Measuring greenhouse gas emissions from wastewater treatment plants is of utmost importance in the context of climate change. However, due to their variability and complexity, it is a particularly challenging task in aerated reactors. The current methods involve capturing gas emissions from the water surface, measuring gas flow rates, and determining the concentration of the emitted gas at that location. Our study proposes a new, more efficient method that eliminates the need for gas flow rate measurements and additional equipment. The proposed technique uses a gas analyzer and a specially designed floating chamber to measure the transient trend of gas concentration within the chamber from the moment it is deployed to when it reaches a new steady state. Our research shows that this method accurately determines methane and carbon dioxide emissions from aerated reactors and potentially other gases emitted in wastewater treatment plants. It is cost effective, versatile, and simplifies the measurement process. This method facilitates the assessment of greenhouse gas emissions in wastewater treatment plants. Our findings are backed by comprehensive testing in the aeration tanks of a full-scale activated sludge plant, across diverse conditions, including fine- and coarse-bubble aeration.

  • A simple chamber method is proposed to measure gas emissions in aerated reactors.

  • The method is designed for surface gas flow rate and concentration determination.

  • It utilizes only a chamber and a gas analyzer for simplicity and cost savings.

  • Together, these features offer a simplified tool for characterizing large-scale reactors.

  • The method was successfully applied to a full-scale wastewater plant.

Methane (CH4) and carbon dioxide (CO2) are greenhouse gases (GHGs) that play a crucial role in climate change. Monitoring their emissions is essential for understanding and addressing their impact in a global context. Emissions generated in the waste sector are a major contributor to the global GHG budget, and it is estimated that wastewater treatment plants generate about 5% of the global CH4 emissions (Ye et al. 2022), which are estimated to be 582 Tg (IPCC 2019). These estimates are primarily based on the Tier 1 method (IPCC 2006), which combines default emission factors, activity data, and/or simple models that require minimal data inputs. Compared to Tier 1, a migration to Tier 2 or Tier 3 methods would be beneficial for improving the precision and reliability of current estimates. This transition holds the potential for substantial benefits in enhancing current estimates. However, it comes at the expense of requiring more intricate and specific data, mandating measurements tailored to the unique characteristics of each process unit, site, facility, and region. Consequently, measuring gas emissions from specific process units in wastewater treatment plants is not only important but also a formidable task due to the diversity of unit operations involved. Among these units, determining emissions from aerated reactors poses challenges. This is attributed to the continuous injection of airflow, demanding precise quantification of the off-gas flow rate and its composition in a highly turbulent and heterogeneous medium.

As documented elsewhere, various methods have been suggested for this purpose (Zhan et al. 2017; Pavelka et al. 2018; Ye et al. 2022). Most of these methods rely on capturing gas emissions from the reactor's water surface, combined with measuring the gas flow rate and using an analyzer to determine gas composition. These methods are commonly based on ‘open flux chambers’ (OFCs), devices floating on the water's surface to capture the gas emitted and connected to a gas analyzer. While the question of which gas analyzer to use is straightforward and essentially a commercial choice, measuring the gas flowing through OFCs might be challenging, and several approaches have been suggested, as listed by Parravicini et al. (2022). There are three main methods for measuring the gas flow rate in an OFC, and the first and most obvious one is the direct measurement of the gas flow rate (Chandran 2011; Foley 2015; Pan et al. 2016; van Loosdrecht et al. 2016), which can be difficult due to the highly dynamic and variable nature of the gas flow. Measuring the gas flow rate may also require a pressure measurement device in addition to a flow meter, since changes in pressure might affect the measurements. The second general method involves using a tracer gas, injected into the OFC at a known flow rate. The tracer gas concentration in the chamber outflow is measured to estimate the dilution factor and determine the actual gas flow rate emitted by the aerated surface (Rodriguez-Caballero et al. 2014; Ribera-Guardia et al. 2019; Tauber et al. 2019). This method generally requires a gas cylinder, a flow controller to deliver the tracer gas at a fixed flow rate, and an additional analyzer for the measurement of the tracer gas concentration, which need to be sensitive enough to measure it at a relatively high dilution ratio. Both direct flow rate measurement and tracer gas methods make the design of OFCs more complex. They require additional equipment, connections, and/or flexible tubing in addition to the target gas analyzer, resulting in a more complex setup operation. In addition, in some cases, these methods are limited to discrete measurement of the gas flow rate, which may be subject to a relatively large error compared to continuous measurement. The third option for measuring the air flow rate is to determine it at the level of the compressors, which is a simple approach but lacks spatial resolution. Lastly, regarding methods for flux measurement, it is worth noting that some analyzers, of the target gas and/or the tracer gas, might require a gas conditioning unit to remove moisture and/or other gases that might interfere with the measurement (Nguyen et al. 2019; Parravicini et al. 2022).

