Combined Effects of Substrate Depth and Vegetation of Green Roofs on Runoff and Phytoremediation under Heavy Rain

Abstract

The quantity and quality of runoff from green roofs have so far been studied using an extensive vegetated roof (substrate depth > 150 mm). However, studies on various substrate depths and vegetation for runoff and phytoremediation in temperate and monsoon climates, where heavy rain is concentrated in a specific season, are limited. Therefore, the purpose of this study was to investigate combined effects of substrate depth and vegetation of an unfertilized green roof on runoff reduction and airborne pollutant purification based on rainfall intensity. A total of 21 plots were implemented on a roof top with three substrate depths (100, 200, and 400 mm), two vegetation s (vegetated or non-vegetated), and control (plot with standard frame only). The runoff reduction increased significantly (p < 0.05) with increasing substrate depth. Vegetated plots had a slightly higher runoff reduction than plots without vegetation. Compared to controls, turbidity and pH tended to increase regardless of vegetation or substrate depth, with the exception of electrical conductivity (EC). However, concentrations of heavy metals (Cu, Zn, Mn, and Cd) in the runoff of vegetated plots were all significantly (p < 0.05) lower than those of un-vegetated plots and controls. These results suggest that as the rainfall intensity increases, the depth of the substrate is more important than vegetation for runoff reduction. In addition, the vegetation can be an effective tool to neutralize acid rain to stabilize pH and effectively reduce EC and heavy metals in the runoff by remediating dissolved air pollutants from rainwater.

1. Introduction

Urbanization is a major driver of change as more forests are converted to impervious surfaces, leading to massive increases in stormwater runoff and nonpoint source pollution [1]. Urban impervious areas such as roads, buildings, and roofs generate significant stormwater runoff in heavy rain conditions that can create flash flooding problems [2]. In particular, roofs generally account for approximately 20~25% of the total urban surface area [3] and 40~50% of total impervious surface area [4]. In urban areas, roof surfaces contribute to excess nutrients and toxic metals in receiving water [5]. Green roofs contribute to flood mitigation in urban areas by reducing surface runoff and allowing delays in peak flow, including decreased loading of contaminants to the wastewater system and surface waters during high intensity hydrologic events [6]. Moreover, green roofs use existing roof space without requiring an extra land, potentially allowing widespread applications [7]. Consequently, green roofs are increasingly becoming important components of low-impact development (LID), sustainable urban drainage system (SUDS), and water-sensitive urban design system (WSUD) [8]. In addition, the use of green roofs is increasingly recognized in many countries as a viable solution to improve environmental quality [9]. Thus, green roofs are emerging as practical strategies to retain stormwater runoff and improve water environment quality of urban cities [8]. For this reason, research on green roofs has grown steadily since 1981, with around 40.9% of articles focusing on service regulation in the context of water, 30.0% on temperature, and 3.5% on air quality [3,10].

Green roof types may vary depending on building structure, climate condition, substrate, and vegetation used [11]. Green roofs are generally divided into intensive and extensive engineering categories based on the depth of the substrate. Extensive green roofs consist of substrate layers with a maximum depth of about 150 mm, whereas intensive green roofs are established with deeper soil layers than extensive ones [12]. This type is thickest and heaviest green roof (usual weight > 300 kg/m²), which requires additional structural support due to increasing roof loads. On the other hand, extensive green roofs (usual weight 50~150 kg/m²) do not required addition support because the soil layer is shallow and light, and can be easily installed during new construction and renovation of buildings. However, an extensive green roof can only accommodate a limited range of plants [10,13]. Thus, the selection of the most suitable system must take into account building characteristics and local climate conditions [14,15].

