Introduction

Pervious pavements have become one of the most used sustainable urban drainage system (SUDS) techniques and have strong ability in the reduction of runoff and pollutants (Brattebo and Booth 2003; Dreelin et al. 2006; Tota-Maharaj and Scholz 2010; Sanudo-Fontaneda et al. 2013, 2014a). Many studies have studied pervious pavements, mainly focusing on (1) the infiltration capacity and pollutant removal rates of the entire permeable pavement structure (Brattebo and Booth 2003; Kuang and Fu 2013; Sanudo-Fontaneda et al. 2014a, b), (2) the decrease of the infiltration capacity due to surface clogging (Fassman and Blackbourn 2010; Coughlin et al. 2012; Sanudo-Fontaneda et al. 2013; Kuang and Fu 2013), and (3) the qualitative impacts of various factors (e.g., retention time, flow rate, season, dry time, and types of materials) on the average removal rates of pollutants (Dierkes et al. 2002; Cates et al. 2009; Tota-Maharaj and Scholz 2010). However, questions that remain unanswered include (1) the study of the influence of the thickness factor on the stormwater infiltration and surface runoff pollution reduction performances in several layers of the permeable pavement systems, and (2) the variations and mechanisms of water infiltration rates and pollutant removal rates of each layer over the rainfall duration.

In this paper, the small-scale permeable pavement layers, including a surface permeable brick layer, coarse sand bedding layers (thicknesses = 2, 3.5, and 5 cm), and single-graded gravel sub-base layers (thicknesses = 15, 20, 25, and 30 cm), were built to analyze the stormwater infiltration and surface runoff pollution reduction of each layer. This study sought to measure the stormwater infiltration and surface runoff pollution reduction of individual layer of permeable pavement and determine the effect of thickness on the water infiltration and pollution reduction. Moreover, the variation of the stormwater infiltration rate and pollutant removal rate of each layer over the rainfall duration was also examined.

This research could give aid to the design of the permeable pavement system for better efficiency of stormwater infiltration and pollution reduction and potentially promote the understanding of water infiltration and pollution reduction of the entire permeable pavement.

Materials and methods

Experimental materials

The permeable pavement layers were designed in accordance with the Engineering Technical Code for Rain Utilization in Building and Sub-District (GB 50400-2006) (Ministry of Construction of the People’s Republic of China 2008) (shown in Table 1). The ceramic permeable brick has a compressive strength of 48.9 MPa, a bending strength of 7.8 MPa, and a permeation coefficient of 0.0312 cm/s. They were laid flat by using a rubber hammer without damaging the corner of the bricks. Roughness deviation was not more than 5 mm, and the elevation difference between two adjacent bricks was not more than 2 mm. The coarse sand in experiment has a mud content of 0.9 %, a clod content of 0.4 %, a flat and elongated particle content of 9.5 %. The coarse sand bedding layer was also laid flat and slightly hand-compacted. The single-graded gravel used in this experiment has a mud content of 1.6 %, a clod content of 0.7 %, a flat and elongated particle content of 1.8 %. The gravel sub-base layer was compacted by grinding with thickness deviation less than 20 mm.

Table 1 Characteristics of each structural layer in the permeable pavement
Full size table
Table 2 Particle size distribution of the sand in the bedding layer
Full size table

Experimental setup

The layer with a specific depth was laid in a plastic container with dimensions of 75.5 cm long × 55 cm wide × 48 cm deep. An outlet pipe with a diameter of 2 cm was installed at the bottom of the container, and a 1 % slope was applied in the horizontal direction to ensure smooth drainage. The rain was simulated with a single pipe with six nozzles that can cover the whole pavement surface. The rainfall intensity range of the screw rotating sprayer could be adjusted from 15 to 200 mm/h. The pipe was connected to a water pump with a flowmeter to control the spray speed and monitor the rainfall intensity. The flow rate range of the flowmeter (LZB-10 glass rotameter) is 6∼60 L/h (shown in Fig. 1).

