Enhanced greywater treatment in a two-stage green wall system by incorporating waste-derived adsorbents

Greywater reuse offers a sustainable approach to alleviating freshwater scarcity, but concerns about treatment reliability limit its adoption. This study investigates a two-stage green wall system for greywater treatment, integrating vegetated units planted with Carex secta and vertical flow post-treatment columns (PTCs) packed with waste-derived adsorbents—calcinated kina (sea urchin) shell, ZnCl 2 -activated apricot stone biochar, and natural zeolite. The system’s performance was evaluated for removal of organic matter (COD, TOC), nutrients (ammonia, nitrate, nitrite, phosphate), and xenobiotics (anionic surfactants, aromatic compounds) under varying hydraulic and pollutant loads. Substrate characterization, plant health assessment, microbial analysis, and physicochemical water quality monitoring were used to elucidate plant–microbe–media interactions. PTCs effectively polished residual pollutants after partial removal in green wall units, achieving removal rates of 34% of phosphate, 99.8% of ammonia, 95% of surfactants, and 25%–37% of aromatic compounds (UV (294)). While nitrate leached from green wall units (−34%), PTCs mitigated this by removing 64% of nitrate. TOC and COD were further reduced by 17% and 14%, resulting in overall removal efficiencies of 95% and 96%, respectively, across the two-stage system. Under stress conditions of 2-week drying followed by high pollutant loading, PTCs maintained pollutant removal, whereas green wall units showed reduced nutrient removal. Across the planted green wall, ammonia increased from 0.02 to 1.63 mg/L and phosphate from 3.42 to 4.44 mg/L. The non-planted control deteriorated more strongly, with ammonia increasing to 8.88 mg/L and phosphate to 8.32 mg/L. After stress recovery, vegetated units rapidly regained treatment efficiency, while controls released ammonia tenfold and phosphate twofold. This novel integrated treatment design demonstrates a scalable, low-maintenance solution for effective on-site greywater treatment and reuse.

Rising economic and social development is resulting in increasing water demand on already constrained freshwater resources (Church et al., 2015). Treated greywater reuse presents a promising strategy to reduce pressures on conventional water supplies (Galvão et al., 2024) and on wastewater treatment facilities, while conserving potable water for critical needs (Madungwe and Sakuringwa, 2007). Greywater, which is distinct from toilet blackwater, results from bathing, laundry, and dishwashing. It accounts for 50%–80% of total domestic wastewater volume (Eriksson et al., 2002) and contains up to 30% of the organic carbon load and between 9%–20% of the nutrients (Fountoulakis et al., 2016) but has a lower pathogenic risk. These attributes support greywater treatment and its potential reuse (Mohit and Remya, 2024; Patil et al., 2022; Paulo et al., 2013). However, its wider adoption remains constrained by variable regulations, potential health risks, inconsistent treatment performance, maintenance requirements, and uncertain costs (Khajvand et al., 2022; Vuppaladadiyam et al., 2019). These barriers highlight the need for reliable, low-maintenance treatment systems that can produce consistently safe effluent for decentralized reuse (Benami et al., 2016; Khajvand et al., 2022). Nature-based solutions (NBS), such as constructed wetlands, green roofs and green walls, have gained prominence in treating greywater for their low-energy demands, small footprint, and co-benefits including biodiversity enhancement and thermal regulation (Arden and Ma, 2018; Pradhan et al., 2019a). Among these, green walls offer unique vertical integration into urban infrastructure (Gattringer et al., 2016), overcoming spatial limitations associated with horizontal flow wetlands (Gholami et al., 2023a). They hold significant potential for household greywater reuse, as greywater can be safely applied for fertigation, providing essential nutrients to support plant growth while simultaneously treating the wastewater at source (Chung et al., 2021). However, green walls are unable to remove all nutrients, particularly phosphorus, and can leach pollutants under inherent high loading or stress conditions such as drying and rewetting cycles (Prodanovic et al., 2019a). Prodanovic et al. (2019a) observed total phosphorus (TP) removal in greywater of 27%–53%, with concurrent nitrate and nitrite leaching, in systems planted with shallow-rooted or poorly performing species (e.g., Ophiopogon japonicus). Fowdar et al. (2017) reported increasing phosphorus leaching over time in systems planted with Phormium tenax, with TP removal of only 7%. Lakho et al. (2021) reported chemical oxygen demand (COD) and yellow discoloration leaching from organic soils, indicating release of recalcitrant compounds under operational stress. Gholami et al. (2023b) reviewed over 30 studies on the topic finding that these pollutant removal limitations are usually attributed to media saturation, insufficient contact time, or the use of sub-optimal plant and substrate combinations.

Adsorbents, like zeolite and biochar, are frequently applied in the polishing stages of conventional wastewater treatment (Metcalf et al., 2013) and in principle, could help overcome limitations with greywater treatment in NBS. They have enhanced surface areas and interconnected pores offering high chemical affinity for pollutants. Abd-ur-Rehman et al. (2022) reported 81% ammonium removal from greywater using zeolite with an adsorption capacity of 200 mg/g. Zeolite also has a very high surface area and porous structure that enhances affinity for various xenobiotics (XOCs) (Abd-ur-Rehman et al., 2022). Gholami et al. (2025a) found rapid phosphate precipitation (98%) with calcareous shell adsorbents, while phosphate (67.7%), nitrate (48.9%), sodium dodecyl sulphate (SDS) (98.6%), and TOC (59.8%) were removed with a ZnCl 2 -activated apricot stone biochar in synthetic greywater batch treatment studies (Gholami et al., 2025b). Although showing early promise, incorporating these adsorbents into nature-based systems like green walls to augment pollutant removal efficiency, has not been reported. CK and ZAB were previously evaluated against other waste-derived and commercial adsorbents in dedicated batch studies (Gholami et al., 2025a; Gholami et al., 2025b). CK achieved a maximum phosphate adsorption capacity of 41.98 mg P/g through calcium-mediated precipitation, while ZAB achieved maximum adsorption capacities of 39 mg/g for phosphate, 35 mg/g for nitrate, 200 mg/g for SDS, and 53 mg/g for TOC. These values compare favourably with several reported calcareous materials and biosorbents, although comparisons are influenced by differences in water matrix, pollutant concentration, adsorbent dosage, pH, contact time, and experimental configuration. Broader comparisons of calcinated materials are available in the recent systematic review by Tat Wai et al. (2024).

In the present study, the materials were selected for their complementary functions: CK targeted phosphate through calcium-mediated precipitation, ZAB adsorbed nitrate, surfactants, and residual organic compounds, and zeolite supported ammonium removal through cation exchange. Therefore, a vegetated vertical flow unit was augmented with a subsequent post-treatment column (PTC) containing different waste-derived adsorbents. The system was subjected to stressors including drying–rewetting cycles and high pollutant and hydraulic loading rates to assess its treatment performance under varying realistic conditions. Accordingly, the main objectives of this study were to: (i) test the performance of low-cost adsorbents in enhancing greywater treatment in a two-stage green wall system; (ii) assess the effect of different flow and pollutant loading conditions on pollutant removal; and (iii) characterize the adsorbents before and after greywater treatment to understand pollutant fate. This design aims to enhance nutrient and organic pollutant removal beyond the capacity of existing green walls or other NBS, producing treated effluent suitable for safe non-potable reuse.