It has been reported that stormwater performance of a green roof is based on factors such as roof slope, media depth, and roof surface [16]. Green roof’s capacity to retain runoff depends on its physical configuration. It is also greatly influenced by local climatic events including precipitation characteristics [17]. Regional climatic conditions such as seasonality of rainfall can significantly alter the runoff performance of a green roof [10]. In addition to seasonal effects, the frequency and intensity of rainfall can also affect runoff reduction [18]. Moreover, substrate depth has long been considered a factor influencing green roof rainwater retention [19]. Rainwater is initially stored on a green roof (vegetation + substrate). Extra water is carried over to the nearby ground. Excess water then flows into the sewage system, which usually helps reduce the maximum flow and volume compared to conventional roofs [2]. Green roof substrates account for 80~90% of a green roof water storage capacity and 80~90% of total system weight [20]. However, runoff reduction studied so far used extensive green roofs with substrate depth typically less than 150 mm [2,8,21,22].

Green roofs can reduce nutrient and pollutant runoff by absorbing and filtering pollutants because green roofs are typically composed of multiple layers of components including vegetation, substrate, filter fabric, drainage material, root barrier and insulation [8,14,23]. In particular, substrate of green roof is very important because it can help capture rainfall runoff and growth of vegetation with regard to nutrient retention [2]. On the other hand, it has been found that total nitrogen and total phosphorus are leached from the medium composed of 15% compost due to rooftop greening [22]. Liu et al. [8] have also reported that total suspended solids, total nitrogen, and total phosphorus concentrations of a green roof runoff are all significantly higher than those of a conventional roof runoff. Indeed, the behavior of green roof as a sink of pollutants is still a controversial issue among researchers [24]. Unfortunately, studies conducted on green roofs with deeper substrates to improve runoff quality are relatively limited [10]. Moreover, previous studies have focused on runoff quality in terms of fertilizer nutrients in relation to eutrophication in green roofs such as nitrogen, phosphorous, and suspended solids [8,16,22,25].

Thus, the objective of current study was to investigate performances of green roofs depending on substrate depths of greater than 150 mm and effects of vegetation on runoff quantity and quality based on rainfall intensity. We believe that the current study provides useful data to determine rainfall performance of an extensive or intensive green roof in temperate and monsoon climates where heavy rainfall is concentrated in a specific season. Results of this study encourage the adoption of green roof as one of stormwater management solutions, including improving the urban environment.

2. Materials and Methods

2.1. Study Site

Field experiments were conducted on a rooftop−platform of Konkuk University Complex Practice Building in Chungju, Chungcheongbuk-do (latitude 35°49′ N; and longitude 127°08′ E) located in the central part of South Korea. Korea generally has a humid continental climate (hot-humid summers, cool−dry winters) with an annual precipitation of 1100~1360 mm (of which 50% is concentrated in summer from June to August) and a rather dry climate in spring, autumn, and winter from September to May [2]. Although the antecedent dry weather period (ADWP) is also considered a major factor influencing the rainfall performance of a green roof, summer (June ~ August) was chosen as the study period to facilitate runoff sample collection [9]. Indeed, outside of this period, low temperatures and dry winds often severely limit viable vegetation in South Korea. During this study period, 40 rainfall events were recorded with a total precipitation of 440.2 mm. Monthly rainfall intensities in June, July, and August were approximately 63.3, 92.6, and 284.3 mm/h, respectively (rainfall events of 9, 14, and 17, respectively).

2.2. Experimental Setup and Sampling

Green roofs experiments were usually conducted on roof surfaces or on green roof platforms. In this study, a green roof platform simulated a roof surface. Unlike the existing roof surface, the bottom of green roofs surface is open to the atmosphere. Thus, a full−scale vegetation roof with the same assembly is possible [5,26]. The design of experimental green roofs was based on a plot that considered two factors. The first factor was substrate depth, for which depths of 100, 200, and 400 mm were used to represent extensive and intensive green roofs profiles [27]. The second factor was vegetation (either vegetated or non−vegetated). Therefore, a total of 21 plots were implemented on the rooftop, each with 3 controls (plots with standard frame only for rainfall sampling), 9 vegetated, and 9 non−vegetated plots. Plots were constructed by platforms with wood laminate, which consisted of waterproof layer with internal dimensions of 50 cm length × 50 cm width × 50 cm height on top of a 2 mm thick steel frame. The plot consisted of a drainage layer, a filter fabric, a growing substrate, and a vegetation from the bottom up [28]. Weight−loading of green roofs is one of the biggest barriers to widespread absorption. It can and reduce the possibility of retrofitting existing buildings. Increasing the substrate depth can increase the weight of the green roof, reducing its applicability to low load-bearing buildings [20]. Thus, a light−weight material was selected as a growth material in consideration of its bearing capacity. The volume ratio of perlite:peatmoss:vermiculite was set to be 1:1:1. The vegetation was planted with Kentucky bluegrass (Poa pratensis), which is highly resistant to drought stress conditions and widely used in green roofs [29]. An acrylic V−flume water channel, which could facilitate the collection of stormwater runoff, was installed at the bottom of the drainage layer and connected to a plastic sampling container with a volume of 20 L (Figure 1). Immediately after collection, all samples were transported back to the laboratory for analysis and prepared for testing water quality, which was carried out within a week of collection in all cases.