Fig. 1
figure 1

Schematic diagram of the experimental setup

Full size image

Test methods

According to the Engineering Technical Code for Rain Utilization in Building and Sub-District (GB 50400-2006) (Ministry of Construction of the People’s Republic of China 2008), the rainfall should be determined in terms of local rainfall data of more than 10 years, and the rainfall intensity should be more than 2 years of return period. Consequently, a rainfall of 60-min duration was simulated using the rainfall intensity of a 20-year return period in Tianjin (q = 201.14 L/(s·ha)). Several studies have found that suspended solids (SS), total phosphorus (TP), chemical oxygen demand (COD), ammonia nitrogen, and total nitrogen (TN) were the main pollutants in stormwater runoff in the urban area (Brezonik and Stadelmann 2002; Gnecco et al. 2005; Shen et al. 2009; Zhang and Li 2008). So these water quality parameters were chosen to be investigated in this paper. Raw water (influent water) was prepared based on the traditional surface runoff measurements from July to September 2012 in Tianjin, in North China (Table 3), and stored in a 50-L water tank before each test. The sediment used in the simulated rainfall was kaolin, which particle size distribution is shown in Fig. 2. The chemicals were all purchased from chemical network management platform of Tianjin University (http://219.243.47.181/chem).

Fig. 2
figure 2

Particle size diameter distribution of the kaolin in the simulated rainfall

Full size image
Table 3 Raw water quality
Full size table

The rain began spraying down from nozzles to initiate the timing, and the outflow was taken at 5-min intervals referring to the previous research studies by Bean et al. (2007), Hu (2009), Yang et al. (2013), and Zhou (2007). The effluent volume and sampling time were recorded simultaneously. Stirring was always conducted to avoid natural sedimentation of SS and nonuniform concentration of pollutants in raw water, and the laboratory conditions were controlled during the tests (temperatures 18∼23 °C, humidity 70∼80 %). Eight permeable pavement-layer experiments were carried out, and each one was repeated three times for improving the subsequent statistical analysis of the results in this study (shown in Table 4). In addition, The event mean concentration (EMC) of effluent volume and various pollutants from each layer was calculated to analyze the infiltration rate and pollutant removal rate. The water quality was analyzed within 24 h after sampling, and the measurement methods are shown in Table 5.

Table 4 The number of tests and repetitions carried out for each layer during the research
Full size table
Table 5 Laboratory methods used in analysis of samples
Full size table

Statistical analysis

The results of effluent volume and pollutant concentration were processed by EXCEL2010, and figures between infiltration rate and both pollutant removal rate and rainfall duration were generated by ORIGIN8.0 software.

Quality assurance

All the samples and blanks were analyzed in duplicates for quality assurance control of laboratory analyses in this study. Only the relative standard deviation (RSD) values below 10 % were accepted, while other samples outside this range were reanalyzed. A standard reference material (SRM) GSB 07-1982-2005 from Institution for Environmental Reference Material Ministry of Environmental Protection was applied for calibration and analytical control. The parameters were identified relative to external standards. The detection limits of TP, COD, ammonia nitrogen, and TN were 0.01, 15, 0.025, and 0.05 mg/L, respectively.

Results and discussion

Permeable brick layer

Infiltration rate of the permeable brick layer

The infiltration rate of the permeable bricks is shown in Fig. 3. The steady infiltration rate of the brick pavers was 51.0 mm/h, and no outflow was observed before 5 min. The outflow will occur when the permeable bricks are dripping wet. About 10 min later, the infiltration rate became steady and approached the saturated hydraulic conductivity, which confirmed the results from Davies et al. (2002), Nnadi et al. (2014), Rodriguez-Hernandez et al. (2012), and Sanudo-Fontaneda et al. (2013) regarding with the time needed to reach the steady state.

Fig. 3
figure 3

Infiltration rate of the permeable bricks

Full size image

Pollutant removal of the permeable brick layer

The pollutant removal rates of the permeable brick layer as a function of the rainfall duration are shown in Fig. 4, and the water quality of the effluent is shown in Figs. 5, 6, 7, 8, and 9.

