Sustainability of global small-scale constructed wetlands for multiple pollutant control

Rapid growth in wastewater generation (360–380 billion m³ annually, projected to rise 24% by 2030 and 51% by 2050) and the high fraction discharged untreated (48%) threaten water security and impede progress toward UN SDG 6.3. Constructed wetlands (CWs) are globally recognized, nature-based, cost-effective solutions for wastewater treatment, with substantially lower capital, energy, and maintenance costs than conventional plants. However, uncertainty persists about their feasibility and long-term effectiveness to concurrently remove multiple pollutants due to potentially synergistic, antagonistic, or neutral interactions among removal processes and temporal variability in performance. Land requirements also constrain large CW deployment, prompting increasing focus on small CWs (<8 ha) and their landscape-scale roles. This study addresses three questions using global small CW data: (i) Can small CWs effectively remove multiple pollutants, and what are the relationships among pollutant removals (synergy, antagonism, neutrality)? (ii) Are these relationships sustainable and influenced by temporal thresholds? (iii) What key design, hydraulic, and environmental factors drive multiple-pollutant removal, and do they have thresholds?

Prior work has documented CWs’ effectiveness and economics relative to conventional wastewater treatment, and explored mechanisms (physical, chemical, biological) and factors such as area, temperature, and hydraulics. Yet findings on multi-pollutant interactions are mixed, with reports of synergistic, antagonistic, or neutral relationships and asynchrony in nutrient removal. Long-term sustainability of performance remains underexplored, with most studies focusing on single pollutants and early operational years. The substantial land requirement of CWs is a noted barrier, motivating interest in small and micro wetlands, recognized by international policy (e.g., Ramsar resolution) for their potential importance in nutrient processing and as biogeochemical hotspots. A global, quantitative synthesis of small CWs’ multi-pollutant removal interactions, their persistence, and governing thresholds has been lacking.

Data source and selection: Publications (2000–2022) were retrieved from Web of Science Core Collections using terms related to constructed/treatment/artificial wetlands and reed/floating beds; non-English studies were excluded. Two global datasets on CW water quality and small water bodies were also used. Inclusion criteria: outdoor CWs treating real wastewater, engineered applications (not lab experiments), reported basic site parameters (location, area), CW area < 8 ha, reported removal for at least two of COD, TN, TP, NH4+-N, and operational or investigation duration. This yielded 196 publications and 364 CW sites across 5 continents. Extracted variables: location (lat/long, country, continent), CW type (FWSCW, SSFCW, FCW, HCW), area (m²), removal efficiency (%) and loading rate (g m⁻² d⁻¹) for TN, TP, COD, NH4+-N (computed from inflow/outflow when needed), hydraulic loading rate (HLR, m/d), hydraulic retention time (HRT, days), air temperature (°C), operational/investigation duration (months), elevation (m). Statistical analyses: Publication bias was assessed via funnel plots and Egger’s test. Pairwise pollutant removal relationships were modeled using linear mixed-effects models (LMEMs) with study site as a random effect; each of the 12 pollutant pairs was modeled bidirectionally. Model fit was evaluated by OLS regression of observed vs. predicted values (R² ~0.76–0.88). Standardized fixed-effect coefficients assessed direction/strength; slope_PRE quantified the slope between two removals (positive for synergy, negative for antagonism). Sustainability over time: Duration (months) was included as a moderator in LMEMs; relationships were predicted at minimum/mean/maximum durations using the Johnson–Neyman technique to identify temporal thresholds where significance changed. Trends of slope_PRE vs. duration were tested with OLS (slope_OLS). Influencing factors: Area, HLR, HRT, and air temperature were tested as moderators in LMEMs; slope_PRE vs. each factor was evaluated by OLS, and Johnson–Neyman analysis identified factor thresholds for significance and potential shifts to antagonism. Structural equation modeling (SEM) quantified direct effects of area, HLR, HRT, and temperature on interactions (covariance) among pollutant removals; model fit used RMSEA and CFI, with pathways iteratively pruned based on significance and ecological plausibility. Random Forest analysis assessed relative importance of factors. Group comparisons used non-parametric Wilcoxon tests due to assumption violations (Shapiro–Wilk, Levene). Analyses were conducted in R with nlme/lmerTest/lavaan/randomForest and related packages; visualization used ggplot2.