Greywater characteristics

Synthetic greywater containing multiple pollutants was prepared using a modified recipe from Diaper et al. (2008), that includes various household products and is tailored to include representative concentrations of nutrients (ammonia, nitrate, phosphate), organic matter (COD, TOC), and xenobiotics (anionic surfactants, UV-absorbing compounds). Pollutant concentrations simulated light greywater, representing bathing and hand basin wastewater following Khanam and Patidar (2022). A full composition of the synthetic greywater is provided in Supplementary Table S1, of the Online Supplementary Material. A household blender was used to homogenize the mixture at low speed. Concentrations were doubled during the second phase of the study to assess the impact of higher pollutant loads (stress condition).

Experimental design and operation

The experimental set-up is depicted in Figure 1. Planted and non-planted primary treatment columns comprise PVC drainage piping (25 cm height × 10 cm diameter, 5 cm freeboard) filled with 17 cm of a perlite and coco coir mixture and topped with 3 cm of LECA for drainage (Figure 1). These materials are low cost, light weight, and suitable for supporting plant growth. Carex secta, a New Zealand native species, was selected for its high nutrient removal capacity evidenced by other studies on Carex’s overall performance in green wall systems (Prodanovic et al., 2019a). Their effectiveness is attributed to traits like higher growth rates, longer root lengths, and extensive root systems, which enhance nutrient uptake in NBS (Gholami et al., 2023a; Prodanovic et al., 2019a). These columns received 800 mL/day of synthetic greywater, corresponding to a wetted-area hydraulic loading rate (HLR) of 89 mm/day—within the optimal range (55 and 100 mm/day) for pollutant removal in green wall systems (Gholami et al., 2023a).

The post-treatment columns (PTC1–PTC3), each 25 cm in height and 3 cm in diameter, incorporated two waste-derived adsorbents, namely, calcinated kina (sea-urchin) shells (CK) and zinc-chloride activated apricot stone biochar (ZAB), along with zeolite. All materials were locally sourced and sieved to ensure uniform particle size: zeolite and CK (1.18–2.00 mm), ZAB (0.60–1.18 mm), coco coir (0.60–2.36 mm), and perlite (2.36–3.18 mm) (Supplementary Figure S1). Each column was packed with 6 cm of a layered mixture of ZAB, zeolite, and CK in a 1:1:1 volumetric ratio, topped with 2 cm of lightweight LECA for flow distribution (Figure 1).

Effluent from the primary treatment green wall columns was collected in the separate reservoirs and subsequently distributed to the three post-treatment columns (PTC1–PTC3) using independent adjustable peristaltic pumps. The pumps operated intermittently to deliver target flow rates of 200, 600, and 1,200 mL/day to PTC1, PTC2, and PTC3, respectively, corresponding to HLRs of 283, 849, and 1,415 mm/day and hydraulic retention times (HRTs) of 16.96, 5.65, and 2.83 h, respectively. Flow rates were monitored regularly and adjusted as required to maintain dosing volumes within ±5% of target values. To ensure equal starting conditions under the elevated loading applied in Phase 2, the adsorbents in all three PTCs were replaced with freshly prepared ZAB, zeolite, and CK at the start of Day 32.

An initial 2-week ‘conditioning phase’ was undertaken with tap water, supplemented with activated sludge from an aeration tank of the municipal Christchurch Wastewater Treatment Plant. This supported plant establishment and biofilm development, promoting rapid nitrification and denitrification once greywater was introduced during the experimental runs.

Experiments were carried out in two phases. In Phase 1 (Days 1–31), synthetic greywater was applied to evaluate treatment performance under expected pollutant loadings. This was followed by a 14-day drying period (no greywater feed or irrigation) to simulate water-stress conditions that may occur while occupants of buildings with green walls are away for periods of time (e.g., holidays). Following that, Phase 2 (Days 32–44) was initiated during which greywater with double the pollutant concentrations was applied to assess the system’s response under high-load conditions. Throughout both phases, greywater was applied in intermittent cycles on a 2-h on/2-h off schedule over a 24-h period, providing 12 h of active pumping with 2-h rest intervals between each application. All experiments were conducted under room temperature (25° C–28° C), which was found to be optimal for successful nutrient removal orchestrated by increased microbial activity and plant uptake under optimal HLR conditions (Gholami et al., 2023a; Gholami et al., 2023b).

Analyses

Water quality measurements

Duplicate water samples were collected from the greywater influent and from the effluent of each column. Key physicochemical parameters were also monitored, including pH and oxidation-reduction potential (ORP) using a YSI pH100 m, electrical conductivity (EC) via a YSI EC30 m, and dissolved oxygen (DO) using a YSI Pro550A optical meter, calibrated under ambient conditions. Chemical oxygen demand (COD) was measured using HACH digestion vials (0–1,500 mg/L), digested in a HACH DRB200 thermoreactor, and analysed via HACH DR3900 spectrophotometer. Ammonia (NH 3 –N) was quantified using HACH Nessler method kits. Nitrate (NO 3 −), nitrite (NO 2 −), and phosphate (PO 4 3-) concentrations were determined using ion chromatography (ICS-6000 HPIC, Thermo Scientific). Anionic surfactants were analysed with HACH TNTplus Vial Test kits as SDS equivalents. TOC was measured using a Shimadzu TOC-L analyser, and aromatic compounds were measured as UV absorbance at 294 nm (UV 294) using a Thermo Fisher UV–Vis spectrophotometer.

Media characterisation

All three adsorbents (CK, zeolite, and ZAB) were characterised before and after greywater treatment as follows. Their surface morphology was examined using scanning electron microscopy (SEM; JSM-IT300LV, JEOL). Their crystalline structure was analysed by X-ray diffraction (XRD) using a Rigaku Smartlab 3 kW diffractometer. The instrument used CuK(α) radiation, and measurements were obtained over a 2θ range of 5°–90°, at a step size of 0.05° and at a scanning speed of 5°/min. In addition, functional groups on each adsorbent were identified using Fourier transform infrared (FTIR) spectroscopy (Spectrum 100, PerkinElmer) over the 400–4,000 cm -1 wavenumber range.

DNA extraction, sequencing and analysis

Genomic DNA was extracted from selected samples collected at the top, middle, and bottom levels of both planted and non-planted columns using the NucleoSpin Soil Kit, and concentrations were quantified via NanoDrop One spectrophotometry (Thermo Scientific). Samples were labelled according to column type and sampling depth. The prefixes “P” and “NP” denote planted and non-planted columns, respectively, while the suffixes “T”, “M”, and “B” indicate samples collected from the top, middle, and bottom sections of the columns. Only DNA samples with A260/A280 purity ratios between 1.8 and 2.0 were selected and analysed at the University of Macau for library preparation and shotgun metagenomic sequencing. The V4 region of the 16 S rRNA gene was amplified using the modified 515F/806R primer pair (Caporaso et al., 2011; Apprill et al., 2015; Parada et al., 2016). This primer set was selected because it provides broad coverage of bacterial and archaeal taxa and has been widely applied in environmental microbiome studies, including wastewater and soil systems, enabling robust community characterization and comparison with existing datasets (e.g., Walters et al., 2016; Zhou et al., 2024). Furthermore, the modified 515F/806R primer set forms the basis of the Earth Microbiome Project protocol, supporting standardized and reproducible microbial community analyses.