Fig. 4
figure 4

Pollutant removal rates of the permeable bricks

Full size image
Fig 5
figure 5

Effluent concentration of SS over the rainfall duration from a permeable bricks and bedding sands and b sub-base gravels

Full size image
Fig 6
figure 6

Effluent concentration of TP over the rainfall duration from a permeable bricks and bedding sands and b sub-base gravels

Full size image
Fig 7
figure 7

Effluent concentration of COD over the rainfall duration from a permeable bricks and bedding sands and b sub-base gravels

Full size image
Fig 8
figure 8

Effluent concentration of ammonia over the rainfall duration from a permeable bricks and bedding sands and b sub-base gravels

Full size image
Fig 9
figure 9

Effluent concentration of TN over the rainfall duration from a permeable bricks and bedding sands and b sub-base gravels

Full size image

The average removal rate of SS was 79.8 %. The bricks rapidly removed particles by filtration, sorption, and interception (Hou 2008).

The average removal rate of TP was 71.2 %. The calcium components in the ceramic permeable bricks could adsorb part of the phosphorus (Geller 1997; Zhu 1999). The change in the TP removal rate was consistent with that of the SS, which indicates that TP could also be adsorbed by particles, consistent with the conclusions of Xiong (2007), Cates et al. (2009), and Eck et al. (2012).

The average removal rate of COD was 24.1 %. The removal rate was negative during the first 15 min because residual organics previously deposited in the permeable bricks desorbed by scouring. Then, the removal rate increased and ultimately stabilized due to reaching an adsorption-desorption balance (Hou 2008).

The average removal rate of ammonia nitrogen was 3.25 %. Kaolin performed the function of cation exchange (Han 2007), so ammonia was removed mainly through ion exchange. However, only a small amount of ammonia was removed because of the limited number of ion sites (Han 2007). The average removal rate of TN was 3.25 %, consisted with Eck et al. (2012).

Sand bedding layer

Infiltration rate of the sand bedding layer

The infiltration rates of sand bedding layers with three different thicknesses are shown in Fig. 10. The steady infiltration rates of different thicknesses in descending order are 32.3 mm/h (5-cm layer) > 30.9 mm/h (2-cm layer) > 24.5 mm/h (3.5-cm layer). In the 2-cm layer, the water infiltration rate was more prominently influenced by gravity during the initial 20 min because fewer particles were intercepted than for the other thicknesses (Figs. 5a and 11) (Pan et al. 2010). Gravitational drainage can drive the lower water to move quickly under the hydrodynamic pressure (Zhang 2005; Liu 2008), and in consequence, the infiltration rate increased dramatically. The infiltration rate decreased soon afterward because the air in the structure could not exhaust effectively under the rapid flow of water (Lv and Hao 2008). After that, the matric suction increased and the gravity weakened due to the increase of the particulate matter filtered (Liu 2008; Pan et al. 2010). The infiltration rate of the 5-cm layer was similar to that of the 2-cm layer. It could be concluded that no air was retained in the 3.5- and 5-cm layers because during the initial 30 min, more particles were retained in the two layers than in the 2-cm layer (Figs. 5a and 11), so no rapid flow appeared. The matric suction was more obvious in the 3.5-cm layer because of the particles accumulated; moreover, the effect of the suction is weaker than that of gravity (Pan et al. 2010), so the infiltration rate of the 3.5-cm layer increased slowly.

Fig. 10
figure 10

Infiltration rates of the sand bedding layers with different thicknesses

Full size image
Fig. 11
figure 11

The removal rates of SS and TP in sand bedding layers of three thicknesses

Full size image

Pollutant removal of sand bedding layer

The pollutant removal rates of the sand bedding layers are shown in Table 6, and the water quality of the effluent are shown in Figs. 5, 6, 7, 8, and 9. The removal rates of SS, COD, and ammonia nitrogen of the 5-cm layer were higher than those of the 3.5- and 2-cm layers, while the removal rate of TP in the 3.5-cm layer was the highest. The sand bedding layer could hardly remove TN.

