Potential of bioretention plants in treating urban runoff polluted with greywater under tropical climate

Bioretention systems are among the most popular stormwater best management practices (BMPs) for urban runoff treatment. Studies on plant performance using bioretention systems have been conducted, especially in developed countries with a temperate climate, such as the USA and Australia. However, these results might not be applicable in developing countries with tropical climates due to the different rainfall regimes and the strength of runoff pollutants. Thus, this study focuses on the performance of tropical plants in treating urban runoff polluted with greywater using a bioretention system. Ten different tropical plant species were triplicated and planted in 30 mesocosms with two control mesocosms without vegetation. One-way ANOVA was used to analyze the performance of plants, which were then ranked based on their performance in removing pollutants using the total score obtained for each water quality test. Results showed that vetiver topped the table with 86.4% of total nitrogen (TN) removal, 93.5% of total phosphorus (TP) removal, 89.8% of biological oxygen demand (BOD) removal, 90% of total suspended solids (TSS) removal, and 92.5% of chemical oxygen demand (COD) removal followed by blue porterweed, Hibiscus, golden trumpet, and tall sedge which can be recommended to be employed in future bioretention studies.


Introduction
Rapid urbanization in developing countries has brought several negative impacts to the water-receiving bodies due to draining from agriculture operations, rapid commercial development, and industrial revolutions (Goh et al., 2019;Liu et al., 2014). Runoff in these areas is often polluted with greywater, domestic sewage, and industrial wastes, and the opportunities to control the sources of pollution sometimes are constrained and difficult to achieve (Müller et al. 2020).
Studies conducted by Chow and Yusop (2014), Hongbing et al. (2012), and Wang et al. (2017) showed that urban runoff in developing countries has more pollutants compared to developed countries.
Since the 1990s, various stormwater best management practices (BMPs) have been introduced to treat urban runoff, and bioretention is one of the most popular systems due to its flexibility in size, design, and vegetation selection. Vegetation such as trees, woody shrubs, herbaceous plants, or grass has played an essential role in bioretention to remove pollutants through the phytoremediation process (Read et al. 2008). In temperate countries like the USA and Australia, various guidelines with comprehensive plant selection have been identified and developed, providing detailed information on plant characteristics and removal abilities (NJDEP 2009;PGC 2007;Water By Design 2014). However, there is a lack of information or guidelines for developing tropical countries regarding plant selection and plants' nutrient removal abilities in their stormwater BMP guidelines. Urban Stormwater Management Manual for Malaysia (Manual Saliran Mesra Alam, MSMA) has suggested a list of potential plants in the manual; however, there was a lack of explanation and description regarding the suitability and efficiency of tropical plants to be used in stormwater BMPs, especially in the bioretention system. Although many studies have discussed about treating runoff pollutants using bioretention (Randall and Bradford 2013;Read et al. 2008;Wu et al. 2017), there is a lack of studies to explore the ability of bioretention plants to withstand extreme pollutants, especially under tropical climates. Goh et al. (2017) have studied the application of the bioretention system with enhanced bioretention media under tropical region; however, it is limited to a single plant species (Hibiscus) and specific pollutant removal (total suspended solids (TSS), total nitrogen (TN), and total phosphorus(TP)). The main idea of their research is to study the effectiveness of different enhanced media in treating nutrient-rich pollutants.
Thus, this study focused on the performance and effectiveness of different bioretention plants in treating urban runoff polluted with greywater by assessing additional important parameters such as biological oxygen demand (BOD), chemical oxygen demand (COD), and ammoniacal nitrogen (AN), despite verifying the results of the previous studies (TSS, TN, and TP). This research will widen the scope of urban runoff treatment by providing additional information on treatment strategies to deal with runoff polluted with various sources of pollutants and the suitable plant species to be employed in the phytoremediation process in bioretention facilities.

Vegetation selection
In this study, ten native tropical plant species were chosen based on the suggestion by National Landscape Department (Jabatan Landskap Negara, JLN) and MSMA for stormwater BMPs (DID 2012), which is shown in Fig. 1c. The ten selected plants are Hibiscus rosa-sinensis L., Heliconia rostrata Ruiz & Pav, Alternanthera ficoidea (L.) P.Beauv., Allamanda schottii Pohl, Stachytarpheta jamaicensis (L.) Vahl, Chrysopogon zizanioides (L.) Roberty, Hippeastrum puniceum (Lam.) Voss, Carex appressa R.Br., Heliconia marginata (Griggs) Pittier, and Canna indica L.. The species were selected based on availability, ease of maintenance, low maintenance costs, easy propagation, and resistance to periodic inundation. The plants were cultivated in a nursery to a required height and with healthy root development before being transferred into mesocosms for the experiments.