Media characterization

Figure 2 presents SEM images of CK (a), zeolite (b), ZAB (c), perlite (d), and coco coir (e) before greywater treatment, while Figure 3 shows the XRD and FTIR characterization of CK, ZAB, and zeolite before and after treatment. SEM imaging revealed the smooth yet porous morphology of perlite, which supports effective hydraulic conductivity, while coco coir exhibited a rough, folded structure conducive to microbial colonization and biological activity (Figure 2e), in alignment with previous studies (Pradhan et al., 2020). Calcination of kina shells at 800° C produced rough, flaky, and porous surfaces (Figure 2a). These were successfully enriched with CaO and Ca(OH) 2, as confirmed by XRD peaks at 47.5° 2θ (Figure 3a). After greywater treatment, CK showed evidence of hydroxyapatite (Ca 5 (PO 4) 3 OH) and dicalcium phosphate dihydrate (CaHPO 4 ·2H 2 O) precipitation, with peaks at 25.9°, 28.1°, 39.8°, and 49.5° 2θ (Figure 3a), consistent with our previous batch study (Gholami et al., 2025a). FTIR analysis further supported the presence of phosphate within the calcinated kina shell layer, demonstrating a strong P-O bond formation band at 1,043 cm -1 post-treatment (Figure 3b). In contrast to the previous batch experiments, where loosely bound calcium flakes detached from the CK surface and were washed away as suspended precipitates (Gholami et al., 2025a), the column setup reported here retained these precipitates between particles. Together, the XRD and FTIR results indicate that phosphate removal by CK was primarily driven by CaO/Ca(OH) 2 dissolution followed by precipitation as hydroxyapatite and dicalcium phosphate, consistent with our previous kinetic and equilibrium analyses (Gholami et al., 2025a).

SEM of ZAB confirmed the development of extensive micro- and mesopores facilitated by ZnCl 2 activation, with XRD peaks at 32°, 34°, and 36°. These peaks indicate ZnO and Zn(OH) 2 phases, confirming the successful zinc impregnation (Figure 2c) (Minz and Mishra, 2023; Sangon et al., 2024). While some minor FTIR and XRD peak changes occurred in the ZAB adsorbent after treatment, the absence of any significant peak changes suggested that contaminants were primarily adsorbed within the porous matrix rather than removed through surface complexation. Based on the developed porous structure of ZAB and our previous batch experiments, pollutant removal likely involved a combination of pore filling, electrostatic interactions, hydrogen bonding, hydrophobic interactions, and π–π interactions (Gholami et al., 2025b).

SEM images of zeolite showed more angular, fragmented particles with high surface roughness (Figure 2b) and interparticle porosity. This may enhance mass transfer, although ammonium removal is primarily driven by internal porosity and material cation exchange capacity. XRD analysis of zeolite showed sharper diffraction peaks around 10°, 15°, 25°, and 30° (Figure 3a), corresponding to a highly crystalline aluminosilicate framework (Djozing et al., 2024). FTIR analysis of zeolite confirmed some changes with the emergence of a small new peak at ∼1,400–1,450 cm -1, which corresponds to N–H bending vibrations of adsorbed ammonium ions (Figure 3b), consistent with previous reports (Bakatula et al., 2015; Yadav et al., 2023). The appearance of a broad band at 3,500 cm -1 — attributed to O–H stretching of hydroxyl groups—also may suggest increased surface hydration or structural hydroxylation associated with ammonium adsorption (Khamidun et al., 2022). New peaks observed at ∼2000 and ∼2,900 cm -1 may correspond to C≡C stretching vibrations and C–H stretching modes, respectively, and could indicate the adsorption of organic matter (e.g., aromatic compounds) (Liu, 2021). The framework Si–O–Si vibrations remained intact, confirming the preservation of structural integrity post-treatment, which is important for longer-term enhanced pollutant treatment within NBS.

Pollutants removal efficiencies

Phosphate and nitrogen removal

Phosphate

Phosphate removal was generally limited in both planted and non-planted columns (compared to the PTCs with adsorbents), with the planted unit achieving 34% removal (effluent: 3.34 ± 0.32 mg/L), compared with only 17% in the non-planted column (4.20 ± 0.38 mg/L; p < 0.001) (Figure 4a). Carex secta is believed to have somewhat enhanced phosphorus removal through a combination of direct plant uptake, microbial assimilation in the rhizosphere, and biofilm-mediated adsorption which are well-documented in green wall systems (Pradhan et al., 2020; Prodanovic et al., 2019a). Pradhan et al. (2019b) found that phosphate removal was 31%–36% lower in unplanted compared to planted columns. Prodanovic et al. (2019a) reported that substrates alone in green walls removed 88% of particulate phosphorus in top levels of green wall biofilters, though its effectiveness in soluble phosphate removal remained limited. Without vegetation, phosphate removal depends on biofilm development, media-driven sorption, filtration, and precipitation, leading to less consistent and generally higher phosphate concentrations, aligned with previous studies (Pradhan et al., 2020; Prodanovic et al., 2017).

The influent greywater (Phase 1) phosphate concentration was 5.05 ± 0.95 mg/L. Most phosphate was rapidly removed in the PTCs (PTC1: low HLR, PTC2: medium HLR and PTC3: high HLR) due to strong alkaline-driven precipitation with Ca(OH) 2 from the CK (Gholami et al., 2025a). This removal behaviour was supported by the CK characterization (Section 3.1). PTC1 maintained complete phosphate removal throughout the study, reflecting slower stoichiometric consumption of Ca(OH) 2 under lower cumulative phosphate loading. However, in PTC3, less phosphate was removed by Day 12 (2.1 mg/L) and in PTC2 by Day 20 (2.04 mg/L) (Figure 4a).

Ammonia

Influent greywater exhibited an average ammonia (NH 3 –N) concentration of 10.11 ± 2.32 mg/L during Phase 1 (Days 1–31) (Figure 4b). Although concentrations in the PTC effluents remained undetectable, this was primarily due to near-complete removal upstream in the green wall units. Nevertheless, the inclusion of zeolite in the PTCs, known for its high cation exchange capacity (CEC) and rapid NH 4 + adsorption (>80–98% removal) (Abd-ur-Rehman et al., 2022; Anangadan et al., 2024; Kumari et al., 2024), likely provided an important polishing step, particularly under elevated ammonia and stress conditions (e.g., Days 32–44; see Section 3.4). As the aqueous ammonia species (NH 3 and NH 4 +) exist in pH-dependent equilibrium (pKa ≈ 9.25) (Yang et al., 2021), most nitrogen in the pH-neutral greywater was present as NH 4 +, which is the form removed through zeolite cation exchange.

