in Figure 12 given that the CaO% (Table 3) was greater in C-EPS (Rig 4) than CCA (Rig 3). Deng and Wheatley (2018) reported that 2-5 mm recycled concrete aggregate (RCA) removed more than 90% of phosphorous from effluent and suggested that RCA could be used for both wastewater treatment and phosphorous recovery.
Change in sulphate (SO − 4 ) Surface waters tend to contain SO 2− 4 . Sources include atmospheric deposition of oceanic aerosols, leaching of sulphur compounds, industrial discharges or atmospheric precipitation. Bacteria can use sulphate as an oxygen source which is converted to hydrogen sulphide (H 2 S, HS − ) under anaerobic conditions (Chapman & Kimstach, 2002). Box and whiskers plots for influent and effluent SO 2− 4 concentrations along with average removal efficiencies are presented in Figure 13. Influent SO 2− 4 concentrations ranged from 0 to 77 mg/L with a mean of 17.8 ± 20.73 mg/L. The ANOVA (Table 10) showed significant variations of SO 2− 4 (p < 0.01) across the four (4) rigs. Rigs 1 and 2 recorded significant increases in SO 2− 4 (121% and 66% respectively) while Rig 3 and Rig 4 recorded 33% and 74% reductions in SO 2− 4 respectively. Increases in SO 2− 4 from Rigs 1 and 2 can most likely be attributed to the dissolution of SO 2− 4 ions from the aggregates as water infiltrated through the rigs. Pilon et al. (2019) found a 157% increase in SO 2− 4 from a pervious concrete pavement installed as a parking stall in Alcoa, TN, USA. They attributed this increase to the degradation of hydrocarbons or other organic compounds within the pavement structure or the creation of SO 2− 4 by oxidation of another form of sulphur such as sulphide (S 2− ). The high removal rates of SO 2− 4 by Rigs 3 and 4 on the other hand, can most likely be attributed to the reaction of SO 2− 4 ions and the cementitious sub-base materials leading to the formation of calcium sulphate (CaSO 4 ) and other salts which adhere to the sub-base materials. This reaction is essentially an attack by the SO 2− 4 ions on calcium hydroxide [Ca(OH) 2 ], tricalcium aluminate (C 3 A) and hydrated aluminate phases of the cement paste. The Ca(OH) 2 can convert to CaSO 4 as per Equation (2) and further to the subsequent growth of Figure 13. Sulphate results (a) box and whiskers plots for inflow and outflows (b) bar charts for mean removal efficiencies. ettringite crystals as per Equation (3) (Miron & Magaña, 2017).

Change in turbidity
The turbidity or cloudiness, of water may be defined as the interference of light passing through water by suspended matter such as silt, clay, organic matter, organic compounds, or dissolved inorganics (Alley, 2007;Hammer & Hammer, 2007;Sawyer et al., 2003). Suspended particles can be a significant health concern when heavy metals and hydrophobic chemicals such as pesticides adsorb to the particles (AWWA, 2011). Box and whiskers plots for influent and effluent turbidity values along with average removal efficiencies are presented in Figure 14. Influent turbidity values ranged from 6 to 184 mg/L with a mean of 57.25 ± 54.21 mg/L. The ANOVA (Table 10), Pearson's correlation tests (Table 11) and the Mann-Whitney two-independent samples tests (Table 12) showed no statistically significant difference (p > 0.05) between rigs with respect to turbidity effluent values. All rigs had positive removal rates for turbidity with Rig 4 recording the highest (57%) while Rig 3 the lowest (10%). Rigs 1 and 2 reduced turbidity by 37% and 31% respectively. Some studies (Chowdhury, Sharvelle, & Beecham, 2016;Tota-Maharaj & Scholz, 2010) have reported turbidity removal rates in excess of 90% from permeable pavements while Pilon et al. (2019) reported no significant change in turbidity removal rates.