In this study, we introduce a simplified method for precisely quantifying point-specific gas flow rates and target gas concentration at the surface of aerated reactors. Our approach eliminates the need for extra equipment beyond an OFC and a gas analyzer, removing the necessity for flowmeters, tracer gases, pressure measurements, or gas conditioning units. The main objective pursued was to reduce the equipment list required and, therefore, to cut costs. Simultaneously, we aimed to simplify the design of the OFC by reducing the number of connections needed, enhancing overall versatility and ease of operation. The underlying concept of the method is to use a simple mass balance approach to characterize the target gas concentration within the OFC during the transient period observed from the moment the OFC is deployed until it reaches a steady state. During that period, the target gas concentration will pass from values close to atmospheric concentration to the concentration of the off-gas at the reactor surface. The rate and magnitude of this change provide information about the off-gas flow rate and its composition, respectively.

The testable hypothesis of the present work is that gas emissions from aerated reactors can be measured in a simpler manner than the standard methods previously described. The method was tested for CH4 and CO2 emissions in a full-scale activated sludge plant, specifically in two aeration tanks, hereafter referred to as aerated reactors. Both reactors treated the same wastewater and were operated under a plug-flow configuration. One reactor was equipped with coarse-bubble aeration involving high and heterogeneous gas flow rates, while the other was operated with fine-bubble aeration, with a relatively lower and more homogeneous gas flow rate. Two OFCs were specifically designed for these two scenarios; their details are presented in this report (supplementary material). As demonstrated in this work, the developed method effectively measured GHG emissions from aerated reactors. Importantly, its potential significance extends beyond GHG to include the ability to measure various target gases, pending the integration of a suitable detector. This expanded list encompasses ammonia, hydrogen sulfide, volatile organic compounds, and off-gas oxygen measurements, offering valuable insights into the functioning of the reactor.

Simplified open flux chamber

The innovative approach suggested here employs a relatively standard open flux chamber strategically positioned on the surface of an aerated reactor. The gas emitted at the surface of the reactor gradually replaces the atmospheric air in the chamber's headspace. The mass balance of that process can be described as follows. Once the floating chamber (Figure 1) is positioned at the surface of the water, the chamber headspace at a given CH4 and CO2 concentration (CC), respectively, and , receives the off-gas flow rate (QG), proportional to the aeration rate of the wastewater plant and the area of the chamber (AC). This gas inflow contains CH4 and CO2 at a given concentration (CG), respectively, and , which mixes with the gas already contained in the chamber.
Figure 1

Conceptual scheme of the simplified OFC chamber.

Figure 1

Conceptual scheme of the simplified OFC chamber.

Close modal

Both and , will then change over time from an initial value to a final value equal to CG. To maintain a consistent chamber volume, a vent allows for continuous gas extraction at a rate equivalent to the inflow rate, ensuring pressure equilibration. A gas analyzer, connected to the chamber, samples a small flow rate from the chamber headspace for analysis and returns it to the chamber.

The mass balance of the gas phase within the chamber can be written as follows:
(1)
where VC is the volume of the chamber headspace, and QV is the flow rate of the gas extracted through the vent.
Assuming mass conservation and a constant chamber volume, the total gas inflow can be considered equal to the total gas outflow (QG = QV), and Equation (1) can be rewritten as follows:
(2)
In Equation (2), the term QG/VC is the inverse of the gas residence time (θG) within the chamber, in such a manner that:
(3)
Assuming that CG is a constant, over the measurement time, we can change the derivative constant, as follows:
(4)
After integration, we obtain:
(5)
And finally:
(6)
To visually represent Equation (6), Figure S1 (supplementary material) illustrates the CC trend observed immediately after placing the chamber on the water surface. The trend evolves from an initial value, , to a pseudo-steady state, CG. It is crucial to emphasize that Equation (6) is not contingent on any specific initial condition. Consequently, by calibrating Equation (6) with any segment of experimental data, values for CG and θG can be determined. This, in turn, facilitates the calculation of the flux emitted by the water surface, as outlined in Equation (7). It is noteworthy that the calibration process of Equation (6) does not necessitate any flow rate measurements:
(7)
To account for gas concentration in the atmospheric air that is blown to the reactor, Equation (7) must be corrected considering the ambient air concentration (CA), respectively, and :
(8)

