From waste to energy: a closed-loop two-phase anaerobic digestion system for sustainable dairy wastewater management

Abstract

The treatment of dairy wastewater (DW), characterized by high organic load and lipid/protein content, remains challenging due to the energy-intensive nature of aerobic processes and instability of anaerobic methods. This study developed a self-regulating two-phase anaerobic digestion (TPAD) system integrating an anaerobic baffled reactor (ABR) with an up-flow anaerobic sludge blanket (UASB) reactor. Sequential phase separation in the ABR enables microbial self-organization for staged lipid adsorption, protein denaturation, and hydrolysis-acidification, ensuring stable UASB input. Laboratory-scale operation achieved exceptional chemical oxygen demand (COD) removal (97.06-99.01%). Full-scale implementations across three Chinese provinces demonstrated robust performance, with COD removal of 78.13-93.46%, high methane content (83.20-83.94%), sludge reduction >75.00%, and reductions in energy consumption (64.71-85.03%) and greenhouse gas emissions (88.01-97.09%) compared to conventional systems. Microbial analysis confirmed functional spatial divergence. The TPAD system presents a regionally-proven, versatile, and scalable solution to transform DW management from a disposal cost into a biogas-generating process.

Dairy wastewater treatment requires efficient and sustainable solutions. Yuqi Gong and colleagues reported a self-regulating anaerobic system that achieved high pollutant removal and methane production, while significantly reducing energy use and emissions.

Introduction

The global dairy industry, a vital component of food security, is projected to produce 1.085 billion tons by 2033¹. This growth is shadowed by a huge environmental challenge: wastewater generation rates that can exceed 2.5–10 times the volume of processed milk². Dairy wastewater (DW), characterized by high concentrations of recalcitrant organics, suspended solids (SS), and nutrients, threatens water quality and contributes to greenhouse gas (GHG) emissions². In regions like China, where rapid consumption growth intensifies these pressures, the development of innovative and sustainable treatment solutions is urgently needed to reconcile industrial development with national environmental goals, underscoring the complex climate–water–energy nexus in treatment systems.

Conventional DW treatment predominantly relies on anaerobic digestion (AD) for recovering biogas. However, single-phase AD systems, such as the up-flow anaerobic sludge blanket (UASB) reactor, face major challenges when treating dairy waste³. The high and fluctuating organic loads, along with complex compositions rich in lipids and proteins, often lead to inhibited hydrolysis kinetics⁴, microbial washout⁵, and reactor acidification⁶. Integrated anaerobic-aerobic bioreactors can achieve better pollutant removal than traditional AD methods⁶. Nevertheless, inefficiencies in the anaerobic phase (e.g., sludge bulking and foaming) inevitably necessitate carbon-intensive aeration in the subsequent aerobic process⁷. This, in turn, leads to high sludge production and operational costs⁸. Collectively, these limitations cause process instability, limit resource recovery, and generate substantial GHG emissions^9,10, hindering the transition to sustainable and economically viable treatment models.

Two-phase anaerobic digestion (TPAD) systems, which spatially separate the hydrolysis-acidification and methanogenesis phases, present a promising strategy for enhancing process stability and resource recovery. However, many existing TPAD configurations rely on external controls or additives to maintain performance. Recent studies, for instance, have often depended on reactor modifications or exogenous additives (e.g., CaO₂ or BC/nZVI) to enhance performance, achieving high COD removal in laboratory settings^11–13. Nonetheless, this reliance introduces operational complexities, cost concerns, and potential long-term microbial instability^14–16. This research builds directly upon prior work, including our own demonstration of an anaerobic baffled reactor (ABR)–UASB reactor configuration for DW treatment¹⁷. While that earlier study confirmed the system’s efficacy, it was constrained by a fixed HRT and offered limited mechanistic insight. The present study addresses these specific limitations by introducing a self-regulating TPAD paradigm that achieves operational stability through inherent microbial self-organization, thereby eliminating dependencies on external controls. This study, therefore, aims to develop a robust, additive-free system that overcomes these inherent drawbacks to create favorable conditions for subsequent treatment.

To address these gaps, we propose a microbial self-regulation paradigm for sustainable DW treatment. This approach integrates ABR and UASB reactor technologies into a self-organizing TPAD system that operates without external controls or additives, establishing closed-loop material-energy cycles. The system’s efficacy stems from the distinct yet complementary roles of its two core components^18–20. The compartmentalized design of the ABR enables sequential lipid adsorption, protein denaturation, and hydrolysis–acidification, which stabilizes organic retention and microbial activity. Meanwhile, the downstream UASB reactor achieves enhanced methane production through the self-enrichment of acetoclastic methanogens, a process driven by functional microbial selection rather than artificial intervention.

This study develops a self-regulating TPAD system that advances beyond previous configurations. Its novelty lies in stable operation under flexible HRT without exogenous additives, validated at both ambient (laboratory-scale) and mesophilic (full-scale) temperatures. We employ a multidisciplinary approach—integrating computational fluid dynamics (CFD) modeling, microbial community analysis, metabolic pathway mapping, and life cycle assessment (LCA)—to provide a comprehensive mechanistic understanding of system performance. The findings establish a robust, scalable technical framework that transforms wastewater management from an energy-intensive cost into a resource-recovery process. By aligning technological innovation with circular economy principles, this work contributes directly to Sustainable Development Goals (SDGs) 6, 7, and 12.