The planted column removed 99.8% ammonia (to 0.020 ± 0.041 mg/L; Figure 4b) and 73% TN (10.47 mg/L to 2.86 mg/L, based on sum of ammonia and oxidized nitrate species) compared with the non-planted column which removed 91% ammonia (to 0.92 ± 0.46 mg/L) (p < 0.001) and 45% TN (10.47 mg/L to 5.78 mg/L), highlighting the role vegetation can play in ammonia removal uptake, as was observed for phosphate removal. Pradhan et al. (2019b) similarly found that vegetation contributed approximately 25% to TN removal. TN decreased further to 1.02 mg/L (90% removal) in PTC1, indicating strong polishing within the post-treatment column. Given the low influent ammonium concentrations, the observed TN reduction cannot be attributed solely to ammonium exchange on zeolite. Instead, it likely reflects combined removal pathways including adsorption and microbial transformation processes associated with the mixed adsorbent media, particularly biochar, which may facilitate the removal of dissolved organic nitrogen and nitrate species.

Some plants prefer ammonium as a nitrogen source due to its lower energy needed compared to nitrate assimilation, especially in early stages (Cui et al., 2017). In addition, the role of adsorptive removal by coir with high CEC (Pradhan et al., 2019b) likely contributed to rapid early-stage NH 4 + reduction, particularly in the first 5–7 days (Figure 4b). However, the primary mechanism behind the observed ammonia removal is attributed to the early establishment of microbial nitrification, supported by favourable redox conditions: DO levels of 4–6 mg/L, pH 6.9–7.1, and ORP >150 mV, particularly in the upper media layers (Fowdar et al., 2018); Prodanovic et al. (2023) similarly reported near-complete nitrification (93% ammonia removal) in deep green wall biofilters.

In the PTCs, although nitrate concentrations in the green wall effluent were reduced by 64%, levels increased steadily up to 31day coinciding with the 2-week drying period. This trend in declining nitrate reduction likely show limited capacity to remove nitrate with increasingly pollutant loading. For example, nitrate rose from 1.73 mg-N/L to 14.11 mg-N/L in PTC1 (Figure 4d). The potential mechanisms for nitrate removal include electrostatic interactions, hydrogen bonding, and pore-filling within the high surface area mesoporous matrix of ZAB in PTC columns (Sangon et al., 2024; Varela et al., 2024). However, these physico-chemical pathways likely became saturated over time reflected as reducing nitrate removal. According to Figure 4c, nitrite also remained undetectable in all PTCs, indicating that complete upstream nitrification occurred in the green wall columns and that no microbial nitrification or denitrification processes developed in the post-treatment media. This absence of nitrite also confirms that biologically mediated mechanisms in the PTCs were limited, likely due to the removal of organic carbon which is mainly removed in green wall units (see Section 3.2.2).

In comparison, nitrate accumulation or leaching was observed in both planted and non-planted columns, consistent with efficient nitrification and limited denitrification. The influent had an average nitrate concentration of 9.52 ± 1.14 mg/L. The planted column showed a lower nitrate peak (25.29 mg/L on Day 30; mean: 12.61 ± 4.84 mg/L) (−32%), compared to the non-planted column (peak: 36.81 mg/L on Day 23; mean: 19.80 ± 10.72 mg/L, p = 0.008) (−108%), indicating partial nitrate uptake by Carex secta or enhanced denitrification potentially stimulated by root-derived carbon exudates in lower column regions (Figure 4d). Cui et al. (2017) demonstrated that plant nitrogen species preference can shift during different growth stages, with grasses often transitioning from ammonium uptake in early stages to nitrate uptake later in development. However, further nitrate reduction may have been limited by reduced availability of labile organic carbon in deeper column regions and short hydraulic retention times, which can restrict denitrification potential. While dissolved oxygen measurements at the outlet indicated generally oxic conditions, internal redox conditions within porous media cannot be directly inferred and are likely spatially heterogeneous. The presence of biochar and other porous substrates may also support the formation of localised micro-anoxic zones. Therefore, nitrate transformation pathways in the system are likely governed by a combination of carbon availability, biofilm development, and microscale redox variability rather than uniform oxygen limitation alone. Similar outcomes were reported by Lakho et al. (2022) in a continuous green wall system, linked to high oxygen availability from air exposure on both the front and back surfaces, but nitrate accumulation in the effluent—from 0.7 mg/L at inflow to 5.9 mg/L.

Nitrite concentrations remained low in the planted column (0.26 ± 0.76 mg/L), while the non-planted column exhibited transient nitrite accumulation (peak: 5.29 mg/L on Day 9; average: 1.85 ± 1.54 mg/L) (Figure 4c), likely due to delayed nitrite-oxidizing bacteria (NOB) activity relative to ammonia-oxidizing bacteria (AOB). Similar observations were made by Fowdar et al. (2018). However, nitrite and nitrate dynamics in such systems are influenced by complex interactions among plant uptake, microbial biofilms, and substrate characteristics. Pradhan et al. (2020) reported that, despite substantial TN removal, nitrate leaching and occasional nitrite accumulation (up to ∼1 mg/L) occurred, depending on plant species (e.g., Ruellia brittoniana, Alternanthera dentata) and substrate types (e.g., perlite, coco coir).

Organic matter (COD and TOC) removal

In the green wall units, COD and TOC removal was primarily driven by aerobic microbial degradation in biofilms, supported by oxygen diffusion in upper media layers, along with filtration and adsorption onto porous substrates (Fowdar et al., 2017; Prodanovic et al., 2019a). Media composition and microbial colonization also play a key role, with 40%–70% COD removal reported in systems using coco coir and perlite, especially in upper layers (Prodanovic et al., 2020). The planted column achieved average effluent COD and TOC of 30.69 ± 8.06 mg/L (86%) and 12.30 ± 3.71 mg/L (83%), while the non-planted column recorded 27.52 ± 14.61 mg/L (89%) and 10.98 ± 5.46 mg/L (85%), respectively. Similarly, TOC and COD removal efficiencies in green wall systems reported in literature ranged from 40% to as high as 92%, depending on vegetation species, media type, and system maturity (Fowdar et al., 2017; Pradhan et al., 2019b; Prodanovic et al., 2019b). Although the planted system showed more consistent performance and lower variance, differences were not statistically significant (COD: p = 0.26; TOC: p = 0.38), suggesting that vegetation enhanced system stability but did not markedly outperform filtration and microbial degradation. Pradhan et al. (2019b) showed that plants contributed approximately 10%–15% to COD and TOC removal in green wall systems by enhancing microbial activity and oxygenation in the rhizosphere, while the primary removal mechanism was microbial degradation (Pradhan et al., 2019b; Prodanovic et al., 2019a).