2.3. Runoff Reduction and Quality Analyses

Although evapotranspiration is a key parameter influening stormwater retention capacity [27], runoff reduction is generally defined as the percentage of total rainfall control associated with retention and evapotranspiration effect of substrate and vegetation [2]. In other words, the retained water in green roof is equivalent to evapotranspirtation over several wet/dry cycles [10]. Therefore, evapotranspiration was not independently considered in this study. As a result, the runoff reduction was calculated using the following equation [5,30]:

$$\mathrm{Runoff}\mathrm{reduction}(\%)=\frac{\mathrm{Rainfall}\mathrm{per}\mathrm{plot}\left(\mathrm{mL}\right)-\mathrm{Runoff}\mathrm{per}\mathrm{plot}\left(\mathrm{mL}\right)}{\mathrm{Rainfall}\mathrm{per}\mathrm{plot}\left(\mathrm{mL}\right)}\times 100$$

(1)

Water quality parameters of plots runoff were focused on pH, EC (electric conductivity), turbidity, and heavy metals (copper, zinc, manganese, and cadmium) in this study. A portion of each runoff sample was analyzed for pH (AZ−86505, Bench, China), electric conductivity (Control−510, EUTECH, Singapore), and turbidity (2100AN, HACH, USA), respectively. Heavy metal concentrations in runoff samples were measured using an atomic absorption spectrophotometer (ICP, Optima 5300DV, Perkin Elmer, Hopkinton, MA, USA).

2.4. Statistics

Data were subjected to analysis of variance (ANOVA) with SPSS 19.0 software package (SPSS Inc., Chicago, IL, USA). Significance was indicated by Duncan’s Multiple Range Test (DMRT) at the 5% level. Multiple linear regressions analyses were performed to evaluate potential combined effect of substrate depth and vegetation on runoff quantity and quality. All figures were carried out using SigmaPlot 15.0 (Systat. Software, Inc., Cary, NC, USA).

3. Results

3.1. Runoff Reduction

All green roofs were more effective in precipitation retention than the control (hard roofs). However, there was a difference in the amount of runoff among June, July, and August (24~78, 23~80, and 24~48%, respectively). The amount of runoff reduction decreased with increasing rain intensity. There were significant differences in runoff reduction according to vegetation and substrate depth. All green roof systems in this experiment were reduced approximately 23~80% more than controls. Moreover, results showed 39% and 50% rainfall reduction in runoff from non−vegetated and vegetated green roof system, respectively. Runoff reduction was the lowest for substrate with depth 100 mm + non−vegetated (23%). It was the greatest for substrate depth with 400 mm + vegetated (80%). At the same substrate depth, the vegetated system performed better in reducing runoff than the non−vegetated system (

Table 1. Runoff reduction from different substrate depths and vegetation plots based on rainfall intensity.