Table 6 Pollutant removal rates of sand bedding layers with three thicknesses
Full size table

The removal rates of SS and TP in the sand bedding layers of three thicknesses are shown in Fig. 11. The infiltration rate of the 2-cm layer was much higher than that of the 3.5- and the 5-cm layers during the first 20 min (Fig. 10). The rapid movement of water affected the adsorption and retention of SS, leading to a sharp decline in its removal. In addition, the removal of TP was mainly dependent on adsorption of particles (Xiong 2007; Cates et al. 2009; Tota-Maharaj and Scholz 2010; Eck et al. 2012). So there was a significant amount of TP captured as SS washed out at this period (Figs. 6a and 11), leading to the dramatic increase of its effluent concentration (Fig. 6a). Then, with the decrease of the infiltration rate, the removal rate of SS increased and reached a steady state after 30 min. The similar situation happened on the removal rate of TP from 2-cm bedding sands meanwhile. The removal rate of TP in the 3.5-cm layer was the highest because of the significant quantity of retained particles and the lowest infiltration rate (Xiong 2007; Cates et al. 2009; Tota-Maharaj and Scholz 2010; Eck et al. 2012).

The COD removal rates in the sand bedding layers are shown in Fig. 12a. COD was removed rapidly because of the adsorbent site vacancies in sand during the early stage of precipitation. Soon after, the removal rates increased slowly with the decrease of the adsorption sites and ultimately reached a steady state. Srimurali et al. (1998) studied the removal of fluorides by adsorption onto bentonite and charfines and reached similar conclusions.

Fig. 12
figure 12

COD removal rates in a sand bedding layers and b gravel sub-base layers

Full size image

The effluent pH varied from 5 to 7.85, and ammonia exists mainly in ionic form under these conditions (Li et al. 2007), so the ammonia could be removed by cation exchange with kaolin (Han 2007). However, the ion exchange capacity of kaolin is limited (Han 2007), so the removal rate of ammonia declined significantly as the rainfall continued. The removal rate of ammonia in the 5-cm layer was slightly higher than that in the other thicknesses due to the great amount of adsorbent. It could be concluded from Fig. 13 that nitrate was prone to be desorbed by flushing and could not be removed in the sand bedding layers. Figure 13 indicates that the effluent of TN was influenced mostly by nitrate during the initial 20 min and thereafter by both nitrate and ammonia.

Fig. 13
figure 13

Removal rates of ammonia, TN, and NO3-N in sand bedding layers of three thicknesses

Full size image

Gravel sub-base layer

Infiltration rate of the gravel sub-base layer

The infiltration rates of gravel sub-base layers of four thicknesses are shown in Fig. 14. The steady infiltration rate of each layer in descending order was 42.3 mm/h (30 cm) > 36.7 mm/h (25 cm) > 27.3 mm/h (15 cm) > 19.5 mm/h (20 cm). The infiltration rates of the 25- and 30-cm layers were higher than those of the 15- and 20-cm layers during the first 30 min because the former two depths were less clogged (Figs. 6b and 15). Thereafter, the infiltration rates increased slowly with the accumulation of particles. The infiltration process of the 15- and 20-cm gravel sub-base layers was consistent with that of the 3.5-cm sand bedding layer because of particle accumulation.

Fig. 14
figure 14

Infiltration rates of gravel base layers of different thicknesses

Full size image
Fig. 15
figure 15

SS and TP removal rates in gravel base layers of four thicknesses

Full size image

Pollutant removal of the gravel sub-base layer

The pollutant removal rates of gravel sub-base layers of four thicknesses are shown in Table 7, and the water quality of the effluent are shown in Figs. 5, 6, 7, 8, and 9. For the removal of TP, COD, and ammonia, the 20-cm layer was higher than the others, while the 15-cm layer was highest for SS removal. For TN, there was no stable removal observed.