In the PTCs, organic matter (COD and TOC) was effectively polished downstream of the green wall units. During Days 1–31 (Phase 1), influent COD ranged from 165 to 274 mg/L and TOC from 50 to 81 mg/L. COD was reduced to 15.26 ± 3.82 mg/L (93%) in PTC1, 14.13 ± 4.43 mg/L (94%) in PTC2, and 19.39 ± 5.71 mg/L (91%) in PTC3. Corresponding TOC values were 7.95 ± 1.28 mg/L (88%), 7.48 ± 1.43 mg/L (88%), and 8.18 ± 1.57 mg/L (87%) (Figures 5a,b). These results reflect consistent polishing performance attributed to the combined effects of ZAB adsorption, zeolite cation exchange, and CK-driven filtration and precipitation. The final effluent quality aligns with US EPA (2012) non-potable reuse standards, which specify BOD < 30 mg/L, TSS < 30 mg/L, pH 6–9, and turbidity < 2 NTU.

Xenobiotics (XOCs) removal

Similar to organic matter, the PTCs achieved highly effective removal of residual XOCs, including anionic surfactants and aromatic compounds (UV 294). During Days 1–31 (Phase 1), influent anionic surfactant concentrations averaged 59.5 ± 13.7 mg/L, and UV 294 absorbance was 0.16 ± 0.07 AU. Anionic surfactants dropped to ≤0.3 mg/L (>99%) in all PTCs, and UV 294 to <0.045 AU, with PTC1 and PTC2 frequently reaching <0.01 AU (99%) (Figures 6c,d). These outcomes reflect similar pollutant removal mechanisms to those described for organic matter removal, including adsorption, ion exchange, and filtration by ZAB, zeolite, and CK.

For UV294, moderate reductions in aromatic absorbance were observed: 0.12 ± 0.03 AU (25%) in the planted column and 0.10 ± 0.05 AU (37%) in the non-planted column. I As UV294 is commonly associated with aromatic and chromophoric organic compounds, these reductions indicate partial removal of potentially recalcitrant organic matter within the treatment columns. In a previous study, Pradhan et al. (2019b) reported approximately 95% UVA removal after 50 days of operation in a green wall system planted with ornamental species primarily through adsorption onto media and microbial degradation facilitated by plant–microbe interactions. Although the UV294 reductions observed in the present study were lower than those reported by Pradhan et al. (2019b), the results nevertheless demonstrate attenuation of aromatic organic constituents. The comparatively lower removal may be attributed to the lower initial UV absorbance of the influent greywater and the short hydraulic retention time in the vertical columns, which likely limited contact time for adsorption and microbial transformation.

Water quality physicochemical parameters

pH

The influent greywater exhibited a neutral pH of 7.29 ± 0.27 (Figure 6a) during Phase 1 (Days 1–31), within the optimal range (6.5–8) for plant growth and microbial activity (Yin et al., 2016). Both planted and non-planted columns showed slight acidification, averaging 7.03 ± 0.11 (planted) and 7.21 ± 0.23 (non-planted) (p < 0.001), with a lower pH in the planted column attributed to nitrification-linked proton release and limited buffering from denitrification. This was also evidenced by high ammonia removal but concurrent nitrate accumulation in both planted and non-planted columns (see Section 3.2.1). Under this condition, the green wall units produced insufficient bicarbonate to buffer the protons released during nitrification, leading to a noticeable drop in pH (Szwerinski et al., 1986). A lower pH in the planted column is also likely influenced by Carex secta –mediated radial oxygen loss (ROL), CO 2 uptake, and rhizosphere microbial activity. These processes collectively contribute to acidification by promoting the oxidation of reduced minerals (e.g., ferrous iron and sulphide to ferric iron and sulphate, respectively), releasing organic acids through root exudates, and enhancing aerobic microbial metabolism via oxygen transport into the root zone (Bai et al., 2023; Liu et al., 2022; Sun et al., 2024). Although not explicitly discussed in earlier studies, Fowdar et al. (2017) and Prodanovic et al. (2020) observed similar pH changes, where non-vegetated systems consistently demonstrated slightly higher pH values ranging from 6.5 to 7.4 in both systems.

In contrast, in the PTCs, the pH significantly increased due to the alkaline nature of CK which releases hydroxide ions (OH −) from the reaction of CaO with water (Gholami et al., 2025a). PTC1 (low HLR), PTC2 (medium HLR), and PTC3 (high HLR) reached average pH values of 8.72 ± 0.62, 9.31 ± 0.59, and 8.69 ± 0.74, respectively (Figure 6a). All three columns showed a clear declining trend over time, with PTC3 peaking at 10.63 on Day 1 and dropping to 7.16 by Day 31. The faster drop in PTC3 was linked to its higher hydraulic loading, which accelerated leaching and washout of alkaline compounds. The pH declines over time across all the PTCs also corresponded dswith decreased phosphate removal, confirming the pH sensitivity of phosphate removal.

EC

The influent greywater had an average electrical conductivity (EC) of 285 ± 14 μS/cm during Phase 1, typical of domestic greywater and well below salinity thresholds that hinder plant growth (>3,000 μS/cm) (Yuvaraj et al., 2021). Throughout this period, the average effluent EC in the planted and non-planted columns was 275 ± 14 μS/cm and 282 ± 18 μS/cm, respectively (Figure 4b), with no significant difference (p = 0.17), indicating minimal ion retention concurring with other studies on the topic Prodanovic et al. (2020).

In contrast, EC increased in the PTCs due to initial leaching from CK, ZAB, and zeolite despite rinsing those materials with ultrapure water before the experiments commenced. On Day 1, EC peaked at 834 μS/cm (PTC1), 609 μS/cm (PTC2), and 524 μS/cm (PTC3). Over 31 days, average EC declined to 380 ± 114, 341 ± 77, and 309 ± 57 μS/cm in PTC1–3, respectively, with more rapid EC declines in higher-flow columns (e.g., PTC3 reaching 284 μS/cm by Day 31). This trend reflects initial mineral release followed by system equilibrium attained—consistent with biochar-based filtration systems (Gholami et al., 2025a; Sami et al., 2023). Although Day 1 values exceeded the 700 μS/cm guideline for non-potable reuse (EPA, 2012), EC dropped below this limit by Day 2 in all PTCs, indicating no long-term salinity risk.

DO

DO is essential for aerobic degradation and nitrification, as it serves as the primary electron acceptor for aerobic heterotrophs and nitrifying bacteria, with effective wastewater treatment observed at DO levels above 0.5 mg/L (Ung et al., 2022). From Days 1–31, the planted column maintained significantly higher average DO levels (6.08 ± 0.36 mg/L) than the non-planted column (5.76 ± 0.37 mg/L; p = 0.0012) and than the greywater influent (Figure 6c). This difference, notable early on (e.g., Day 5: 6.06 mg/L), persisted in Phase 1. The elevated DO in the planted column is attributed to ROL by Carex secta, which enhances rhizosphere aeration and supports microbial activity (Bai et al., 2023; Gholami et al., 2023a). These findings align with previous observations by Attalage et al. (2022), who observed higher DO levels in vegetated filters—up to 7.4 mg/L compared to 5.1 mg/L in bare-soil columns—attributing this to enhanced air diffusion through root structures. These aerobic conditions facilitated improved pollutant removal, including lower phosphate levels and near-complete ammonia removal (Section 3.2.1), consistent with previous findings in vegetated systems (Prodanovic et al., 2019a). However, the consistently high DO (>2 mg/L) likely suppressed denitrification in both columns, which requires low-oxygen or anoxic conditions (Wei et al., 2022), resulting in nitrate accumulation within the green wall units (Section 3.2.1). DO was not measured in the PTCs as they primarily relied on physico-chemical removal processes rather than biological activity, given their short hydraulic retention times.