Experimental setup

Several chamber designs were tested. A chamber with a relatively standard shape (Figure S2) was proven effective for reactors with fine-bubble aeration. That chamber (OFCA) was self-made from a steel sheet to ensure a significant weight to offer stability (weight, 6 kg; AC, 0.071 m2; VC, 0.011 m3), and a tire tube was used as a floater. In moderately agitated reactors, the OFCA chamber had a draft of 12 cm, which was sufficient to prevent contact between the atmosphere and the chamber headspace. The chamber was maintained at a fixed position with two cords controlled from the perimeter path of the reactor. In reactors with coarse-bubble aeration, the strong mixing and intense water currents caused OFCA to flip and tilt easily. From assay/errors, a tubular OFC chamber (OFCB) was designed (Figure S3; total weight, 10 kg; AC, 0.035 m2; VC, 0.009 m3). This chamber included a relatively long tube, a draft of 81 cm, and a 4 kg counterweight made of stones packed in a plastic tissue bag. The OFCB floaters were made of three cylindrical foam tubes (pool noodles), closely encircling the chamber. To ensure that the OFCB remained in a fixed position and to prevent flipping or excessive tilting, three cords were secured to the top of the chamber, and three more cords were fixed at the bottom, operated from the reactor's perimeter path.

The chambers were connected in a closed circuit to an ultraportable greenhouse gas analyzer (UGGA, model 915-0011, Los Gatos Research, USA) that measured CH4 and CO2 concentration at a frequency of 1 Hz, with data acquisition in an internal hard drive. The closed-circuit configuration ensured that the gas flow rate sampled was returned to the chamber after measurement without affecting the gas mass balance. This analyzer is equipped with an internal pump (1 L min−1) and has a limit of detection (LOD) of 10 ppb (parts per billion) for CH4 and 300 ppb for CO2. It is worth noting that the UGGA also measured water vapor so that the off-gas CO2 and CH4 concentrations were provided per unit of dry air. It is also important to mention that no gas treatment was required, except the use of a hydrophobic filter (Acrovent 0.22 μm, Pall Corp, Mexico) placed at the input of the UGGA to prevent liquid water entrance. The connection between the chambers and the UGGA was made with a 6 mm external diameter, 4 mm internal diameter polyurethane dual tubing (PUN-6X1-DUO, Festo, Mexico), with a variable length (5, 10, or 15 m) depending on the measurement location selected.

Wastewater treatment plant

The simplified OFC method was thoroughly tested in the ‘San Juan de Aragon’ wastewater treatment plant (WWTP), which is a conventional activated sludge plant with two independent treatment trains, differing only in the type of aeration used in the biological reactors (secondary treatment), i.e., one train operated with coarse-bubble aeration and one train operated with a recently installed fine-bubble aeration system. This WWTP is located within Mexico City (lat. 19.4575; long. −99.0975) and treats primarily urban wastewater with some industrial effluents at a mean flow rate of 12,500 m3 day−1. The WWTP includes primary settlers, aerobic reactors, and secondary settlers with a final chlorination unit. The biological reactors have a plug-flow U-shaped configuration with a total length of 66 m and a total volume of 3,124 m3 each. During the measurement campaign, the hydraulic residence time of the reactors (θH) was 5.98 ± 0.74 h.

Experimental protocol

The simplified OFC method was tested in both reactor types of the WWTP. The UGGA was operated continuously. Before starting flux measurements, , and were measured in the compressor room for 5 min. For each flux measurement, the chamber was suspended above the desired position using the cords and then submerged in a vertical position for about 5 min. Each measurement was repeated three times, and between each measurement, the chamber was ventilated for 1 min above the water surface. A total of 90 measurements were conducted within each reactor, involving triplicate measurements at 10 specific locations on 3 different dates. This comprehensive approach covered the entire length of the reactors. In this study, measurement locations were correlated with the water residence time, using the ratio t θH−1 where 0t θH−1 ≤ 1.

Data treatment and sensitivity analysis

Data provided by the UGGA were treated according to the following protocol. Data corresponding to each measurement were extracted, considering the measurement delay, i.e., the time required for the sampled gas from the chamber to reach the detector, typically 15 s. For each CH4 and CO2 dataset, Equation (6) was adjusted to the experimental data, using CG and θG as calibration parameters, minimizing the root mean square error (RMSE) with the Excel solver function.