Results

Design of the TPAD system

The TPAD system was engineered as a resource-positive and low-carbon solution that bridges wastewater treatment with circular economy principles. This integrated system combines an ABR with a UASB reactor. It operates at ambient temperatures (room temperature at laboratory scale; 35 ± 2 °C in full-scale engineering) without requiring external resources or energy inputs. The core mechanism relies on compartmentalized ABR chambers packed with biofilm carriers (suspended plastic media with high specific surface area), which leverage microbial self-organization to execute sequential biochemical processes: (1) lipid adsorption and protein coagulation/denaturation, followed by (2) hydrolysis and acidification. (Fig. 1a). The chambers were configured with specific carriers to promote distinct functions sequentially. The initial chambers are optimized for lipid adsorption and protein coagulation, while subsequent chambers foster the hydrolysis and acidification of the retained organic matter, thereby facilitating the phased removal and transformation of complex substrates. These interconnected processes establish a closed-loop mechanism, embodying a “Low Extraneity Reservoirs” design philosophy that minimizes reliance on external inputs. This design achieves three pivotal functions: buffering hydraulic load fluctuations, maintaining operational stability without external controls, and minimizing material loss while maximizing methane-rich biogas production (Supplementary Tables S1 and S2). The system’s robust stability stems from the functional selection of self-organized microbial communities, which is driven by the phased reactor design.

Fig. 1. Schematic diagram and flow chart of TPAD (two-phase anaerobic digestion) system.

a Schematic diagram of laboratory scale TPAD system (VFAs volatile fatty acids, COD chemical oxygen demand, SS suspended solids, UASB up-flow anaerobic sludge blanket). b Flow chart of scale-up engineering of TPAD system. Laboratory studies employed a single TPAD system. In full-scale engineering trials (1500 m³ DW/d), the Anhui, Hunan, and Yunnan projects implemented TPAD systems coupled with contact oxidation tanks, while the Kunming (800 m³ DW/d) single-stage anaerobic project utilized an ABR (anaerobic baffled reactor) combined with a contact oxidation tank.

DW exhibits complex composition, characterized by pH variability, inorganic ions (e.g., Fe, Ca, Na, and K), recalcitrant organic compounds, and SS^3,21,22. The TPAD system addresses this complexity through intrinsic process intensification. Unlike the traditional TPAD process, the designed front-phase reactor primarily focuses on retaining organic matter and facilitating its hydrolysis. Specifically, it mitigates lipid-protein interactions via compartment-specific adsorption and coagulation, while optimized HRT and sludge bed dynamics prevent reactor acidification^5,23 and sludge bulking²⁴ (Fig. 1b). Laboratory and engineering scale studies demonstrated that enhanced organic stabilization through retained activated sludge processes. The resulting volatile fatty acids (VFAs; Supplementary Table S3), SS, and ammonia nitrogen (NH₃–N) were then channeled to the UASB reactor under optimized conditions.

Full-scale implementations across three Chinese provinces (Yunnan, Anhui, and Hunan) validated the operational robustness. To ensure effluent compliance with discharge standards, the anaerobic TPAD effluent was subsequently treated in an aerobic biological contact oxidation tank. This polishing step effectively removes residual pollutants, such as remaining organic matter and nutrients (e.g., NH₃–N and TP), which are not fully eliminated by anaerobic digestion alone. Residual sludge from the whole system was dewatered using a filter press prior to landfill disposal. The system sustained long-term stability through concurrent sludge retention, nutrient recovery, and scalable biogas production. Compared to conventional systems^21,25, this configuration reduced aeration energy demands by 32.01–41.10%^26,27 while mitigating critical challenges like sludge bulking²⁴ and foaming^28,29. Economic viability stems from optimized resource utilization, which yields superior loading capacity, simplified process management (requiring less operator intervention)^30,31, and major energy savings (Supplementary Tables S4–S12 and S17).

Operational performance of TPAD at laboratory scale

The TPAD system’s efficacy was rigorously evaluated under controlled laboratory conditions, with a focus on sludge retention, organic removal efficiency, and resilience to organic shock loads. Given the critical influence of inlet configuration and HRT on these parameters in wastewater treatment systems^32–34, this study first employed CFD simulations using Ansys Fluent to analyze the flow dynamics. The simulations revealed minimal sludge washout during inoculation and showed that effective sedimentation led to the formation of stable sludge beds within the ABR compartments (Fig. 2a, b). Subsequent steady-state analyses confirmed that variations in fluid rheology and HRT (24–30 h) had a negligible impact on degradation efficiency, and inflow patterns similarly showed minimal effects on removal rates (Supplementary Fig. S1).

Fig. 2. Operation performance of the TPAD (two-phase anaerobic digestion) system at the laboratory scale.

a Volume fraction of sludge in ABR (anaerobic baffled reactor, the color from blue to red indicates an increase in volume fraction). b Velocity of sludge in ABR (the color from blue to red indicates an increase in velocity). c Influent COD (chemical oxygen demand) concentrations (yellow square), effluent COD of ABR (green sphere), and effluent COD of UASB (up-flow anaerobic sludge blanket) reactor (purple sphere). d COD removal rate of ABR (purple bar chart), COD removal rate of UASB reactor (pink bar chart), and COD removal rate of the entire system (blue bar chart). e Influent NH₃–N (ammonia nitrogen) concentrations (yellow bar chart), effluent NH₃–N of ABR (purple sphere), and effluent NH₃–N of UASB reactor (green sphere). f pH of influent (pink square), effluent of ABR (blue sphere), and UASB reactor (green triangle).

Furthermore, vector field analyses demonstrated the efficient adhesion of organic matter from the dairy wastewater to the biofilm carriers (i.e., the immobilized biomass), which ensures sustained nutrient availability for the microbial community and underscores the advantage of the ABR’s compartmentalized design (Supplementary Figs. S1 and S2). Collectively, these characteristics enabled the TPAD configuration to achieve enhanced sludge retention^24,35,36 and superior resistance to organic shock loads³⁷ compared to literature benchmarks.

Notably, the system facilitated the transfer of protein-derived granular sludge to the UASB reactor, a mechanism that contrasts with the conventional dominance of dissolved-phase organicse²¹. This process ensured microbial stability and provided a sustained carbon supply to the UASB reactor. Supporting these observations, Supplementary Fig. S2a, f, k demonstrated a progressive increase in sludge contact angle (from 67.2° to 93.9°) during the reaction process. This increase is attributable to the sequential surface encapsulation of activated sludge by proteinaceous and lipid compounds, which were subsequently metabolized by hydrolytic acid-producing bacteria.