Treatments	June		July		August
	Runoff (mL)	Reduction (%)	Runoff (mL)	Reduction (%)	Runoff (mL)	Reduction (%)
Cont. z	14,567 a y	-	14,233 a	-	38,980 a	-
NV100	11,137 b	24 f	11,010 b	23 d	29,563 b	24 d
NV200	8770 d	40 d	8657 c	39 c	25,993 c	33 c
NV400	3907 f	73 b	6600 d	54 b	23,450 d	40 b
V100	9950 c	32 e	9080 c	36 c	26,697 c	32 c
V200	6910 e	53 c	6003 d	58 b	24,463 d	37 b
V400	3170 g	78 a	2820 e	80 a	20,463 e	48 a

Note: ^z Cont., Plots with standard frame only; NV100, Non−vegetated + substrate depth 100 mm; NV200, Non−vegetated + substrate depth 200 mm; NV400, Non−vegetated + substrate depth 400 mm; V100, Vegetated + substrate depth 100 mm; V200, Vegetated + substrate depth 200 mm; V400, Vegetated + substrate depth 400 mm. ^y Different letters in the same row indicate a significant difference as determined by Duncan’s test (p < 0.05).

).

Results of regression analysis indicated that runoff had a negative relationship with substrate depth, while reduction increased significantly with an increase in substrate depth (p < 0.05). Vegetated green roofs showed slightly higher runoff reduction than non−vegetated ones. The reduction in runoff from green roofs was found to decreased by approximately 0.03% with an increase of 1 mm in substrate depth (Figure 2).

3.2. Runoff Quality

Runoff from green roofs generally had a higher pH than controls. The control had an average pH of 5.6, while the measured pH value from runoff averaged 6.9. In addition, vegetated plots had slightly higher pH than non−vegetated plots. The difference in pH between the two was statistically significant (p < 0.05) except for that in June. The pH values of runoff from vegetated plots with different substrate depths were similar. However, the pH of runoff from vegetated plots increased statistically with increasing substrate depth. Mean electronic conductivity (EC) concentrations in control runoff in June, July, and August were 0.05, 0.09, and 0.04 dS·m⁻¹, respectively. EC concentrations were higher in runoffs from non−vegetated plots than in those from vegetated plots (0.17 ~ 0.47 vs. 0.12 ~ 1.34 dS·m⁻¹). Runoffs from plots with various substrate depths showed statistically significant difference (p < 0.05) in EC, with higher values recorded for deeper substrate depths. Runoff turbidity values were highly dependent on rainfall intensity rather than on vegetation or substrate depth. They were increased in August. The washing process of rainfall strongly affected particulate matter. In August, the non−vegetated + 100 mm substrate depth treatment showed the highest turbidity (4.66 NTU), whereas the control had the lowest turbidity (1.43 NTU) (

Table 2. pH, EC, and turbidity in runoff from different substrate depths and vegetation plots based on rainfall intensity.

Treatments	June			July			August
	pH	EC (dSm−1)	Turbidity (NTU)	pH	EC (dSm−1)	Turbidity (NTU)	pH	EC (dSm−1)	Turbidity (NTU)
Cont. z	5.8 e y	0.05 d	1.25 c	5.3 c	0.09 d	1.88 c	5.8 e	0.04 f	1.43 b
NV100	6.8 cd	0.20 c	1.78 bc	6.3 c	0.22 cd	3.25 ab	6.7 d	0.12 e	4.66 a
NV200	7.0 ab	0.36 b	1.95 b	7.0 a	0.47 b	3.15 ab	6.8 c	0.28 c	2.96 ab
NV400	7.1 a	1.34 a	1.92 b	7.2 a	1.14 a	2.47 bc	7.0 ab	0.64 a	3.85 a
V100	6.9 bc	0.31 bc	3.21 a	7.1 a	0.36 bc	3.89 a	6.8 c	0.17 d	4.28 a
V200	6.9 bc	0.25 bc	2.07 a	7.3 a	0.37 bc	3.46 ab	6.9 b	0.32 c	3.90 a
V400	6.8 d	0.34 b	2.73 a	7.2 a	0.17 d	3.36 ab	7.0 a	0.47 b	3.63 a

Note: ^{z, y} See Table 1.

).

Overall, runoff pH, EC, and turbidity showed positive relationships with substrate depth. However, non−vegetated or vegetated green roofs had no significant effect on the relationship between substrate depth and pH or turbidity except for EC. The EC of the non−vegetated runoff was higher than that of the vegetated runoff. In addition, EC value increased 2 dSm⁻¹ when the substrate depth increased by 1 mm (Figure 3).