The goodness-of-fit between the observed and the simulated CG data was evaluated through the nt parameter (Ritter & Muñoz-Carpena 2013), which is an indication of the number of measurements where the variability of the observations is greater than the mean error of the model:
(9)
where SD is the standard deviation of the observed CG values.

According to Ritter & Muñoz-Carpena (2013), the model adjustment was categorized as excellent for nt > 2.2, good for nt in the range of 1.2–2.2, acceptable for nt in the range of 0.7–1.2, and unacceptable for nt < 0.7. From this classification, all data with a nt < 0.7 were discarded. It is worth noting that, after calibration of Equation (6), both and datasets provided a CG and θG value for each gas (CH4 and CO2). In the latter case, and are analogous and were averaged. Our strategy for averaging θG values involved considering the mean of and if both values were determined to be acceptable, with a nt greater than 1.2. In any other case, we selected the θG values determined with the best nt, and only if greater than 0.7.

To investigate whether CG and θG can be independently estimated from a single experiment, we performed a numerical sensitivity analysis of Equation (6). First, we generated a theoretical dataset of CC values with fixed CG and θG parameters, spanning a time range of 0 − 10 times θG with increments of 0.1 θG. Next, we perturbed each CC value by 10% and re-estimated CG and θG for each perturbed dataset. We computed the impact on CG and θG for each perturbation by determining the ratios ΔCCCG and ΔCCθG, respectively.

Statistical comparisons among results were determined according to Tukey's honest significance test. Most variables exhibited a positive skew, which was corrected by log-transforming them to achieve normality for the analysis of variance (ANOVA). This analysis was performed with Origin(Pro) software (Version 2016, Northampton, USA).

Method sensitivity

The numerical sensitivity analysis of θG and CG (Equation (6)) is illustrated in Figure 2. The local maximum in sensitivity observed for θG at t = θG suggests that the accuracy of θG estimation will be highest for the measurements taken around that time. On the other hand, CG sensitivity exhibits an asymptotic pattern and approaches a limit as the measurement time becomes longer. The latter suggests that the accuracy of CG estimation is better after a longer time than θG estimation. Taken together, the numerical sensitivity analysis with different sensibility patterns for θG and CG put forward that both parameters can be estimated from a single experiment if it is ensured that the experimental time is greater than θG.
Figure 2

Numerical sensitivity analysis; ΔCC/ΔP, sensitivity parameter where P is either the parameter θG or CG; t θG−1, time expressed as a fraction of θG.

Figure 2

Numerical sensitivity analysis; ΔCC/ΔP, sensitivity parameter where P is either the parameter θG or CG; t θG−1, time expressed as a fraction of θG.

Close modal

Method deployment

The stability and operability of the chambers were tested in both reactor types, i.e., a fine-bubble reactor and a coarse-bubble reactor. The first chamber (OFCA) demonstrated stability and easy operation in the reactor with fine bubbles. That simpler and lighter chamber design, compared to OFCB, was easily operated by two operators from the perimeter path of the reactor, which were able to maintain the chamber in a fixed position and lift it above the water surface for 1 min, between measurements, for ventilation purposes. However, OFCA was not operable in the coarse-bubble reactor: flipping over, excessive tilting, and difficulty maintaining a fixed position. In that coarse-bubble reactor, the tubular chamber (OFCB) operated well. However, it required three operators, using two cords each (fixed at the top and bottom of the chamber), to ensure a fixed location and reasonable vertical orientation of the chamber. Due to the heavier weight of that chamber, OFCB could not be lifted above water between measurements for ventilation. Thus, the chamber was approximated to one operator that could lift it above water for 1 min, and then the chamber was placed back in position, which took 10–20 s. It is worth mentioning that the OFCB was also working well in the moderately agitated fine-bubble reactor, but in that reactor, the OFCA chamber was preferred, being lighter and easier to operate.

During deployment of the OFC method, measurements of QG, CG, and the resulting F were taken at several locations within the reactor, allowing the characterization of spatial variability of these parameters in the reactors, similar to what was previously done by Aboobakar et al. (2014). Some examples of CC patterns experimentally observed during triplicate measurements are shown in Figure 3. Overall, Equation (6) fitted well the experimental data for both and and therefore indicates that the OFC method is of potential interest for determining emissions from aerated reactors, i.e., including the air flow rate and its concentration in the target gases. Figure 3 also shows the distinctive calibrated asymptotic shape (Equation (6)) for the reactor equipped with fine-bubble aeration (Figure 3(a) and 3(b)) and the reactor with coarse bubbles (Figure 3(c) and 3(d)).
Figure 3

Examples of CC patterns experimentally observed during triplicate measurements of CH4 (a,c) and CO2 (b,d) and best Equation (6) fit in an aerated reactor with fine bubbles (a,b) and coarse bubbles (c,d).