Scanning electron microscopy (SEM) images (Supplementary Fig. S2b, g, l) further revealed distinct microbial morphologies in the two reactors. The sludge in ABR exhibited a granular morphology, dominated by organic-encapsulated bacterial consortia, which is indicative of enhanced sludge retention capacity and a stable sludge bed. In contrast, the UASB reactor maintained flocculent sludge structures with characteristic surface corrugations, a feature that improved contact between the wastewater and biomass, thereby enhancing treatment performance^37,38.

Key operational parameters, including pH and temperature, were critical determinants of system performance^39,40. During the 120-day laboratory trial at ambient temperature, the initial start-up phase (Days 1–30, HRT = 24 h) achieved a COD removal efficiency of 97.26%, reducing the influent concentration from 1300.00 ± 88.56 mg L⁻¹ to 35.30 ± 21.07 mg L⁻¹ in the UASB reactor effluent (Fig. 2c, d). Concurrently, a surge in NH₃–N concentration was observed, from 16.66 ± 7.80 mg L⁻¹ in the influent to 59.60 ± 13.10 mg L⁻¹ after the UASB reactor (Fig. 2e), which serves as a clear indicator of robust hydrolysis of nitrogenous organic compounds.

Moving to the stable operation phase (Days 31–120), the system maintained high performance even under an increased organic loading rate (Influent COD of 1640.87 ± 135.69 mg L⁻¹) and an extended HRT of 30 h. The UASB reactor effluent COD remained low at 47.91 mg L⁻¹, corresponding to a removal rate of 97.06% (Fig. 2c, d). NH₃–N levels continued to rise through the process, from 17.60 ± 2.24 mg L⁻¹ at the ABR inlet to 66.79 ± 4.49 mg L⁻¹ at the ABR outlet, and stabilized at 65.70 ± 3.13 mg L⁻¹ following UASB reactor treatment (Fig. 2e). Importantly, these NH₃–N concentrations remained well below the inhibitory threshold for anaerobic microorganisms (typically >1500 mg L⁻¹ ammonia nitrogen)⁴¹, indicating stable operation without toxicity concerns. The pH values provided further evidence of system stability, declining slightly from 7.52 ± 0.15 in the influent to 6.95 ± 0.14 in the ABR, before recovering to a near-neutral 7.25 ± 0.05 in the UASB reactor effluent (Fig. 2f). This near-neutral pH in the ABR effluent, combined with the high COD removal, points to a well-balanced process between acidogenesis and methanogenesis.

The occurrence of extensive hydrolysis was further corroborated by distinct changes in the sludge’s physicochemical properties (Supplementary Fig. S2). Specifically, a substantial decrease in sludge contact angle indicated increased hydrophilicity, while SEM images revealed the disintegration of microbial aggregates, and EDS mapping showed a diffuse redistribution of nitrogen. These transformations are consistent with the breakdown of complex organic matter, confirming that effective hydrolysis created favorable conditions for the downstream UASB reactor.

The compartmentalized sludge bed architecture of the ABR addressed the challenge of sludge retention by enhancing microbial adhesion mechanisms, maintaining operational stability under effective HRT and varying organic loading rates. Meanwhile, the protein-derived granular sludge transfer mechanism ensures process stability in the UASB reactor during load fluctuations. The TPAD system as a whole maximizes the utilization of organic matter while creating favorable conditions for subsequent aerobic treatment by reducing excess sludge production and energy consumption. These innovations overcome the limitations of traditional systems characterized by high energy consumption and low recovery (Supplementary Table S17), demonstrating higher pollutant removal efficiency and a lower carbon footprint compared to energy-intensive thermophilic alternatives. By integrating technical efficiency with circular economy principles, the TPAD system aligns with China’s dual-carbon goals and offers a scalable and sustainable solution for wastewater management.

Scale-up engineering of TPAD

Three full-scale TPAD systems were deployed in Yunnan, Anhui, and Hunan provinces, China, validating industrial scalability, adaptability, and sustainability. The facilities in Yunnan and Anhui employed sequential batch operations with high-flow influent, maintaining stable operation for 18 months at HRT of 48 h (ABR) and 96 h (UASB reactor) in Yunnan, and 36 h (ABR) and 48 h (UASB reactor) in Anhui. In contrast, the Hunan facility utilized a continuous low-flow feeding (HRT 36-48 h) model that mirrored the laboratory-scale configurations, and stable operation for 12 months. These different operational modes were designed to meet distinct regional discharge standards: the stricter reclaimed water reuse standards (50–120 mg L⁻¹, GB/T 18918-2002) in Yunnan and Anhui, versus the municipal sewer discharge limits (300-500 mg L⁻¹, GB/T 31962-2015) in Hunan.

Pollutant removal efficiencies confirmed the robustness of the TPAD system across all sites. As shown in Fig. 3a, COD reduction reached 91.34% in Anhui (from 1967.78 ± 73.30 mg L⁻¹ to 170.28 ± 12.75 mg L⁻¹), 93.46% in Yunnan (from 1529.24 ± 93.43 mg L⁻¹ to 99.29 ± 10.45 mg L⁻¹), and 78.13% in Hunan (from 2368.52 ± 189.48 mg L⁻¹ to 516.18 ± 62.48 mg L⁻¹). The dynamics of NH₃–N are presented in Fig. 3b. Its concentration increased by 54.52%, 208.83%, and 29.26% after ABR treatment in Anhui, Yunnan, and Hunan, respectively, and subsequently rose by a further 28.18%, 38.65%, and 79.52% during the UASB reactor phase. Meanwhile, SS levels increased substantially within the ABRs—by 129.71% in Anhui, 75.90% in Yunnan, and 267.53% in Hunan—indicating effective solids retention. In contrast, the UASB reactor effluents exhibited SS reductions, ranging from 24.39% to 75.92% compared to the influent concentrations (Fig. 3c).