Heavy metals (Cu, Zn, Mg, and Cd) in runoff from vegetated green roofs were all significantly (p < 0.05) lower than those in runoff from non−vegetated plots and controls. The highest concentrations of heavy metals were recorded for control, which was expected to release almost all incoming precipitation as runoff. Meanwhile, heavy metals in runoff decreased significantly (p < 0.05) with increasing substrate depth. In both vegetated or non−vegetated plots, green roofs with substrate depth > 200 mm performed better in reducing heavy metals than those with substrate depth < 200 mm (

Table 3. Four heavy metals (Cu, Zn, Mn, and Cd) in runoff from different substrate depths and vegetation plots based on rainfall intensity.

Plots	June				July				August
	Cu	Zn	Mn	Cd	Cu	Zn	Mn	Cd	Cu	Zn	Mn	Cd
Cont. z	0.027 a y	0.969 a	0.093 a	0.015 a	0.028 a	0.967 a	0.091 a	0.016 a	0.027 a	0.976 a	0.092 a	0.015 a
NV100	0.021 b	0.717 b	0.085 ab	0.014 b	0.021 b	0.718 b	0.083 b	0.014 bce	0.021 b	0.719 b	0.087 ab	0.014 a
NV200	0.019 c	0.691 c	0.083 bc	0.014 b	0.019 c	0.688 c	0.081 b	0.014 bc	0.019 c	0.687 c	0.082 bc	0.014 a
NV400	0.015 e	0.612 e	0.083 bc	0.014 b	0.015 d	0.610 e	0.081 b	0.013 cde	0.015 de	0.613 e	0.085 bc	0.014 a
V100	0.018 d	0.641 d	0.077 d	0.014 b	0.018 c	0.647 d	0.074 c	0.015 ab	0.018 c	0.642 d	0.080 c	0.014 a
V200	0.016 de	0.619 e	0.073 d	0.012 c	0.016 d	0.613 e	0.070 c	0.012 e	0.016 d	0.611 e	0.069 d	0.013 b
V400	0.013 f	0.581 f	0.073 d	0.012 c	0.014 e	0.580 f	0.073 c	0.013 de	0.013 e	0.582 f	0.074 d	0.012 b

Note: ^{z, y} See Table 1.

).

There was a negative relationship between substrate depth and heavy metal concentration in runoff. Substrate depth showed significantly negative relationships with Cu and Zn concentrations (R² = 0.6~0.8, p < 0.05), while Mn and Cd concentrations were significantly different between vegetated and non−vegetated plots. Overall, vegetation with higher substrate depth was associated with lower concentrations of heavy metals (Figure 4).

4. Discussion

4.1. Runoff Reduction

All green roof systems in this study contributed to a reduction in runoff by 23~80% compared to the control (plots with standard frame only) under heavy rain. In addition, thicker substrate was associated with better runoff reduction capacity. The difference might be due to water retention capacity of the substrate, which increases with growing media depth [12]. The retention effect of the substrate was found to be 1.6~3.6 times greater than that of the drainage layer/protection mat [31]. Studies have shown quantitative benefits associated with green roofs, including a decrease of runoff peak discharge [32], a delay in runoff occurrence [31], and a reduction in runoff volume [33]. These results indicate that green roofs can reduce runoff compared to common roofs [28]. Green roofs with 200 mm and 150 mm growing media depth are 42.8~60.8% and 13.8~34.4% effective in reducing runoff for, respectively [9]. VanWoert et al. [33] have concluded that an extensive (shallow) green roof could retain 96% of total light < 0.2 cm, 83% of total medium 0.2~0.6 cm, and 52% of total heavy precipitation > 0.6 cm. Mentens et al. [34] have analyzed 18 studies and demonstrated a clear correlation between media depth and precipitation retention. Previous analyses have shown that retention capabilities on an annual basis may range from 45% for extensive green roofs (substrate depth: 10 cm) to 75% for intensive green roofs (substrate depth: 15 cm), while traditional and gravel roofs can only have retention capabilities of 15% and 25%, respectively [35].