Change in Total dissolved solids (TDS)
Total Dissolved Solids (TDS) are the inorganic salts and dissolved materials in the filtrate from the TSS test (i.e. with a diameter < 0.45 μm) (Alley, 2007;Butler & Davies, 2011). Water with a high dissolved-solids content tends to have adverse impacts on irrigated crops, plants and grasses (Sawyer et al., 2003). Box and whiskers plots for influent and effluent TDS values along with average removal efficiencies are presented in Figure 15. Influent TDS values ranged from 22.00 to 400.00 mg/L with a mean of 196.82 ± 108.69 mg/L. Effluent TDS removal percentages were negative for all rigs. TDS increased by 48% from Rig 1, 31% from Rig 2, 212% from Rig 3 and 387% from Rig 4. These variations were confirmed by ANOVA (Table 10) which showed significant variations in mean TDS (p < 0.01) across the rigs. In general, TDS is directly related Figure 14. Turbidity results (a) box and whiskers plots for inflow and outflows (b) bar charts for mean removal efficiencies. to EC (Chapman & Kimstach, 2002) and as such the negative removal rates for TDS can be attributed to the dissolution of ions and other mineral fractions on the surface of the materials within the pavement structure (Myers, Beecham, van Leeuwen, & Keegan, 2009) similarly to EC. As with high pH and EC, TDS values were high from Rigs 3 and 4 most likely because of the richness of CaO and other compounds present in the CCA and C-EPS (Table 3).

Change in Total suspended solids (TSS)
Total Suspended Solids (TSS) are the organic and inorganic solid matter maintained in suspension and retained upon evaporation and drying at 103-105°C, when a sample is filtered through filter paper with a pore size of approximately 0.45 μm (Alley, 2007;Butler & Davies, 2011;Sawyer et al., 2003). The finer fractions of suspended solids ( < 63 μm) are extremely effective pollutant carriers. High concentrations of suspended solids pose harmful threats to receiving water, including increased turbidity and interference with numerous types of fish and aquatic invertebrates (Butler & Davies, 2011). Box and whiskers plots for influent and effluent TSS values along with average removal efficiencies are presented in Figure 16. Influent TSS concentrations ranged from 18 to 386 mg/L with a mean of 131.95 ± 114.43 mg/L. Mean effluent TSS concentrations ranged from 46.50 ± 40.02 mg/L (Rig 4) to 64.48 ± 63.82 mg/L (Rig 1). The ANOVA (Table 10), Pearson's correlation tests (Table 11) and the Mann-Whitney twoindependent samples tests (Table 12) showed no statistically significant difference (p > 0.05) between rigs with respect to TSS effluent concentrations. As expected, TSS removal percentages were relatively high for all rigs. TSS removal in permeable pavements is highly credited to sedimentation and mechanical filtration through the pavement structure. The entrapment (and thus removal) of most sediments is largely dependent on the size of the particulate matter and occurs within the top layers of the pavement structure (Brown, Chu, van Duin, & Valeo, 2009;Lucke & Beecham, 2011). Rig 1 had the lowest removal percentage of 52% whilst Rig 4 has the highest rate of 64%. Rigs 2 and 3 had removal percentages of 53% and 58% respectively. Based on the clarity of the samples collected, Rig 4 was seen to produce the highest TSS removal rates. Additionally, the increased filtration surface area of the C-EPS in the sub-base of Rig 4 provided for increased filtration of particulate matter. The presence of a geotextile layer between the bedding layer and the base course layer in all rigs also contributed to the removal of TSS. Numerous studies (Pratt, 1997;Rahman et al., 2015b;Tota-Maharaj et al., 2012) have reported improved TSS removal efficiencies when geotextiles have been used in permeable pavements. Brown et al. (2009) assessed the processes and characteristics of solids removal in two types of permeable pavements in Canada and found that both pavement types removed 90% to 96% of suspended solids. Pilon et al. (2019) reported a 97% removal of TSS from a pervious concrete section of a parking lot in Alcoa, TN, USA.