Fig. 3. Operation performance at Yunnan, Anhui, and Hunan provinces of the TPAD (two-phase anaerobic digestion) system.

a Influent COD (chemical oxygen demand) concentrations (green bar chart), effluent COD of ABR (anaerobic baffled reactor, pink bar chart), effluent COD of UASB (up-flow anaerobic sludge blanket) reactor (blue bar chart), removal rate of ABR (orange square), removal rate of UASB reactor (blue sphere), and removal rate of the entire system (pink triangle). b Influent NH₃–N (ammonia nitrogen) concentrations (green bar chart), effluent NH₃–N of ABR (pale blue bar chart), and effluent NH₃–N of UASB reactor (dark-blue bar chart). c Influent SS (suspended solids) concentrations (pale blue bar chart), effluent SS of ABR (dark-blue bar chart), and effluent SS of UASB reactor (green bar chart). d Correlation heatmap of reactor nodes of ABR in Yunnan (the color from green to purple indicates an increase in positive correlations).

A comparative analysis of the three full-scale TPAD systems quantitatively demonstrates the process’s robustness and adaptability, while also revealing how operational modes dictate performance trends (Fig. 3, Supplementary Table S1). The COD reduction efficiency was systematically higher in the batch-operated Anhui (91.34%) and Yunnan (93.46%) plants, designed for reclaimed water reuse, than in the continuous-flow Hunan system (78.13%), which targeted less stringent sewer discharge standards. This performance difference is primarily attributed to the fundamental difference in feeding strategy. The sequential batch mode with high-flow influent likely promotes a more plug-flow-like hydraulic pattern, reducing short-circuiting and allowing for longer effective contact time between the wastewater and the biomass. This mechanism promotes longer effective retention and enhanced hydrolysis of complex solids, whereas the continuous low-flow feeding in Hunan results in a more completely mixed regime, potentially reducing the effective contact time.

This interpretive framework also clarifies the contrasting nitrogen transformation patterns. The dramatic surge in NH₃-N following the ABR in Yunnan (208.83%) and Anhui (54.52%) is characteristic of the rapid hydrolysis and ammonification triggered by the batch feeding shocks. In contrast, the more subdued increase in Hunan (29.26%) reflects the dampening effect of continuous feeding on organic load fluctuations, resulting in a more gradual and stable release of ammonia. Notably, the consistent upward trend in NH₃–N across all sites confirms the ubiquitous occurrence of hydrolysis and ammonification, despite differences in dynamics.

Similarly, the substantial increase in SS within the ABRs (e.g., 267.53% in Hunan) and its reduction after the UASB reactors (up to 75.92% removal compared to the influent) across all projects underscores effective solid retention and subsequent biodegradation. The variations in the magnitude of these changes further highlight the influence of local wastewater characteristics and specific operational parameters on the internal process dynamics, while simultaneously confirming the fundamental stability and effectiveness of the TPAD mechanism under varying conditions.

Remarkably, these efficiencies were maintained even though HRTs were shorter than literature benchmarks¹⁰, confirming the industrial viability of the TPAD system. Analyses of the sludge and process parameters provided further insights into the system’s operation. SEM and elemental mapping revealed compact sludge aggregates enriched with C, N, O, P, and S (Supplementary Fig. S2). Furthermore, strong positive correlations (Pearson’s r > 0.80, p < 0.05) were observed between COD and SS across reactor nodes (Fig. 3d; Supplementary Fig. S3). The elevated NH₃–N levels at ABR outlets (1.54×, 3.09×, and 1.32× influent concentrations in Anhui, Yunnan, and Hunan, respectively) prevented acidification (Supplementary Table S3) while optimizing pH (7.89–8.74) for methanogenesis (Supplementary Fig. S4). Notably, the ABRs in the Anhui and Hunan projects exhibited limited production of VFAs, with acetate concentrations of 0.60 mg L⁻¹ and 0.40 mg L⁻¹, and propionate concentrations of 0.20 mg L⁻¹ and 0.10 mg L⁻¹, respectively (Supplementary Table S3). This contrasts with conventional two-phase systems where maximizing acidogenesis is a primary goal (Supplementary Table S17). In this study, the ABR was strategically optimized for operational stability against issues like sludge flotation. For the specific composition of dairy wastewater, achieving a metabolic balance where hydrolysis was dominant over rapid acidification was critical. This controlled environment prevented the inhibition of downstream processes and ensured stable feeding conditions for the subsequent UASB reactor phase. Concurrently, the progressive increase in SS concentration from influent to effluent of ABR (204.44 mg L⁻¹ to 469.61 mg L⁻¹ in Anhui; 207.92 mg L⁻¹ to 764.17 mg L⁻¹ in Hunan) facilitated solid sedimentation and the formation of a stable sludge bed. Meanwhile, SEM analysis (Supplementary Fig. S2) indicated that lipids were primarily adsorbed onto the packing materials and sludge surfaces, while hydrolyzed proteins were immobilized through coagulation and integration with the activated sludge. These processes enhanced organic–sludge interactions, with the SS particulates serving as a nutrient source for the UASB reactor, thereby stabilizing overall system performance. In addition, the above processes contributed to pH stabilization in subsequent treatment stages. The UASB reactor effluent maintains near-neutral values (7.37 in Anhui and 7.36 in Hunan), creating favorable conditions for methanogenesis. With a sufficient carbon source supply and stable operating conditions, UASB reactors achieved enhanced biogas production with high methane purity (83.94% in Anhui and 83.20% in Hunan).