Effectiveness: Average removal efficiencies for small CWs were COD 65.2% (n=214; 95% CI 62.1–68.4), NH4+-N 57.1% (n=201; 53.5–60.8), TN 50.2% (n=269; 47.2–53.2), TP 50.5% (n=276; 47.4–53.7). The 75th percentile removal efficiencies were COD 84.0%, TN 68.8%, TP 72.7%, NH4+-N 78.6%; outflow loading rates were significantly lower than inflows. Synergies: LMEMs showed significant positive pairwise interactions across pollutants (standardized coefficients 0.27–0.64; p<0.001). slope_PRE indicated TN removal increased with TP, COD, NH4+-N removal (0.24–0.72; p<0.001); TP removal increased with COD, NH4+-N removal (0.28–0.71; p<0.001); COD removal correlated with NH4+-N removal (0.24–0.39; p<0.001). SEM corroborated positive interactions (b=0.31–0.59; p<0.001). Across CW types, 43/48 pairwise interactions were significantly positive; across continents, 48/56 were positive, indicating global generality. Sustainability over time: Johnson–Neyman analysis identified temporal thresholds for 11 of 12 relationships, with synergistic associations persisting approximately 34.3–142 months (~3–12 years) depending on pollutant pair and CW type. Example thresholds: TN–TP synergy ~113–122 months; COD–NH4+-N ~34–43 months. Beyond these durations, synergies weakened, became non-significant, or turned antagonistic. Thresholds of drivers: Significant threshold effects were detected for area, HLR, HRT, and temperature on synergies. Across the 12 relationships, minimum (optimal) thresholds to sustain all synergies were: area 17,587 m² (~1.7 ha), HLR 0.45 m/d, HRT 8.2 days, temperature 20.2 °C. As factors exceeded these minima, some synergies became non-significant; when exceeding upper thresholds (area up to 43,349 m²; HLR up to 3.56 m/d; HRT up to 96.6 days; temperature up to 30.2 °C), all 12 synergies collapsed or shifted to antagonism. OLS trends indicated strengthening or weakening of synergies depending on factor and pair. Mechanisms and relative importance: SEM showed direct negative effects of area (b ≈ −0.18 to −0.30), HLR, and temperature on interactions, and a positive effect of HRT (b ≈ 0.17–0.35). Area had the largest mean effect (mean b = −0.23), followed by HRT (mean b = 0.22), consistent with Random Forest results. Small areas favored maintaining synergies. HCWs exhibited the strongest synergistic interactions, followed by FWSCWs, SSFCWs, and FCWs. Appropriate stoichiometric ratios (e.g., inflow COD/TN > 7; COD/TP > 40) can enhance multi-pollutant synergy.

This global synthesis demonstrates that small CWs, despite limited area, can effectively remove organic matter and nutrients and exhibit robust, broadly generalizable synergistic interactions among pollutant removals across climates and CW types. The persistence of these synergies is time-bound, typically lasting 3–12 years, and varies by pollutant pair and CW configuration. Recognizing temporal and design/hydraulic/environmental thresholds is critical: synergies are best maintained below specific thresholds of area, HLR, HRT, and temperature, with area emerging as the most influential driver and HRT as a positive lever. These insights address the core research questions by quantifying multi-pollutant synergies, establishing their sustainability windows, and identifying key mechanisms and thresholds. Practically, they guide CW design and operation toward smaller footprints with adequate retention (HRT ≥ 8 days) and moderate loading (HLR ≤ 0.45 m/d), and inform adaptive management (e.g., aeration, electron donor addition, dredging) when systems approach threshold conditions. Given land constraints and the need for decentralized, nature-based solutions, promoting small-scale CWs is a viable pathway to advance SDG-aligned wastewater management and alleviate water scarcity pressures.

Small-scale constructed wetlands deliver high multi-pollutant removal performance with globally consistent synergistic interactions among organic matter and nutrient removal processes. These synergies are sustainable over multi-year timescales (about 3–12 years) and are governed by clear thresholds in area, hydraulic loading, retention time, and temperature. Area is the principal determinant of synergy (smaller is better), while sufficient HRT enhances interactions. The study provides quantitative thresholds to guide the design, operation, and long-term management of small CWs and supports their deployment as scalable, nature-based solutions for wastewater treatment worldwide. Future work should strengthen long-term monitoring beyond a decade, expand coverage in data-sparse regions (e.g., Africa, Oceania), and extend analyses to other pollutants (microplastics, antibiotics, heavy metals) and context-specific factors (policy, economics, wastewater characteristics, substrates, and plant assemblages).

Geographical data gaps exist, notably limited observations from Africa and Oceania, partly due to missing site attributes (e.g., area, multi-pollutant data). Most available studies cover only early operational years (median ~3 years), constraining long-term inference. The analysis focused on area, HLR, HRT, and air temperature; other influential factors (influent pollutant ratios, substrate composition, plant species, climate extremes, policy and economic contexts) were not comprehensively quantified and may affect generalizability. Additional pollutants (e.g., microplastics, antibiotics, heavy metals) and their synergistic dynamics were not assessed.