Mesocosm fabrication
A total of 32 mesocosms were used in this study, by which 30 of them were vegetated mesocosms (10 species, replicated three times) and two of them were non-vegetated mesocosm, used as a baseline study. Each mesocosm was equipped with 500 mm of filter media comprised of 60% sand, 20% topsoil, and 20% compost according to the standard bioretention composition recommended by MSMA (DID 2012). Gravels with 10-mm thickness were placed at the bottom of the mesocosm, and a layer of geotextile was placed between the filter media to prevent clogging at the outlet pipe, as shown in Fig. 1a. For inlet flow, a perforated rubber pipe was connected to the soil surface to ensure the inlet flow was evenly distributed and 20 cm of freeboard is left to allow ponding of influent in the mesocosm before the sample collection.
A total of 32 containers (83-cm height and 48-cm diameter) were modified from used chemical barrels that have been thoroughly washed and sterilized (Fig. 1b). The mesocosm with the layout shown in Fig. 1c was tested under a constructed greenhouse, covered by a clear and transparent roof in Universiti Sains Malaysia Engineering Campus, Penang.

Influent runoff selection
Urban runoff was collected weekly from the main drain near USM Engineering Campus, Nibong Tebal, Penang, Malaysia. The selected location received a mixture of runoff from the main road and greywater from nearby shop lots, residential areas, and urban agriculture. The polluted runoff is collected and stored in two units of 1-m 3 tanks. During the experiment, the polluted runoff was directed to clear tanks with a 45-L marking to ensure each mesocosm received the same amount of influent.

Experimental procedures and data analysis
Before the experiment, all mesocosm were irrigated weekly for 2 weeks with desalinated tap water for the plant adaptation and settling of bioretention media in the mesocosm. One week before the test, the mesocosm was flushed using polluted runoff to ensure the equilibrium condition of the plants and soil. Each mesocosm was fed with 45 L of influent per week for 6 weeks during the test.
Influent samples were collected weekly, and effluent was collected on weeks 2, 4, and 6, with retention of 24 h after the influent was released into mesocosm for water quality analysis. Effluents from each mesocosm were collected in a 1-L bottle and kept cool at 4 °C for further lab testing. Water quality analysis tests such as biochemical oxygen demand (BOD), chemical oxygen demand (COD), total suspended solids (TSS), total phosphorus (TP), total nitrogen (TN), ammoniacal nitrogen (AN), nitrate, and nitrite were carried out for the effluent samples. Humidity and wind speed were not measured in this study; however, the temperature was monitored during the influent feeding and sampling period.
The removal efficiency of the pollutant was calculated and determined by using the equation below: where:  The overall water quality status was determined using the water quality index (WQI) method by the Department of Environment Malaysia (DOE). The equation is shown below: where: Analysis was carried out by generating box plots in Microsoft Excel 2016 to distinguish the concentration of influent and effluent for different types of vegetation as well as the removal efficiency of the plants. The overall performance of plants was discussed based on BMP manuals for TSS, TN, and TP and water quality index performance for BOD, COD, TSS, and AN.
One-way ANOVA together with post hoc analysis (Tukey HSD) was conducted on the plant species for each pollutant monitored to study the existence of variance between the mean values of treated samples for different plant species. The null hypothesis was constructed and will be rejected if the p-value is less than 0.05 which indicates that there is a statistical difference in the mean values of different plant species. The vegetated mesocosm was compared to the non-vegetated mesocosm. Subsequently, an overall ranking of the ten plant species was tabulated and generated from their mean removal efficiency as measured by the significance level of the studied water quality parameter when compared to the controlled mesocosm via IBM SPSS 26 by totalling the p-values by using the equation below: where: Total score = pTSS + pTP + pTN + pCOD Since the consideration is based on both stormwater BMP manual and water quality index, the equation is calculated by breaking them into 4 categories (solids-TSS, nutrients-TP and TN, and organics-COD) to ensure a fair comparison is achieved by taking into account all the pollutant requirement stated in manual and WQI. AN was excluded from the calculation as it was a part of the TN element and it will be unfair to include both to determine the total scores. Similarly, BOD was neglected in this equation, since COD exhibit both non-biodegradable and biodegradable organic matter. Though dissolved oxygen (DO) and pH are listed in WQI calculation, both were also omitted as they cannot be studied using the removal efficiency method.