Figure 3

Examples of CC patterns experimentally observed during triplicate measurements of CH4 (a,c) and CO2 (b,d) and best Equation (6) fit in an aerated reactor with fine bubbles (a,b) and coarse bubbles (c,d).

Close modal
A much higher variability in CC was observed in the case of the reactor with coarse bubbles, which can be associated with a more heterogeneous aeration and particularly stochastic aeration bursts reaching the surface, a distinctive behavior observed in that reactor. The latter was confirmed by a poorer goodness-of-fit of measurements in the coarse-bubble reactor, as shown in Table 1. Indeed, in the fine-bubble reactors, 46.7% of the measurements were characterized by nt > 1.2 (classified as good or excellent), while the same indicator was 25.0% in the coarse-bubble reactor (Figure 3(b) and 3(d)). Notably, during chamber deployment, the goodness-of-fit of was superior in the upstream region compared to the downstream region. The lower goodness-of-fit in the downstream region is likely attributed to lower resulting from the CH4 stripping (Figure 4(a) and the related description below). Essentially, a reduction in the final in each measurement results in a diminished magnitude of the transient state, which is the underlying principle of the method. In simpler terms, the method works better when emissions are higher. Conducting replicate measurements is advised for situations with low emissions to ensure the method's reliability and robustness.
Table 1

Range and mean nt values (goodness-of-fit) observed during chamber deployment, separated in all data, fine-bubble reactor only, and coarse-bubble reactor only; n = total number of nt measurements

CH4CO2
All data Range of nt 0.065–9.86 0.02–42.15 
 Mean nt 4.53 ± 2.19 (A) 7.3 ± 8.9 (B) 
 N 180 180 
 % of nt > 1.2 (Good/excellent) 36.2% 58.6% 
 % of 0.7 ≤ nt ≤ 1.2 (Acceptable) 24.1% 24.1% 
 % of nt< 0.7 (unacceptable) 39.7% 17.2% 
Fine bubbles Range of nt 0.25–9.86 0.9–42.15 
 Mean nt 3.5 ± 2.9 (A) 13.04 ± 9.5 (B) 
 N 90 90 
 % of nt > 1.2 (Good/excellent) 46.7% 83.2% 
 % of 0.7 ≤ nt ≤ 1.2 (Acceptable) 10.0% 10.1% 
 % of nt< 0.7 (unacceptable) 43.3% 6.7% 
Coarse bubbles Range of nt 0.05–4.57 0.02–6.26 
 Mean nt 1.46 ± 1.33 (A) 1.7 ± 1.45 (B) 
 N 90 90 
 % of nt > 1.2 (Good/excellent) 25.0% 32.2% 
 % of 0.7 ≤ nt ≤ 1.2 (Acceptable) 39.3% 39.2% 
 % of nt< 0.7 (unacceptable) 35.7% 28.6% 
CH4CO2
All data Range of nt 0.065–9.86 0.02–42.15 
 Mean nt 4.53 ± 2.19 (A) 7.3 ± 8.9 (B) 
 N 180 180 
 % of nt > 1.2 (Good/excellent) 36.2% 58.6% 
 % of 0.7 ≤ nt ≤ 1.2 (Acceptable) 24.1% 24.1% 
 % of nt< 0.7 (unacceptable) 39.7% 17.2% 
Fine bubbles Range of nt 0.25–9.86 0.9–42.15 
 Mean nt 3.5 ± 2.9 (A) 13.04 ± 9.5 (B) 
 N 90 90 
 % of nt > 1.2 (Good/excellent) 46.7% 83.2% 
 % of 0.7 ≤ nt ≤ 1.2 (Acceptable) 10.0% 10.1% 
 % of nt< 0.7 (unacceptable) 43.3% 6.7% 
Coarse bubbles Range of nt 0.05–4.57 0.02–6.26 
 Mean nt 1.46 ± 1.33 (A) 1.7 ± 1.45 (B) 
 N 90 90 
 % of nt > 1.2 (Good/excellent) 25.0% 32.2% 
 % of 0.7 ≤ nt ≤ 1.2 (Acceptable) 39.3% 39.2% 
 % of nt< 0.7 (unacceptable) 35.7% 28.6% 

Letters in parentheses indicate significant differences between parameters (p < 0.05), with the same letter signifying no significant difference.