ORP

Oxidation Reduction Potential (ORP) indicates redox conditions and microbial oxidative capacity. During phase 1, the influent greywater averaged 257 ± 43 mV (Figure 6d) and the planted column maintained slightly higher ORP (274.5 ± 63.4 mV) than the non-planted one (257.8 ± 59.4 mV), though the difference was not statistically significant (p = 0.29). Elevated ORP in the vegetated column is likely due to ROL via aerenchyma, which enhances aerobic activity in the rhizosphere (Tokarz and Urban, 2015). Similar findings were reported by Chowdhury and Abaya (2018), where intensive green roof systems with denser vegetation had higher ORP (∼100.5 mV) than extensive systems (∼43.5 mV) for greywater treatment, due to enhanced oxygenation from root activity. Wei et al. (2023) also observed that columns with higher ROL maintained elevated redox states and achieved superior removal of COD, TN, and TP.

ORP dropped sharply within the PTCs. On Day 1, effluent values fell to 244 mV (PTC1), 137 mV (PTC2), and 85 mV (PTC3), compared to 262 mV in the planted column effluent (Figure 6d). While a decrease in ORP may indicate more reducing conditions within the treatment media, the mechanisms responsible were not directly quantified in this study. The observed decline likely reflects the combined influence of media–water interactions, ion exchange processes, and increases in pH associated with the dissolution of alkaline mineral components. ORP fluctuations observed over time indicate that media–water interactions evolved throughout the experimental period. Lower ORP also coincided with the higher HLR treatment (PTC3). Because ORP is inversely related to pH, the greater pH increase observed in PTC3 may have contributed to its lower ORP values. In addition, enhanced phosphate removal in PTC3, potentially through precipitation with calcium released from the media, may have altered solution chemistry and further influenced ORP. The lower ORP observed under the higher HLR condition may therefore reflect greater media saturation and associated geochemical changes rather than increased biological oxygen demand alone.

Turbidity

Turbidity was significantly reduced in both the planted and non-planted green wall columns. Influent levels averaged 10.31 ± 6.36 NTU, dropping to 0.75 ± 0.61 NTU (planted) and 0.75 ± 0.74 NTU (non-planted), both corresponding to a 93% reduction (p = 0.89) (Figure 6e). The consistently low effluent turbidity values indicate effective removal of suspended particles and suggest that the treated greywater would be suitable for a range of non-potable reuse applications where low turbidity is desirable for aesthetic quality and downstream treatment performance. Solid particles were primarily removed by physical filtration and substrate straining, with added stabilization from biofilms and possible rhizosphere effects in the planted column (Gholami et al., 2023a). Similar efficiencies (>90%) have been reported in other vegetated systems (Lakho et al., 2022; Prodanovic et al., 2020). The PTCs did not leach solids and turbidity levels remained below detectable limits regardless of flow.

Effect of drying period and high loading rate

The impact of a 2-week drying period followed by increased pollutant loading (Phase 2: Days 32–44) was assessed relative to the initial operation period (Phase 1: Days 1–31) (Figures 4–6). The PTCs demonstrated strong resilience, maintaining high treatment performance under stress. Ammonia and nitrite remained undetectable, and phosphate removal remained complete in PTC1 and 2, with PTC3 still achieving 80.5% removal (p < 0.001). TOC and COD showed no significant deterioration in removal performance (p > 0.6), indicating overall stability despite hydraulic and loading stress. Nitrate removal was significantly different between phases (p < 0.015) but remained effective overall at 56%–63% removal. UV294 absorbance dropped by 100% in PTC1–2 and 83.1% in PTC3 (p < 0.001), and anionic surfactants were completely removed in all PTCs. Physicochemical parameters also showed minor but statistically significant phase-dependent variation without affecting overall treatment function. pH trends were similar to Phase 1, EC showed a minor increase, and ORP showed no substantial functional change despite minor fluctuations, indicating that the adsorbent matrix buffered against hydraulic and pollutant shocks.

In comparison, the planted green wall column showed a moderate performance decline. Ammonia increased from 0.02 to 1.63 mg/L (p = 0.0001), phosphate from 3.42 to 4.44 mg/L (p = 0.030), and nitrite 0.26 ± 0.76 to 1.06 ± 0.36 mg/L (p < 0.001). Nitrate increased by 30% but was not statistically significant (p = 0.194) (Figure 4d). EC increased significantly by 41.6% (p = 0.0002) (Figure 6b), while ORP changes were not statistically significant (p = 0.016), indicating slightly more oxidizing conditions under elevated pollutant loading rates (Figure 5d). The pH dropped significantly from 7.03 ± 0.11 to 6.46 ± 0.19 (p < 0.001). This is attributed to the greater concentration reduction in ammonium under the doubled pollutant concentrations, and associated acidification, as well as a reduced denitrification efficiency and associated alkalinity recovery (Figure 5a). COD, TOC, UV 294, and surfactants remained stable (all p > 0.1), suggesting retained performance for organic matter and XOC removal (Figure 5). These shifts likely reflect partial loss of microbial activity and re-saturation of media sorption sites, a common behaviour in drying–rewetting cycles (Prodanovic et al., 2019a; Wiel-Shafran et al., 2006).

In contrast, the non-planted column deteriorated markedly. Ammonia surged nearly tenfold (to 8.88 ± 3.32 mg/L) (p < 0.001), phosphate doubled (to 8.32 ± 1.57 mg/L) (p < 0.001), and nitrite increased significantly (to 1.26 mg/L) (p = 0.0022). Nitrate showed no significant change (p = 0.713). These effects (nutrient leaching) are likely due to microbial die-off, preferential flow, and mineralization of previously stored nutrients in unvegetated media (Fowdar et al., 2017; Prodanovic et al., 2019a). UV 294 (p = 0.439), surfactants (p = 0.905), COD (p = 0.411), and TOC (p = 0.052) remained largely unaffected. The results confirmed the planted column exhibited faster recovery compared to non-planted column, likely due to microbial and oxygen contributions from C. secta, which helped restore redox balance and microbial activity, as well as nutrient uptake during the drying period, reducing initial mineralization losses.