Interestingly, in the current study, runoff retention rate decreased with increasing rainfall intensity for the same adsorption substrate depth. As the maximum rainfall intensity increased, the retention decreased much more rapidly in the green roof system without vegetation. Carpenter et al. [36] have reported that a high−intensity rain event can promote saturation of the roof substrate, which can decrease its retention capacity from 98% to 88%. Therefore, increasing growing media depth alone could not substantially increase water retention [35]. Moreover, the effect of reducing runoff of vegetated green roofs was 1.5 times higher than that of non−vegetated plots. These results indicate that green roofs vegetation can also make a positive contribution to runoff reduction. Greater retention was observed, especially in vegetation plots with thicker substrates, consistent with a previous report [29], although vegetation played a less important role in aiding water retention than the media in extensive green roofs [33]. Data presented in the literature showed that the retention capacity of green roofs typically fell within the range of 40~80% [37] or 30~86% [38]. Peak runoff volume can be decreased by 22~93% or 60~80% [28]. These three studies have also demonstrated that vegetated systems could retain more stormwater than their non−vegetated counterparts. In addition, green roof vegetation plays a more important role in enhancing the evaporation rate, which can reduce rainfall runoff by up to 50% [4]. Differences in evapotranspiration rates can also lead to decreased runoff reduction [36]. In this study, the decrease of runoff was significantly related to the substrate depth, with vegetation showing slightly higher decrease in runoff than no vegetation. These results suggest that vegetation has a better effect than non−vegetation on runoff retention during heavy rains when the substrate depth is the same.

4.2. Runoff Quality (pH, EC, Turbidity, and Heavy Metals)

The pH, EC, and turbidity of runoff from vegetated or non-vegetated plots were significantly higher than those observed for the control. In particular, changes in pH suggested that green roofs had the potential to neutralize acidic properties of runoff. Neutralization is an important environmental benefit that contributes to lowering the degree of acidification of natural water [12]. Green roof structure, as a filter layer, reduces acid rain damage by raising the pH levels [39], which meet the standard V (pH 6~9) of Environmental Quality Standards for Surface Water [22]. This is consistent with the findings of earlier research studies [12,23,27,40,41,42]. Acid rain pollution is caused by the release of numerous precursors such as acid anions (SO₄²⁻ and NO₃⁻) from road traffic and industrial emissions [43]. Depending on their shape and size, plant species can sequester air pollutants and consume carbon dioxide to develop their vital function [15]. In general, improved nutrient availability, due to acid rain in the soil promotes mineralization of nutrients (such as Mg²⁺, K⁺, and Ca²⁺) and expands available pool in the soil. While these mechanisms improve soil neutralization efficiency and plant physiology, acid rain in direct contact with leaves can negatively affect leaf performance and chemical composition [44].

Interestingly, vegetated−green roofs with lower substrates were found to be the least sources of dissolved ions based on EC. Beeham and Razzaghmanesh [27] have also observed that EC values in vegetated beds are less than in non−vegetated systems, with intensive systems producing higher EC values than extensive system. This result indicates that vegetation tends to utilize nutrients for growth and uptake other non-essential ions [26]. Meanwhile, Morgan et al. [45] have reported that growing media have greater effect on turbidity than plants. Because the growing medium used for green roofs might release more quickly than what the media can store or the plant can uptake, increased turbidity in the runoff might have resulted from non-vegetated plots [46].

Gong et al. [22] have found that vegetation can reduce turbidity by 27~71% in the first watering. On the other hand, there were no significant statistical differences between outflow turbidities of all non-vegetated roofs and vegetated extensive beds from vegetated groups [27]. Vijayaraghavan et al. [23] have observed that concentrations of most chemical compounds in runoff from green roofs are the highest at the beginning of rain events, showing gradually decreases over time. Concentrations of chemical compounds in green roof runoff strongly depend on the substrate material used in the green roof and the rainfall [47]. In addition, substrate with less organic matter shows better water quality performance [22].