Energy consumption profiles underscored the sustainability of the TPAD system, with comparative analysis revealing a notable dependence on operational mode. The specific energy demand was substantially lower in the continuously-fed Hunan system (0.28 kWh m⁻³) than in the batch-operated systems of Yunnan (0.66 kWh m⁻³) and Anhui (0.68 kWh m⁻³) (Fig. 4a–c). This clear difference highlights the influence of the feeding strategy. The lower energy consumption of the Hunan plant, which successfully met a less stringent discharge standard, demonstrates a key adaptability feature of the process: sustainability can be achieved by optimizing the balance between energy input and specific treatment objectives, rather than solely pursuing maximal removal efficiency at a high energy cost. The primary energy consumption across all systems was attributed to the contact oxidation tanks, sludge dewatering, and pumping systems. Overall, organic retention within the TPAD process reduced energy use by 32.02–41.01% compared to conventional systems³⁰, with the Hunan configuration showcasing exceptional operational adaptability. Meanwhile, Biogas production was highest in the Anhui system, which yielded 3.43 times more methane than the Yunnan system—a difference that corresponded to a 2.97 times greater wastewater treatment capacity (Fig. 4d, Supplementary Fig. S5 and Table S2). The methane content in the biogas reached 83.94% in Anhui and 83.20% in Hunan, exceeding values reported in the literature by 20.01–23.03%^42,43. Furthermore, strong negative linear correlations (Pearson’s |r| > 0.85, p < 0.01) between COD removal and biogas production rates (0.48 m³ biogas kg⁻¹ COD in Anhui; 0.35 m³ biogas kg⁻¹ COD in Hunan) were observed (Supplementary Fig. S3), further confirming the high metabolic efficiency of the process.

Fig. 4. Electricity consumption and biogas production at Yunnan, Anhui, and Hunan provinces of TPAD (two-phase anaerobic digestion) system.

a–c Month average electricity consumption (EC, orange-pink pentagon) and month average treatment water volume (blue hexagon) in Yunnan, Anhui, Hunan respectively, and pink-purple triangle represent electricity consumption per ton of water. d The yellow bar chart and blue bar chart represent biogas production and treatment water volume respectively and pink circle represent biogas production rate.

Microbial community structure and metabolic pathway analysis

High-throughput sequencing of activated sludge from laboratory-scale and full-scale TPAD systems revealed critical microbial consortia and their niche-specific functional roles. In full-scale ABRs, hydrolytic and acidogenic phyla, including Proteobacteria (3.94–56.68%), Bacteroidota (11.37–26.36%), Firmicutes (5.61–8.09%), and Chloroflexi (5.02–39.63%), exhibited functional complementarity in maintaining system stability⁴⁴ Proteobacteria facilitated glucose and long-chain fatty acid degradation⁴⁵. Meanwhile, Bacteroidota and Firmicutes rapidly proliferated in organic-rich environments to hydrolyze polysaccharides and proteins into short-chain fatty acids (SCFAs)⁴⁶. The high abundance of Chloroflexi in Hunan suggests a potential role for (DIET)⁴⁷, a specialized form of IET, in facilitating syntrophic metabolism^48,49. The establishment of efficient IET, potentially including DIET, could be a key reason for the effective degradation of proteins and lipids observed in this project, as reflected in the increased ammonia concentration (Fig. 3b).

The observed geographic variations in microbial dominance directly correlate with functional performance adaptations. In Yunnan, the dominance of Proteobacteria (56.68%) facilitated rapid hydrolysis of readily degradable organics, resulting in high COD removal rates (91.34–93.46%). Conversely, Hunan’s Chloroflexi-rich community (39.63%) appears to enhanced DIET-mediated syntrophic metabolism, which is particularly beneficial for complex substrate degradation under varying organic loads. These distinct community structures represent optimized adaptive strategies. Each configuration achieves efficient wastewater treatment through a specialized metabolic route, explaining the consistent performance.

Geographic variations in phylum abundance were apparent (Fig. 5a–c). Yunnan’s microbial community was dominated by Proteobacteria (56.68%), whereas Anhui displayed a balanced coexistence of Bacteroidota (26.36%) and Chloroflexi (20.18%). Hunan’s microbial ecology was unique, featuring a synergy between Chloroflexi (39.63%) and Bacteroidota (11.37%). Furthermore, the abundance of Patescibacteria (13.80%) in Hunan was markedly higher than in Yunnan and Anhui (≤2.88%), suggesting adaptations driven by the local hydraulic regime⁵⁰ (Supplementary Table S13).

Fig. 5. Bacterial phylum level and archaeal genus level in TPAD (two-phase anaerobic digestion) from Yunnan, Anhui, and Hunan provinces.

a Bacteria phylum of TPAD of Yunnan province (Y-ABR, Y-UASB). b Bacteria phylum of TPAD of Anhui province (A-ABR, A-UASB). c Bacteria phylum of TPAD of Hunan province (H-ABR, H-UASB). d Archaea genus of TPAD of Yunnan province (Y-ABR, Y-UASB). e Archaea genus of TPAD of Anhui province (A-ABR, A-UASB). f Archaea genus of TPAD of Hunan province (H-ABR, H-UASB). (Note: A in the panels refer to anaerobic baffle reactor-ABR, U refer to up-flow anaerobic sludge blanket-UASB reactor).

Although ABRs are primarily designed for acidogenesis, they naturally develop minor methanogenic niches due to microbial migration. This activity contributes to preliminary methane production and helps maintain redox balance, without compromising the overall phase separation efficiency. This phenomenon, observed in other TPAD systems (Supplementary Table S17), indicates that while complete phase separation is theoretically ideal, it is challenging to achieve in practice.

In full-scale UASB reactors, methanogenic archaeal communities demonstrated site-specific specialization (Fig. 5d–f). Acetoclastic methanogens dominated across all sites. Methanosaeta was prevalent in Yunnan (54.04%) and Hunan (56.38%), whereas Methanothrix accounted for 84.46% of the archaeal population in Anhui. Hydrogenotrophic methanogens, including Methanoregula (3.27–25.78%), Methanolinea (4.78–22.11%), and Methanobacterium (4.97–6.04%), occupied secondary roles. A finding that was consistent with laboratory-scale observations (Supplementary Table S14). The exceptional dominance of Methanothrix in Anhui correlated well with its superior biogas yield (Fig. 4d), highlighting the efficiency of metabolic specialization. Substrate availability further shaped these community profiles: elevated acetate levels in Anhui (Supplementary Table S3) supported Methanothrix proliferation, while the diverse methanogen composition in Yunnan aligned with its more heterogeneous organic inputs.