Table 7 Pollutant removal rates in base gravel layers with four different thicknesses
Full size table

High flow rates would reduce the adsorption and sedimentation of media to particles, so the SS removal rates of the 15- and 20-cm layers were higher than those of the others. The TP removal rates of the 15- and 20-cm layers were significantly higher than those of the other two layers, which was consistent with the SS removal rates. Hence, it could be concluded that TP was also adsorbed by particles in the gravel sub-base layer (Fig. 15).

The removal rates of COD in the gravel sub-base layers are shown in Fig. 12b. The removal rate of COD in the 20-cm layer was higher than that of the others because the worst infiltration in the 20-cm sub-base layer favored the adsorption and interception of COD (Hou 2008; Fig. 14).

The infiltration rates were higher in the 25- and 30-cm layers (Fig. 14), which weakened the adsorption and ion exchange (Hou 2008), so the removal rates of ammonia dropped more quickly in the 25- and 30-cm layers than those in the 15- and 20-cm layers (Fig. 16). The results show that the removal rate of TN was the highest in the 25-cm layer, and no removal was observed in the other layers (Fig. 16).

Fig. 16
figure 16

Removal rates of ammonia and TN in base gravel layers of four thicknesses

Full size image

Comprehensive analysis

There was no surface runoff produced in the permeable bricks and sub-base gravels during the experiment. Thus, the reduction of runoff is more effective in the surface and sub-base layers than in the entire pavement structures reported by Rushton (2001) (50 %), Gilbert and Clausen (2006) (72 %), Dreelin et al. (2006) (93 %), and Collins et al. (2008) (98 %). This is because the single layer has a shorter distance and higher porosity than the entire permeable pavement. However, a small amount of runoff existed in the sand bedding layer during the test. This suggests that the bedding layer was a main limiting factor for water infiltrating into the entire permeable pavement, and the sub-base layer played a major role in determining the water storage capacity of the permeable pavement, which is consistent with the study by Park et al. (2014).

The removal rates of SS from the permeable pavement layers agreed with the observations on permeable pavement by Pezzaniti et al. (2009) (94 %), Tota-Maharaj and Scholz (2010) (91 %), and Kuang and Fu (2013) (87.6 %). The removal rates of TP and ammonia were lower than those (78 % and 85 %) in the study of Tota-Maharaj and Scholz (2010). In the study of Tota-Maharaj and Scholz (2010), the depth of each layer of permeable pavement was significantly greater than that in this paper, which leads to higher capacities of adsorption, interception, filtration, and ion exchange; what is more, processes of microbial degradation and nutrient transformation such as nitrification and denitrification could also be observed during a longer rainfall period.

Conclusions

The results presented in this paper indicate that

  1. 1.

    The steady infiltration rate of the brick pavers was 51.0 mm/h, while those of the bedding and sub-base layer mainly depended on the thickness in this experiment. The optimal thickness of the sand bedding layer was 5 cm (the infiltration rate = 32.3 mm/h), and that of the sub-base layer was 30 cm (42.3 mm/h).

  2. 2.

    SS could be effectively removed by all layers, with removal rates from 79.8 to 98.6 %. The 5-cm sand bedding layer had the highest removal rates for COD and ammonia nitrogen (9.0 and 32.5 %), while the 3.5-cm bedding sands had the highest removal rate of TP (74.2 %) resulted from both the retained particles and infiltration rate. The sand bedding layer could hardly remove TN. For the removal of TP, COD, and ammonia, the 20-cm gravel sub-base layer was higher than the others, with the removal rates of 72.2, 26.1, and 10.3 %. For TN, 25 cm was the optimal thickness while the removal rate reached 35.6 %.

  3. 3.

    Considering both the water infiltration and pollution reduction, the recommend optimal thickness of bedding sands is 5 cm, and that of sub-base gravels is 20∼30 cm.

In the future, field tests will be developed to validate the results obtained in the laboratory. Also, further tests will be conducted on other surfaces and different materials for the bedding and sub-base, applying the same laboratory and statistical methodology, with the aim of characterizing the infiltration behavior of the most used permeable pavement layers across the world.