In this study, a significant negative relationship was found between heavy metal concentration and substrate depth. Concentrations of heavy metals in green roof are generally lower than concentrations of heavy metals in urban runoff from hard surfaces [12]. Vijayaraghavan and Raja [26] have indicated that high sorption capacity of green roofs substrate exhibits high removal efficiencies (greater than 97%) for heavy metals ions. Particularly, Mn and Cd in the runoff were affected by vegetation. These results indicate that green roofs with vegetation could effectively reduce various heavy metal concentrations from runoff. The substrate on green roofs can filter out heavy metals, which can then be taken up by plants, thus improving water quality [48]. Liu et al. [25] have reported that the effect of plant uptake on runoff pollutants from green roof is plant species dependent. In this study, Kentucky bluegrass (Poa pratensis) vegetated plots had significantly lower heavy metal concentrations in runoff than non−vegetated plots and controls, similar to those reported by Vijayaraghavan and Joshi [14]. Previous studies have found that green roofs can increase nutrient concentration, but reduce concentrations of other common runoff pollutants including metals [41]. Metal-spiked artificial rain events have revealed that green roofs can act as a sink for various metal ions and generate better runoff [26]. These changes might be due to the removal or fixation of contaminants in soil and water through absorption, volatile decomposition, and stabilization by plants to purify the environment [49].

Rainwater is generally considered as non-polluted, although it might be acidic. It contains other pollutants depending on local pollution sources and prevailing winds [12]. They usually exist in the atmosphere as aerosols, which can be deposited into soil and water by rainfall [49]. Thus, dissolved various chemicals in runoff can provide valuable information about air pollutants and short-term environmental changes [50]. In particularly, dissolved metals can be more toxic to aquatic life [5]. Plants need mineral nutrients to grown. After plant uptake, the substrate binding site becomes empty, allowing the substrate to adsorb additional elements from the incoming contaminated water [26]. From the standpoint of runoff quality, green roofs have potential to control air quality of pollutants by filtration and adsorption through vegetation and substrates [39]. In particular, vegetation has shown the potential to continuously absorb and immobilize various contaminants [7]. Green roofs also contribute to the quality of runoff by reducing the amount of dust, pollutants and nutrients that would be sent to the sewage systems and waterways [15]. If the substrate of green roofs is saturated and the atmospheric humidity rises, it can become a source of pollution [10,12,46,47]. However, the extent of loss is closely controlled by the runoff reduction [36]. In other words, high water retention capacity, green roofs limit pollution exports for most storm events [10]. Therefore, vegetated−green roofs have the potential to be used as pollutant sinks related to phytoremediation in urban environments.

5. Conclusions

Efficient green roofs in urban areas that experience heavy rainfall during certain periods of time can help mitigate pollutants. In this study, combined effects of green roof vegetation and substrate depths (100, 200, and 400 mm) according to rainfall intensity on runoff reduction and air pollutant purification were investigated. The runoff had a negative relationship with the substrate depth, while runoff reduction increased significantly with increasing substrate depth. For the same substrate depth, vegetated green roofs exhibited slightly higher runoff reduction than non−vegetated green roofs. The pH, electronic conductivity (EC), and turbidity of the runoff were positively related to substrate depth. The pH, EC, and turbidity of runoff from vegetated or non−vegetated plots were significantly higher than those observed for the control. However, vegetation had no significant effect on the pH or turbidity, except for EC. On the other hand, heavy metals (Cu, Zn, Mg, and Cd) in runoff from vegetated plots were all significantly lower than those in non−vegetated plots or controls. These results underscore the importance of substrate depth in improving runoff retention over vegetation. Additionally, our research showed that vegetated green roofs can neutralize acid rain to stabilize pH and effectively reduce heavy metals in runoff. In particular, vegetated green roofs with heavy metal removal capabilities can be implemented as pollutant sinks in cities with a temperate and monsoon climate, where heavy rainfall is concentrated during certain seasons. Nevertheless, these results might not be sufficient to explain the dynamic quantitative behavior of green roof systems in different climatic conditions or hydrological applications of green roofs. Therefore, further investigation and long−term monitoring field experiments are needed to determine substrate moisture dynamics, evapotranspiration, and runoff processes of green roofs.