KEGG pathway analysis (full-scale ABR-UASB reactor) identified propionate (map00640) and pyruvate metabolism (map00620) as central acidogenic routes in full-scale systems (Fig. 6a,). Propionate conversion involved succinyl-CoA synthetase (EC:6.2.1.5) catalyzed succinate synthesis, followed by pyruvate generation via tricarboxylic acid cycle intermediates. Hypoxia promoted pyruvate dehydrogenase (EC:1.2.4.1) mediated acetyl-CoA production, with subsequent acetate formation by acetyl-CoA hydrolase (EC:3.1.2.1). Quantitative analysis of key genes revealed distinct phase-specific enrichment patterns across the projects (Fig. 6b, c). A comparative analysis revealed that the relative abundance of gene K01902 (succinyl-CoA synthetase, EC:6.2.1.5) was enriched in the acidogenic phase relative to the methanogenic phase, with increases of 10.34% in Anhui and 3.86% in Hunan. Similarly, the relative abundance of the gene K01067, encoding the pyruvate metabolism key enzyme acetyl-CoA hydrolase (EC:3.1.2.1), was 24.87% greater in the acidogenic phase in Anhui and 3.73% greater in Hunan, compared to their respective methanogenic phases. In contrast, both genes showed lower relative abundance in the acidogenic phase in the Yunnan project and in laboratory-scale tests (Fig. 6b, c and Supplementary Table S15).

Fig. 6. Metagenomic analysis of methane metabolism.

a Methanogenic pathways based on KEGG (kyoto encyclopedia of genes and genomes, map00680). b Propionate metabolism (map00640) related KEGG Orthology (KO) of Yunnan (Y-A, Y-U,), Anhui (A-A, A-U), and Hunan (H-A, H-U) provinces. c Pyruvate metabolism (map00620) related KO of Yunnan, Anhui, and Hunan provinces. d Hydrogenotrophic methanogenic (M00567) related KO of Yunnan, Anhui, and Hunan provinces. e Acetoclastic methanogenic (M00357) related KO of Yunnan, Anhui, and Hunan provinces. f Environmental impacts of the TPAD (two-phase anaerobic digestion) system in Hunan and Anhui compared to the single-phase anaerobic digestion system in Kunming (treatment of 1 ton of COD as a functional unit). (Note: A in the panels refer to anaerobic baffle reactor-ABR, U refer to up-flow anaerobic sludge blanket-UASB reactor).

Regarding methanogenesis, the acetoclastic (M00357) and hydrogenotrophic (M00567) pathways were identified as the principal routes (Fig. 6a), which is consistent with recent findings^13,51,52. Acetoclastic methanogens employ acetyl-CoA synthetase (EC:6.2.1.1) for acetate activation, while hydrogenotrophic methanogens utilize multi-enzyme cascades for CO₂ reduction. The elevated expression of genes associated with the acetoclastic pathway (Fig. 6d, e) corresponded well with the predominance of Methanosaeta/Methanothrix observed in the archaeal community (Fig. 5d–f). Furthermore, the enhanced acetate levels in Anhui (Supplementary Table S3), coupled with its superior biogas productivity (Fig. 4d), confirmed an efficient conversion of propionate to acetate within the system.

Based on above, the ABR-UASB reactor configuration enhanced system functionality through phased substrate channeling. Specifically, acidogenic SCFAs production in ABRs coupled with their efficient utilization driven by the UASB reactor, optimized acetoclastic methanogenesis. This spatial segregation reduced hydrogen partial pressures, thereby minimizing competitive inhibition of acetoclastic archaea. Geographic and operational parameters collectively shaped microbial consortia assembly, exemplified by Chloroflexi-mediated DIET in Hunan and Methanothrix-dominated methanogenesis in Anhui. These adaptive strategies to local substrate regimes highlight the system’s resilience. The operational efficiency is ultimately underpinned by this spatial segregation and the resulting syntrophic partnerships.

Sustainability analysis and environmental applications

Operational data confirmed the superior overall efficiency of the designed TPAD system compared to the conventional single-phase system. This enhanced performance was evident across energy, treatment, and economic dimensions. In terms of energy, the TPAD system achieved a lower specific energy demand (0.28–0.68 kWh m⁻³ for Hunan/Anhui implementations vs. 1.87 kWh m⁻³ for the Kunming benchmark, Fig. 4a–c), representing a 63.64–85.03% reduction. Critically, this reduction was complemented by active energy recovery. As the biogas produced was directly used in on-site gas boilers to generate process hot water, thereby displacing fossil fuel consumption. Regarding treatment performance, the process maintained high COD removal (78.02–99.01%, Figs. 2d and 3a) alongside a substantial reduction in residual sludge production (>75.00%). Economically, the system demonstrated viability with an investment payback period of approximately 4 years. This was realized through combined benefits from biogas revenue (0.35–0.48 m³ biogas kg⁻¹ COD; Fig. 4d) and reduced operational costs for energy and chemicals (Supplementary Table S17). Collectively, these multi-faceted advantages underscore the TPAD system’s enhanced sustainability and practical feasibility.

LCA revealed TPAD’s environmental superiority over the single-phase system (Kunming project), demonstrating lower impact scores across most indicators (Supplementary Fig. S6). Taking global warming potential (GWP) as an example, TPADs achieved 88.01-97.09% reductions in GHG emissions through suppressed CO₂ emissions and high-purity methane recovery (83.20–83.94%) (Fig. 6f and Supplementary Fig. S5). Furthermore, Anhui project outperformed Hunan in environmental benefits due to stricter emission standards, where its UASB reactors introduced additional carbon sources through high-flow influent. These advantages collectively minimized energy/resource consumption and environmental impacts in dairy wastewater treatment.

The Hunan configuration proved particularly impactful as a budget-adaptive model for developing regions. It maintained 78.13% COD removal efficiency while delivering biogas productivity and environmental performance comparable to Anhui system. With its demonstrable reductions in operational energy intensity, direct fossil fuel displacement, verifiable resource savings, and economic viability, the TPAD system establishes itself as a dual-purpose solution. Its alignment with SDGs provides policymakers a ready framework to address the climate-water-energy nexus in global dairy production.