Pollutant removal efficiency across various BMP manuals
Various BMP guidelines were introduced to assess the efficiency of stormwater BMPs in different countries based on their climate region and suitability. The pollutant removal requirement across selected BMP guidelines from other countries (DEP 2006;NJDEP 2009;PUB 2011;DID 2012;Water By Design 2014) are presented in Table 1. MSMA from Malaysia has the highest removal requirement for TN (50%), New Jersey Stormwater BMP manual has set the highest removal requirement for TSS (90%), and Pennsylvania Stormwater BMP Manual from the USA sets 85% of TP removal. In this study, the removal efficiency of plants was compared with both Malaysia's standard requirement as well as the highest removal requirement set by other manuals.

TP removal
All vegetated mesocosm managed to perform well by removing 85% of TP compared to non-vegetated mesocosm from higher influent concentration (6.5-8.5 mg/L) as shown in Fig. 2. This result has also surpassed the TP removal requirement set by Malaysia, Singapore, Australia, and the US BMP Guidelines (Table 1). An analysis by Barrett et al. (2013) found that big muhly and buffalograss increased nutrient removal from synthetic stormwater. Findings from this study show that vegetation plays a vital role in eradicating TP from the system, with removal rates of 77 to 94%, which supports the data from this study. B01-Hibiscus is the best-performing plant in eliminating TP with the highest removal rate (94.3%), at the same time achieving the highest TP removal requirement set by DEP (2006). The observation from this study is well supported by Ali et al. (2021) and Goh et al. (2017), where the authors declared that Hibiscus is ideal to be implemented in bioretention systems because of its ability to withstand drought and high removal efficacy.

TN removal
Nitrogen removal of all mesocosm achieved the highest TN removal requirement of 50%, set by MSMA (DID 2012) which recorded a range of 58-86% removal as shown in Fig. 3. This data is comparable with the study conducted by Goh et al. (2017), where the TN removal lies within the range of 50-80%. Statistically, all the vegetated mesocosms were observed to have a significant difference (p < 0.05) in removing TN (Table 4) compared to the control mesocosm except lobster claw (B02) and crane lily (B09). Although it is complex and challenging to remove nitrogen from influent due to its complex cycle and mostly soluble suggested by Ali et al. (2021), Hatt et al. (2009), Osman et al. (2019, and Zinger et al. (2013), this study shows a positive and consistent result in term of removing N from the influent. All the vegetated mesocosms outperformed the control mesocosm in removing TN from the influent. This data is well supported by Read et al. (2008) and Osman et al. (2019) where the author mentioned that vegetation plays a vital role in removing nitrogen from runoff via an assimilation process where the inorganic nitrogen is extracted out and accumulated in plant biomass. Vetiver (B06) showed the highest mean removal of TN (86.4% compared to other vegetated mesocosms 70-85%) due to its root size and structure and plant biomass. Vetiver is proven to equip with an extensive, Table 1 Pollutant removal requirements across different BMP manuals *80% removal or less than 10 mg/L (90% of all storm events), **45% removal or less than 0.08 mg/L (90% of all storm events), ***45% removal or less than 1.2 mg/L (90% of all storm events); % that is in bold indicates highest removal requirement penetrating root system that can develop quickly, up to 3-4 m in a short period. This extremely thick root system may penetrate the entire filter media and promote aerobic bacteria living that boost nitrogen uptake (Seroja et al. 2018;Truong et al. 2019).

TSS removal
The concentration range of TSS in inf luent is 128-207 mg/L, which is comparable with average urban runoff (150 mg/L) (LeFevre et al. 2015). Only a few mesocosm managed to fulfill the targeted TSS removal set by MSMA (80%) that is B01-Hibiscus, B04-golden trumpet, B05-blue porterweed, and B06-vetiver, which recorded TSS removal of 81.9%, 86.3%, 88.8%, and 89.5% respectively. B06-vetiver almost fulfill the 90% TSS removal (Fig. 4), and the highest removal criteria set by New Jersey Stormwater BMP Manual (NJDEP 2009). Statistical analysis shows that all the mesocosm performed well in removing TSS compared to the unvegetated mesocosm as the p-value was observed to be < 0.05. This result showed a promising outcome in removing TSS from urban runoff using tropical plants. All the mesocosms achieved over 60% of TSS removal from the influent, as filtration is the primary mechanism involved in TSS removal. Bioretention eliminates TSS through sedimentation on the surface, and

Pollutant removal efficiency based on the water quality index (WQI)
According to the National Water Quality Standards for Malaysia (NWQS), the water quality index (WQI) is also used to assess the amount of pollution in water bodies besides the BMP manual. Dissolved oxygen (DO), biochemical oxygen demand (BOD), chemical oxygen demand (COD), ammoniacal nitrogen (AN), suspended solids (SS), and pH levels are all factors considered by the WQI and classification is as shown in Table 2.