Plant growth and its effect on removal efficiency

The planted column consistently outperformed non-planted across most parameters (e.g., ammonia, nitrate, nitrite, phosphate), emphasizing the vital role of C. secta in enhancing system performance. During stable operation (Days 1–31), the planted column achieved significantly greater pollutant removal than the non-planted column, with >98% ammonia removal (p < 0.01) and ∼21% higher phosphate removal (p = 0.030), due to plant uptake and root-mediated microbial support. While COD and TOC removal were statistically similar, the planted system showed greater stability and lower variance, reflecting rhizosphere-driven process enhancement. Higher DO and ORP levels in the vegetated column supported aerobic microbial activity.

Under drying and increased loading (Days 31–44), the planted column remained stable, whereas the non-planted column deteriorated—ammonia rose nearly tenfold and phosphate doubled. This highlights C. secta ’s role in redox buffering and shock resistance. Although surfactant removal was similar, slightly better reductions in the planted system suggest rhizodegradation. The species also showed strong regrowth, with shoots reaching 27 cm in 1 month and 55 cm by month two (Supplementary Figure S2), following complete removal of above-ground biomass at the substrate surface (i.e., shoots were cut flush with the surface of the media bed) at the start of the experimental period (Day 31), confirming its resilience and suitability for long-term use in nature-based treatment systems.

Microbial identification

High-throughput DNA sequencing revealed distinct shifts in microbial community abundance and composition between the planted and non-planted columns, confirming the strong influence of vegetation on microbial structure and function (Figures 7a–c). At the phylum level, Pseudomonadota (formerly Proteobacteria) dominated all samples (50%–82%), followed by Bacteroidota (6%–18%), Verrucomicrobiota (3%–15%), and Actinomycetota (4%–11%) in P and NP systems. This dominance pattern is consistent with previous reports on microbial communities in greywater- and municipal wastewater-treating green walls and constructed wetlands (Chauhan and Prajapati, 2025; Peng et al., 2025; Qadir et al., 2023).

According to Figure 7a, the proportion of Pseudomonadota was markedly higher in the unplanted columns, reaching 82% in NPT1, while the planted columns ranged between 50% and 63%, reflecting the emergence of vegetation-associated taxa. Members of Pseudomonadota represent a highly diverse bacterial phylum containing taxa involved in a wide range of metabolic functions, including heterotrophic degradation of organic matter and, in some lineages, participation in nitrogen transformations. However, functional processes such as nitrification are typically carried out by relatively low-abundance specialist organisms, and are not reliably inferred from phylum-level dominance alone. In the present study, while members of Pseudomonadota were detected, no clear dominance of classical nitrifying genera (e.g., Nitrosomonas or Nitrospira) was observed, suggesting that nitrogen transformations were likely driven by low-abundance functional guilds or distributed across multiple taxa. This is consistent with the efficient ammonia, organic matter, and xenobiotic removal observed in both planted and unplanted columns (Section 3.2), which likely reflects a combined contribution of diverse microbial communities rather than dominance of specific phylum-level groups. Bacteroidota were moderately abundant in both systems (6%–18%) but slightly higher in the planted columns, particularly in the middle layers. Members of this phylum are chemoheterotrophic bacteria that hydrolyse and ferment complex organic matter—including carbohydrates and proteins—thereby contributing to COD removal and supplying labile carbon for heterotrophic denitrifiers (Chauhan and Prajapati, 2025; Huang et al., 2025; Wang et al., 2024). Verrucomicrobiota were enriched in the planted columns (3%–15%), particularly in the upper layers (PT1–PT2), but were also present in the NP bottom layers (NPB1–NPB2). Verrucomicrobiota can contribute to COD and NH 4 + –N removal in vegetated constructed wetlands (Chauhan and Prajapati, 2025). Actinomycetota (4%–11%) and Acidobacteriota (1.5%–4.8%) were more abundant in the lower layers (PB1–PB2 and NPB1–NPB2), reflecting adaptation to microaerobic environments (Huang et al., 2025). Actinomycetota proliferate under elevated nitrogen concentrations and participate in nitrogen cycling (Chauhan and Prajapati, 2025), while also degrading complex organics and accumulating phosphorus (Wang et al., 2024). Acidobacteriota occur widely in both constructed and natural wetlands due to their roles in carbon and nitrogen cycling (Chen and Shiau, 2025) and can remove nitrogen and phosphorus while degrading organic carbon (Ali et al., 2024).

Several additional phyla—Bdellovibrionota, Cyanobacteriota (0.8%–6.9%), Myxococcota (1%–2%), Bacillota (1%–3%), Patescibacteria (<1%), Chlamydiota, and Chloroflexota (0.9%–2.7%)—were detected predominantly in the planted columns, reflecting the complex microhabitats, redox gradients, and root exudates that promote greater microbial diversity under vegetated conditions. Cyanobacteriota were concentrated in the upper planted layers (PT1–PT2). Their photosynthetic oxygen release can support aerobic metabolism and nitrification (Wang et al., 2024). Bacillota were more abundant in the deeper, less-oxygenated layers, consistent with their predominance in anaerobic wetland zones where they mediate fermentative metabolism and organic-matter degradation (Chauhan and Prajapati, 2025; Wang et al., 2024). Other detected phyla, including Myxococcota, Chloroflexota, and Patescibacteria, were present at relatively low abundances (generally <2%) and are therefore not interpreted in detail at the phylum level. Their presence, however, suggests the existence of localized micro-niches that may support diverse metabolic interactions, including syntrophy and slow turnover of organic substrates.

At the genus-level (Figure 7b), Pseudomonas, Mycobacterium, and Cupriavidus were dominant in the unplanted columns—typical aerobic and oligotrophic taxa associated with organic-matter oxidation under high-oxygen conditions. This aligns with the relatively stable but less hydraulically buffered performance observed in the unplanted systems, particularly under increased loading in Phase 2, where nutrient accumulation (e.g., ammonia and phosphate) was more pronounced (Section 3.2). In contrast, the planted columns were characterized by a broader range of rhizosphere-associated genera such as Bradyrhizobium, Burkholderia, Dyella, and Opitutus, reflecting enhanced nitrogen cycling, phosphate solubilization, and carbon degradation near plant roots. Mucilaginibacter was also abundant in several planted-column samples, particularly within the middle and lower layers. Members of this genus are commonly associated with plant rhizospheres and are known for degrading complex polysaccharides and dissolved organic matter, suggesting a potential contribution to the sustained COD and TOC removal observed in the planted systems.

This increased diversity is consistent with the improved resilience of planted systems during Phase 2 (post-drying and increased loading), where ammonia and phosphate removal remained more stable compared with the unplanted columns. Bradyrhizobium and Burkholderia are well-known nitrogen-fixing and phosphate-solubilizing bacteria that promote nutrient availability and plant–microbe symbiosis in wetland ecosystems (Mng'ong'o et al., 2023; Pal et al., 2022). Roseateles was also present at appreciable abundance in several samples. Species within this genus are frequently reported in freshwater and wastewater environments and are associated with aerobic utilisation of dissolved organic carbon, further supporting the role of microbial organic-matter processing within the treatment columns. The coexistence of nitrifying and heterotrophic taxa supports the coupling between aerobic and microaerobic processes within the planted system. Rather than indicating dominance of classical autotrophic nitrification pathways, the detected community suggests the potential importance of heterotrophic nitrification–aerobic denitrification processes. In particular, genera such as Pseudomonas, Cupriavidus, and Dyella include strains capable of heterotrophic nitrification and coupled aerobic denitrification, enabling simultaneous oxidation of reduced nitrogen species and reduction to gaseous nitrogen under fluctuating oxygen conditions. In contrast, Mycobacterium and Burkholderia are associated with more conventional (anoxic) denitrification pathways, further supporting nitrogen removal under microaerobic or anoxic niches within the column. Additionally, Pseudomonas spp. may contribute to phosphorus removal through polyphosphate accumulation, suggesting a potential role in biological phosphorus cycling alongside physicochemical processes.