Discussion

The ambient-temperature TPAD system fundamentally addresses three persistent challenges in DW treatment while establishing a scalable framework for sustainable dairy industrialization. This configuration integrates the compartmentalized hydrolysis-acidification of an ABR, which effectively retains lipids and proteins, with the methanogenic specialization of a UASB reactor. Without chemical additives, the system achieved a high methane content of up to 83.94% in full-scale operation, representing a 21.03% improvement over conventional thermophilic systems⁵³. The higher performance originates from three interconnected innovations with global sustainability implications.

First, microbial self-organization drives two key processes: Proteobacteria-led hydrolysis-acidification in the ABR, and the enrichment of acetoclastic methanogens in the UASB reactor. This synergy prevents reactor acidification and sludge bulking while outperforming single-phase reactors and other TPAD configurations^12,42,53 (Supplementary Table S17). Such autonomous regulation eliminates chemical dependency and reduces the need for specialized operators, which is a benefit for small and medium-sized enterprises.

Second, the system shows strong operational adaptability. It maintains stable COD removal (78.02–99.01%) at shorter HRT than conventional benchmarks (24-96 h vs. 7–14 days^11,12). This allows consistent effluent quality that supports SDG 6, as shown by three implementations meeting local discharge standards.

Third, the system successfully integrates circular economy principles through biogas recovery (0.35–0.48 m³ kg^-1 COD) and a greater than 75.00% sludge reduction, it approaches carbon neutrality within a 4-year payback period, surpassing strategies focused solely on VFAs⁵⁴. An LCA accounting for both phases confirmed this, showing an alignment with SDG 7 via a 64.71–85.03% reduction in energy demand (0.28–0.66 vs. 1.87 kWh ton^-1).

The design principles of the present TPAD system suggest its promising potential for treating other high-strength wastewaters (e.g., from breweries, soybean processing, or mixed food-industry effluents; Supplementary Table S17), enhancing its international relevance. A key advantage for such applications is the ABR’s role as a robust pretreatment stage, whose compartmentalized structure is effective in retaining and degrading a wide spectrum of complex organics (e.g., carbohydrates, proteins, and lipids), thereby creating favorable conditions for subsequent stable methanogenesis in the UASB reactor. This mechanism, reliant on physical retention and microbial self-organization, endows the system with greater resilience to variations in wastewater composition compared to processes dependent on specific chemical additives or stringent temperature control. The enriched facultative and anaerobic hydrolytic bacteria (e.g., Proteobacteria), while investigated here in DW, are ubiquitous in systems treating diverse food wastewaters. This suggests a broader applicability of the microbial self-regulation principle beyond the dairy industry. Future work should validate the system’s performance with mixed food effluents of varying carbon-to-nitrogen ratios and lipid contents, and investigate the associated microbial community dynamics. This research is crucial for promoting broader adoption across the global food industry.

The TPAD system’s decarbonization potential extends beyond recent technical advantages. Given projected dairy sector emissions (868–2604 Mt CO₂ yr⁻¹ by 2033; Supplementary Table S17), the contribution of wastewater treatment (8.68–78.12 Mt CO₂ yr⁻¹) represents a strategic intervention point. Implementing the designed TPAD could mitigate 7.64–75.80 Mt CO₂ yr⁻¹, equivalent to 0.88–2.91% of total dairy lifecycle emissions. This mitigation potential is particularly important for small and medium enterprise clusters, where the system’s 88.01–97.09% GHG reduction capability could enable carbon credit generation under evolving climate policies. Furthermore, the modular design offers climate resilience (adaptable from temperate to alpine tundra zones) and compatibility with international standards (e.g., China’s GB/T 18918-2002 vs. 31962-2015), supporting global deployment and benefiting emerging dairy regions through SDG 12-aligned resource recovery.

Industrial validations confirm 12–18 months of operational stability without exogenous additives. For instance, the cost-effective Hunan configuration achieved 78.13% COD removal and energy recovery (0.35 m³ biogas kg⁻¹ COD) under stringent budget constraints, outperforming membrane-dependent systems²² and recirculation-based TPAD designs³⁴. The substantial sludge reduction also lowers digestate disposal costs and facilitates reuse, offering clear economic and environmental advantages for small and medium-sized enterprises. These demonstrations show that microbial self-regulation does more than prevent technical failures. It democratizes access to advanced treatment by bridging the technological gap between large industrial processors and small-to-medium enterprise operations. By converting baseline emissions into climate action opportunities, the TPAD repositions wastewater management from a compliance cost to a value generator, empowering small and medium-sized enterprises and developing economies to align operational viability with SDG 13 commitments.

In conclusion, the TPAD system bridges laboratory innovation and industrial scalability through microbial self-regulation, transforming wastewater management into a strategic asset. As quantitatively compared with other TPAD and conventional systems in Supplementary Table S17, our configuration excels in methane yield, operational stability at ambient temperature, and sludge reduction without exogenous additives. Its technical reliability stems from self-organizing microbial consortia, while modular, low-capital-expenditure designs ensure economic accessibility. Scalable GHG mitigation across dairy industry tiers enhances climate relevance, collectively addressing the climate–water–energy nexus within dual-carbon frameworks. The intrinsic microbial resilience ensures long-term adaptability, positioning TPAD as a cornerstone for sustainable industrialization.

Methods

TPAD systems simulation

CFD technology has become a key tool in water treatment research, capable of accurately analyzing complex fluid dynamics^55–57. This study employed Ansys Fluent’s Euler multiphase model and RNG k–ɛ steady-state flow numerical model to conduct two-phase flow simulations of hydraulic flow and sludge retention in the ABR during inoculation. The specific operational procedures are as follows: (1) Scale modeling of the internal fluid in the ABR (length 320.00 mm, width 160.00 mm, height 160.00 mm); (2) Meshing using poly-hexcore; (3) Setting boundary conditions with velocity inlet and pressure outlet; (4) Designating the remaining model as internal fluid; (5) Defining water phase properties (with densities of 1050.00 kg m⁻³) through material interfaces; (6) Then changing the secondary phase name from “water” to “sludge” (with densities of 0.02 kg m⁻³), while configuring sludge characteristics (water content 0.60, particle size 0.00020 m, particle viscosity 0.20 Pa s).