BOD removal
Apart from physical mechanisms (such as filtration), BOD removal relies on biomass growth in the developing medium, which then absorbs BOD. From the experiment, it was shown that some plants could reduce the BOD concentration from upper-class V (63-80 mg/L) to class IV (B04-golden trumpet) and class III (B05-blue porterweed and B06-vetiver) whereas the rest of the plants and control mesocosm could only reduce BOD to lower class V (21-27 mg/L), as presented in Fig. 5. The biomass accumulates with time, enhancing the bioretention system to reduce the concentration further down to class II and class 1. This finding is significant because it suggests that the bioretention system with plants may require time to mature (Khan et al. 2012). Though BOD removal could not achieve the targeted class 11 (1-3 mg/L) set by DOE (2020), the mean removal efficiency of certain plants looks promising in eliminating BOD from influent. Plants such as B01-Hibiscus, B04-golden trumpet, B05-blue porterweed, B06-vetiver, and B10-canna lily recorded > 80% of BOD removal (80.6%, 82.7%, 85.4%, 89.8%, and 80.6%, respectively) from the influent, which is better than the previous results reported by Hunt et al. (2008) (63% by average). Results from a previous study conducted by Takaijudin et al. (2021) are on par with these results, where the author observed a maximum of 88% of  BOD removal. This indicates that aerobic decomposition by bacteria in the filter media is vastly improved, facilitating BOD removal from the mesocosms (Fowdar et al. 2017). Supporting this notion, higher BOD concentration in influent might undermine the effectiveness of bioretention system; hence, the removal efficiency should be taken into consideration when evaluating the performance of plants in the bioretention system.

COD removal
B05-blue porterweed and B06-vetiver recorded the highest removal efficiency of more than 90%, from class V (115-145 mg/L) to class 1 (< 10 mg/L), and the rest of the vegetated mesocosm was able to improve COD to class II (10-25 mg/L) as shown in Fig. 6. These results are slightly more significant than the observation by Qiu et al. (2019) (84.1%) in their studies of the enhanced removal of nutrients in Beijing, China. Similarly, Hong et al. (2017) observed > 90% of COD removal in their pilot-scale study to treat urban runoff in Korea. The best top 4 ranked plants were observed to have a significant difference in removing COD compared to the control mesocosms (p < 0.05) as shown in Table 4. COD is typically removed from a bioretention system by adsorption, filtering, microbial breakdown, and plant absorption. Adsorption plays a crucial part in these activities because it provides a relatively immediate capture mechanism and a momentary storing reservoir to aid slower activities like biodegradation and plant uptake (Hong et al. 2017). Massive and branching root structures below ground for both vetiver and blue porterweed could have enhanced the pollutant removal through plant uptake.

TSS removal
All the vegetated mesocosms improved the water quality as shown in Fig. 7 from class IV (150-300 mg/L) to class II and below, except for B02-lobster claw (class III). B06-vetiver and B05-blue porterweed improved the water quality from class IV to class 1, better than other plants because of their aboveground plant biomass, which can trap and accumulate more sediments before the influent penetrates through the filter media. These results were well supported by past studies where TSS removal was observed to be consistently high (> 90%) regardless of the research design and setup (Barron et al. 2020;Blecken et al. 2010;Bratieres et al. 2008;Fowdar et al. 2017;Hatt et al. 2009;Hsieh and Davis 2005;Lim et al. 2021;Søberg et al. 2020).

AN removal
Influent AN concentration (6.5-8.5 mg/L) in this study falls under class V of the Water Quality Index (WQI) Malaysia (> 2.7 mg/L), which is generally higher compared to the typical AN concentration in Melbourne (0.27 mg/L) (Taylor et al. 2005) and in Malaysia (2.59 mg/L) (Chow and Yusop 2014) for mixed residential and commercial areas. Despite all vegetated mesocosms reducing AN to class III (< 0.9 mg/L), there was one mesocosm that managed to increase the water quality from class V to class II which is B05-blue porterweed. Despite AN's excellent removal efficiency (> 80%), the effluent concentration for most mesocosms is still above class IV, possibly due to high influent concentration, as shown in Fig. 8. This observed result (69-89% of AN removal) has outperformed the study conducted by Blecken et al. (2007) and Geronimo et al. (2015), where AN removal was reported to be 51.7% and 40-54% respectively and falls short behind Zhang et al. (2011) andFan et al. (2019) where the author reported 95% and > 95% of AN removal respectively.