Change in Electroconductivity (EC) Electroconductivity (EC) or conductivity, measured in microsiemens per centimetre (μS/cm), is a measure of the concentration of dissolved ions/salts present in a given solution (Sawyer et al., 2003;Tota-Maharaj, 2010). It is a measure of the ability of water to conduct an electric current and is sensitive to variations in dissolved solids, mostly mineral salts (Chapman & Kimstach, 2002). Conductivity itself is not an aquatic or human health concern, but because it is easily measured, it can serve as an indicator of other water quality issues (Tota-Maharaj, 2010). Mean EC concentrations and removal efficiencies for all rigs are presented in Figure 17. The results showed that EC values increased for all rigs. Rig 1 and Rig 2 had slight increases (33.5% and 17.2% respectively) whereas significant increases were observed from Rig 3 and Rig 4 (908% and 1895% respectively). These variations were confirmed by ANOVA (Table 10) Figure 17. Electroconductivity results (a) box and whiskers plots for inflow and outflows (b) bar charts for mean removal efficiencies.
which showed significant variations in mean EC values (p < 0.01) across the rigs. In general, the increases can be attributed to the dissolution of ions and other mineral fractions on the surface of the materials within the pavement structure (Myers et al., 2009). The composition of basalt and quartzite aggregates in the sub-base of Rigs 1 and 2 respectively, did not permit high levels of dissolution of ions. This was not the case for the CCA and C-EPS in Rigs 3 and 4 respectively. As with high pH, EC values were high from Rigs 3 and 4 most likely because of the richness of CaO and other metal compounds present in the CCA and C-EPS (Table 3).

Conclusion
This study evaluated the suitability of Crushed Concrete Aggregates (CCA) and a novel filter material referred to as Cement-bounded Expanded Polystyrene (C-EPS) as suitable sub-base materials in permeable pavements. The evaluation compared the physical characteristics of the recycled materials and pollutant removal efficiencies of permeable pavement rigs containing CCA and C-EPS to those made up of natural aggregates of basalt and quartzite aggregates. Natural stormwater runoff from various locations across Trinidad, West indies was used as the influent. This ensured that different levels of pollutant concentrations were assessed. Water quality parameters were analysed using the IBM statistical software SPSS version 20. Based on the results of this study, the following conclusions are made.
(1) Based on the physical measurements obtained, CCA demonstrated potential as a suitable construction material to substitute or compliment traditional quarried materials in permeable pavement applications.
(2) C-EPS has relatively low compressive strength, < 1.0 MPa. Hence, from a structural view point, C-EPS is recommended for use in permeable pavements in non-traffic areas such as building aprons, sidewalks, footpaths, landscapes, pedestrian access and bicycle lanes. To accommodate light vehicular traffic, structural modifications to the pavement structure such as an increase in the thickness of the base layer may be considered and evaluated.
(3) In terms of pollutant removal efficiency, Rigs 3 and 4 which contained CCA and C-EPS respectively, performed equally or better than their counterparts. On average, Rig 3 removed COD by 5%, NO 3 − N by 20%, PO − 4 by 18%, SO 4 by 33%, turbidity by 10% and TSS by 57%. On average, Rig 4 removed COD by 6%, PO 3− 4 by 32%, SO 2− 4 by 74%, turbidity by 57% and TSS by 64%. However, noticeable increases in pH, TDS and EC measurements were noted. These increases were driven by the cementitious nature of the recycled materials. Despite the reported increases, all mean effluent pollutant concentrations were within the Maximum Permissible Levels (MPLs) of water pollutants discharged into the environment according to the Environmental Management Authority (EMA), Trinidad and Tobago and the United States Environmental Protection Agency (US EPA). In this regard, the CCA and C-EPS performed satisfactorily as sub-base materials in the permeable pavement rigs. (4) The results showed that, in general, CCA and C-EPS can be used as suitable filter materials in the sub-base reservoir of permeable pavements. However, from an environmental view point, further analysis through leaching tests on these materials is recommended.