Configuration and operation of laboratory-scale and engineering projects

The laboratory-scale TPAD system comprised a four-compartment ABR and UASB reactor, inoculated with 600 mL of activated sludge from a municipal wastewater plant and operated for 120 days without sludge discharge (Fig. 1a). The start-up stage with a 24 h HRT and influent COD of 1300.00 ± 88.56 mg L⁻¹, and the stable operation phase with a 30 h HRT and influent COD of 1640.87 ± 135.69 mg L⁻¹. Peristaltic pumps control the water intake and discharge, and sampling ports for water and sludge were installed in the ABR and UASB reactors for real-time monitoring.

The treatment process workflow of the engineering projects deployed in Yunnan, Anhui, and Hunan was illustrated in Fig. 1b. Specific unit configurations of anaerobic systems for individual projects were detailed in Supplementary Table S1. Biogas production was quantified via flow meters, while acetate/propionate concentrations and composition were analyzed using gas chromatography (GB 17820-2018).

Wastewater composition and analytical methods

Laboratory substrates simulated industrial effluent by blending milk with micronutrient salts (e.g., ZnSO₄·7H₂O, Fe(NH₄)₂SO₄, CoCl₂·6H₂O, MnCl₂·4H₂O, NiCl₂·6H₂O, and CuSO₄·5H₂O) and supplements (e.g., KH₂PO₄ and urea) (Supplementary Table S16). During the initial start-up phase, the influent exhibited mean concentrations of 16.66 ± 7.80 mg L⁻¹ NH₃–N and 9.01 ± 0.63 mg L⁻¹ total phosphorus (TP), with concomitant COD and pH values of 1300.00 ± 88.56 mg L⁻¹ and 7.24 ± 0.06, respectively. Stable operation phase maintained elevated nutrient levels at 17.6 ± 2.24 mg L⁻¹ NH₃-N and 10.16 ± 1.29 mg L⁻¹ TP, alongside increased organic loading (COD: 1640.87 ± 135.69 mg L⁻¹) and stabilized alkalinity (pH: 7.52 ± 0.15). The influent water quality of engineering projects varies with production conditions.

In the laboratory-scale trials, influent and effluent samples were collected every other day and analyzed for key parameters (e.g., COD, NH₃–N, pH) to monitor the long-term performance of the TPAD system. Following collection, samples underwent centrifugation (5000 rpm, 10 min) to isolate supernatants, which were subsequently filtered through 0.45 μm membranes prior to further analyses. pH was measured using a portable pH meter (Leici, YHBJ-262, China). COD, NH₃–N and TP were determined according to standard methods (APHA, 2005). In engineering projects, real-time COD, NH₃–N, TP, SS, and pH data were acquired via online sensors. Methane concentration in the biogas was analyzed using a gas chromatograph equipped with a thermal conductivity detector (Haixin, Shanghai, China). VFAs were quantified by gas chromatography with a flame ionization detector (Shimadzu, Kyoto, Japan) employing an HP-FFAP column (Agilent, Santa Clara, CA, USA).

High-throughput sequencing analysis

Activated sludge samples from both laboratory and engineering-scale reactors were collected and subjected to 16S rRNA amplicon sequencing. The experimental workflow comprised the following steps: DNA extraction was performed using the E.Z.N.A™ Mag-Bind Soil DNA Kit, targeting the V3–V4 hypervariable regions of the bacterial 16S rRNA gene. Following DNA extraction, PCR amplification was initiated using forward primer CCTACGGGNGGCWGCAG and reverse primer GACTACHVGGGTATCTAATCC. The paired-end sequencing reads were first merged based on overlapping regions, followed by demultiplexing, quality control, and filtration. PCR amplification was conducted on the Sangon Biotech platform (Shanghai, China). KEGG Orthology (KO)-based metabolic processes and their associated enzyme quantification profiles were systematically mapped through KEGG’s bioinformatics annotation system.

LCA assessment

Following ISO 14040/14044 standards, SimaPro 9.5 and the Ecoinvent 3.9.1 database were used to evaluate the environmental impacts. The TPAD (Anhui/Hunan) with a single-phase ABR system (Kunming, without biogas collection and utilization) was comparatively analyzed. System boundaries encompassed energy/chemical consumption, biogas recovery, and sludge management (Fig. 1b), all sourced from actual project monitoring and statistics. Considering the negligible environmental impact of AD plant construction on the overall system, this factor was excluded. GWP was quantified using CML-IA baseline V3.09/EU25, excluding CO₂ from biological processes⁵². Environmental impact results were normalized to the functional units of 1-ton COD removal and 1-ton wastewater treatment. Detailed life cycle inventory data were provided in Supplementary Tables S4–S6.

Reporting summary

Further information on research design is available in the Nature Portfolio Reporting Summary linked to this article.

Author contributions

S.P. Ji directed the project. S.P. Ji and Y.Q. Gong designed the experiments and wrote the paper. L. Li and X.C. Guo conducted modeling of ABR reactor using Ansys Fluent. Y.Z. Guo, Y.F. Ren, and P.P. Huang performed LCA analysis. S.P. Ji, F.Z. Jiang, Y.B. Liu, and T.F. Xiao further provided constructive suggestions for the paper revision.

Peer review

Peer review information

Communications Engineering thanks Debkumar Chakraborty, Camila Menezes and the other, anonymous, reviewer(s) for their contribution to the peer review of this work. Primary Handling Editors: [Philip Coatsworth].

Data availability

Data will be made available on reasonable request.

Competing interests

The authors declare no competing interests.

Supplementary information

The online version contains supplementary material available at 10.1038/s44172-025-00568-2.