Overall performance of plants based on the water quality index (WQI)
According to DOE (2020), the targeted water quality based on WQI is class II and below which is classified as unpolluted. Based on Table 3, all the plants improved the water quality from class V (influent) to class II and class III. However, only two plants could further ameliorate the water quality to the targeted class II (76.5-92.7) which are B05-blue porterweed and B06-vetiver. These plants performed better in removing the major pollutants from the influent as both plants recorded > 85% of removal in all water quality tests.

Ranking of plants based on the overall performance
The ANOVA conducted on the effluent samples of different plant species provided the frequency distribution (F) data 4.628, 10.595, 11.667, and 8.859 respectively for COD, TSS, TP, and TN at a 95% of confidence level. The critical value for F was found to be 2.30, which is lower than the F value, thus supporting the rejection of the null hypothesis. This confirmed the influence of plants on pollutant removal performance in the bioretention system. All the selected plants were ranked 1-10; 1 was the most effective, and 10 was the least effective among all the plants. Based on Table 4, it was indicated that B06-vetiver ranked first in terms of overall performance, followed by B05-blue porterweed with only a slight difference in the total score. Truong et al. (2019) reported that the vetiver's root system is well-formed and exceptionally deep and robust, allowing it to withstand harsh circumstances such as high nutrient loads. This statement supports our results as vetiver was observed to have high removal of pollutants despite having a high influent concentration. Delis et al. (2015) and Seroja et al. (2018) also suggest that vetiver can be a great asset to be employed in the bioretention system. The author reported that long roots and stems of vetiver could enhance the removal of nutrients. The second-ranked blue porterweed's ability to remove pollutants was never tested in previous studies. Thus, this plant has become an excellent addition to bioretention systems in the tropical climate with a bigger aboveground structure with branching stems and leaves as well as a deeper and massive root structure which may facilitate the pollutant removal in this study.
The next three in the ranking were B01-Hibiscus, B04golden trumpet, and B08-tall sedge. In Malaysia, the Hibiscus is the most popular plant due to its excellent resilience to dry weather and good removal efficiency; it is ideal for use in bioretention systems (Ali et al. 2021). The effect of the plant on nutrient removal efficiency was studied by Goh et al. (2015), who discovered that the enhancement media and plant increased TP removal from 84.9 to 93.3% and TN removal from 57.4 to 80.4%. The fourth-ranked golden trumpet is also an excellent addition to the list of potential tropical plants in the bioretention system as its long stems and numerous leaves enhanced pollutant removal. In addition, the golden trumpet is fast-growing, and the aboveground plant structures certainly impact the bioretention system. Tall sedge is ranked fifth though it is indicated as a superior plant in past research for its excellent performance in stripping nutrients via the bioretention system (Bratieres et al. 2008;Ellerton et al. 2012;Fowdar et al. 2017;Herzog Read et al. 2010). Tall sedge performed better in removing nutrients (TN and TP), which was observed to eliminate 84.3% and 93.4%, respectively. However, it was not enough as it performed slightly lower in BOD and COD removal than other tropical plants. The current study was done in a tropical climate and for urban runoff contaminated with greywater rather than stormwater treatment, which might explain the opposite outcome. Another explanation might be increased evapotranspiration from tall sedge, with a large stem, where all plants received the same amount of influent, resulting in a root zone with low water content and high organic concentration. Plants with small stems and leaves like B03-Alternanthera, B07-Amaryllis, and B09crane lily were underperforming in removing pollutants as all three were ranked in the bottom 4. To summarize, plants with higher aboveground biomass, extensive and deep root structures, and fast-growing capabilities are some of the key physical characteristics in plants which greatly influence the removal of pollutants in a bioretention system under tropical climate.

Conclusion
The pollutant removal performance of ten different tropical plants was evaluated and compared with various stormwater BMP guidelines from other countries and the WQI method from the Department of Environment Malaysia (DOE). Results showed that most plants could fulfill the highest requirement set by guidelines from other countries and improve the water quality from class V/IV to class III/II, and even some plants were able to achieve class 1.
The 10 tropical plants were ranked based on their performance via one-way ANOVA based on the pollutant removal effectiveness. B06-vetiver performed well and ranked first, followed by B05-blue porterweed, B01-Hibiscus, B04golden trumpet, and B08-tall sedge. Plant species with high stems, leaves, and massive roots ranked in the top 5. This research focuses mainly on the performance of the pollutant removal by tropical plants and suggests the best-performing plants to be used in the bioretention system in tropical countries. The plant's growth rate, as well as their concentration accumulation in plants, was not studied. This data will help construct bioretention facilities in the future by providing complete information about potential plants for treating urban runoff in the tropical region. In the future, it is recommended to compare the performance of various vegetation planted together for on-site application for a bioretention pilot study in the tropical region.