The genus-level heatmap (Figure 7c) further highlights the compositional divergence between planted and unplanted columns. Samples from the planted system (PT–PM–PB) clustered together and exhibited higher taxonomic richness and evenness, whereas unplanted samples (NPT–NPM–NPB) formed a distinct, more homogeneous cluster dominated by Pseudomonas, Cupriavidus, and Ensifer. Within the planted columns, the upper and middle layers (PT1–PM2) were enriched in Bradyrhizobium, Burkholderia–Caballeronia–Paraburkholderia, Dyella, Opitutus, and Mucilaginibacter, typical rhizosphere-associated genera involved in nitrogen fixation, phosphate solubilization, and carbon degradation (Wang et al., 2022). This enrichment coincided with the strong nutrient and organic-matter removal observed throughout the planted columns. In contrast, the deeper layers (PB1–PB2) showed enrichment of Reyranell and other facultative taxa, indicating the establishment of microaerobic and fermentative communities under reduced oxygen conditions (Mustafa et al., 2024). This depth-related segregation supports the coexistence of aerobic and microaerobic guilds promoted by root oxygen release and exudate-derived carbon, consistent with previous observations that vegetation enhances microbial heterogeneity and redox microzonation in constructed wetlands (Kinnunen et al., 2025).

Preliminary practical and economic considerations

A preliminary assessment was undertaken to evaluate the practical viability of adding PTCs to the green wall system. Based on phosphate breakthrough under the PTC1 loading, the adsorbent mixture was projected to provide approximately 70–75 days of treatment, corresponding to an indicative quarterly replacement interval. Increasing media depth or regenerating adsorbent could potentially extend this service life. Scaling the experimental media demand to typical household greywater production suggests a requirement of approximately 2–3 kg of mixed CK, ZAB, and zeolite per person per month, although this estimate will vary with influent quality, flow rate, treatment targets, and regeneration efficiency.

Published green-wall installation costs range from approximately NZD 640–2,300 per m 2, depending on design, scale, materials, plumbing, and labour (Gholami et al., 2023b; Perini and Rosasco, 2013). The PTC would add costs for adsorbent preparation, column housing, pumping, media replacement or regeneration, and spent-media management. Although the media are waste-derived, CK calcination and ZAB activation require energy and chemicals. According to current New Zealand commodity pricing, zeolite typically costs NZD 75–350/t bulk, and commercially around NZD 1.5/kg for consumer sized packing (20 kg bag). Similarly, biochar costs are estimated to be around NZD 2–4/kg, but granular activated carbon–a more reasonable comparison due to the additional activation step, is closer to NZD 13/kg for 25 kg bags.

In this testing we used singular mixed-media PTC columns comprising equal volumetric proportions of each material. The benefit of these systems is that exhausted media can be replaced without disturbing the vegetation or green-wall substrate. However, the use of three separate materials adds complexity that warrants further study and optimization. It is expected that the proportions could be optimized to meet the typical profile of either domestic or commercial greywater such that the expected lifetime of each material is similar. This would allow simple operation by replacing a singular media unit periodically. Alternatively, the materials could be separated, with each being replaced at defined intervals, or as needed if low-cost sensors can be incorporated to indicate saturation.

This study demonstrates a novel two-stage greywater treatment system comprising a Carex secta-planted green wall followed by a replaceable post-treatment column containing complementary waste-derived adsorbents. Unlike conventional green walls, the system combines biological treatment with targeted physicochemical polishing to enhance the removal of residual phosphate, nitrate, surfactants, and organic compounds. The novelty lies in the whole-system integration and its resilience under variable hydraulic loading, drying, and increased pollutant loading, rather than in the individual adsorbents.

The green wall units alone achieved near-complete ammonia removal (99.8%) and reduced phosphate and nitrate by 17% and 70%, respectively, compared to the non-planted system. While COD and TOC removal was comparable between planted and non-planted columns, vegetation provided greater redox buffering, particularly under drying and high-loading stress. The non-planted column experienced phosphate doubling and tenfold ammonia spikes under the increased concentration/loading applied after a drying stage, whereas the planted system remained resilient.

The PTCs played a critical role in polishing effluent from the green wall. Under low-flow conditions, they achieved complete removal of phosphate, ammonia, nitrite, surfactants, and UV-absorbing compounds, while also reducing COD and TOC by 14% and 17% beyond the upstream green wall stage—bringing overall removal to 94%–95%. Phosphate was fully removed through calcium-driven precipitation by CK, while nitrate was reduced by 56%–63%, likely through adsorption by ZAB as demonstrated in earlier batch-testing.

Under typical conditions, phosphate removal reached 100% under low-flow (PTC1) and exceeded 80% in higher-flow columns (PTC 2 and PTC3). Assuming similar influent loads and Ca(OH) 2 availability, the projected phosphate exhaustion time in PTC1 was ∼70–75 days. This duration, which could be extended by increasing column length or layering, offers a feasible replacement cycle (∼quarterly) in decentralized applications. Based on measured removal rates and typical household greywater volumes (globally 100–150 L/person/day), material requirements could be scaled to ∼2–3 kg of CK/ZAB/zeolite mixture per person per month, indicating manageable material quantities. Post-treatment also achieved full removal of ammonia, nitrite, and anionic surfactants, with UV 294 reduced by up to 100% and COD and TOC by 95%–96%. These outcomes confirm the robustness of post-treatment columns, under variable hydraulic and pollutant loads, as compact design modules for treating greywater pollutants in decentralized reuse systems. However, whole-system feasibility will depend on sorbent production/functionalisation and regeneration energy, and because pathogens were not assessed, reuse-oriented deployment would require a disinfection barrier (e.g., UV) that adds energy and operational demands.

To support system optimization and real-world adoption, future research should focus on quantifying long-term material exhaustion rates under seasonal and fluctuating conditions, testing regeneration and beneficial reuse of spent media (e.g., as a second-generation soil amendment/fertilizer where safe), and understanding biofilm–media interactions. Pilot-scale systems should be validated with real greywater under outdoor conditions to capture variability in influent composition and climate. This should include targeted sensitivity and safety testing for ZAB, particularly potential Zn leaching under the range of pH and ionic strengths expected in greywater, as well as robustness to different surfactant classes, NOM fractions, and long-term continuous operation. Together, these steps will strengthen the case for PTC-integrated green walls as sustainable, low-maintenance solutions for decentralized greywater treatment.