Figure 4

CH4 concentration (a), air flow rate (b), and the resulting flux (c) of the gas escaping the coarse-bubble reactor and the fine-bubble reactor. For clarity, only error bars above each data point are shown (one standard deviation, triplicate measurements). Please note the logarithmic (a,c) and decimal (b) scales.

Figure 4

CH4 concentration (a), air flow rate (b), and the resulting flux (c) of the gas escaping the coarse-bubble reactor and the fine-bubble reactor. For clarity, only error bars above each data point are shown (one standard deviation, triplicate measurements). Please note the logarithmic (a,c) and decimal (b) scales.

Close modal

Regarding , the goodness-of-fit of Equation (6) (Figure 3(b) and 3(d)) was generally better, and nt was significantly higher for than for (Table 1). We also observed a generally better fit in the fine-bubble reactor than in the coarse-bubble reactor, but, on contrary to measurements, we did not observe a decrease in nt from the upstream region to the downstream region. We assume that the latter is caused by the lower observed decrease of , along the residence time, compared to (data not shown).

The main results of CH4 parameters are shown in Figure 4, revealing clear trends in both the coarse-bubble and the fine-bubble reactors. It should be recalled that both reactors were operating in plug-flow mode with a θH of approximately 5.98 ± 0.74 h, meaning that the observed trends represent the evolution of water exposition to continuous aeration. As a first result of the exposition to aeration, a rapid decrease in by a factor of approximately 14 in the coarse-bubble reactor and 104 in the fine-bubble reactor was observed. This decrease can be attributed to the stripping of CH4 presents the influent water, and it is worth noting that was close to atmospheric air in both reactors, for t θH−1 ≥ 0.55. This asymptotic trend is a clear indicator that CH4 emissions were most certainly caused by the stripping of the dissolved CH4 accumulated during the previous steps of the treatment train.

The features of the method presented here include measuring the air flow rate without using an actual flowmeter. The results in Figure 4(b) indicate that the OFC method is a convenient tool for characterizing and optimizing aeration in activated sludge reactors. The results indicate notable differences between the reactors with fine- and coarse-bubble aeration. The coarse-bubble reactor had a significantly higher air flow rate, at 2,561 ± 1,574 m3 h−1 (coefficient of variation; CV = 61%), more than 3.1 times higher than in the fine-bubble reactor, where it was 809 ± 432 m3 h−1 (CV = 53%). Moreover, the fine-bubble aeration system, which was recently installed in the WWTP, had a more homogeneous airflow, indicating optimization of the aeration system. These differences between coarse- and fine-bubble aeration systems are what was previously reported, i.e., a better air distribution at a lower energy cost (Amaral et al. 2017, 2019; Rosso et al. 2023).

Regarding CH4 emissions (Figure 4(c)), a noticeable decay was observed in the upstream region of both reactors. For instance, over the first third of the water residence time (t θH−1 < 0.3), in the coarse-bubble reactor, declined from 15,251 ± 5,980 to 5,492 ± 1,355 mg m−2 h−1, representing a 64% decrease. In the fine-bubble reactor, dropped more sharply, decreasing from 7,904 ± 3,455 to 166 ± 25 mg m−2 h−1, signifying a 98% decrease. According to the mass transfer theory, this discrepancy indicates that fine bubbles are more effective in stripping out the CH4 contained in the influent water.

Comparing the CH4 fluxes observed in this study to existing literature poses a complex challenge. This complexity arises not only from the prevalence of reporting plant-scaled emission factors in prior works, expressed as CH4 emitted per unit of wastewater treated or pollutant removed, but also due to the substantial variations in influent characteristics, treatment train configurations, and operational regimes across these earlier reports. Emissions reported by Bao et al. (2016) were in the range of 255 − 8,944 mg m−2 h−1, and those reported by Masuda et al. (2015) were in the range of 389 − 1,650 mg m−2 h−1, both in the oxic reactor of an anoxic–oxic train. Yan et al. (2014) and Ren et al. (2013) reported ranges of CH4 emissions of 500 − 1,500 and 590 − 1,048 mg m−2 h−1, respectively, both in the aerobic reactor of an anaerobic–anoxic–oxic (A2O) treatment train. Lastly, Aboobakar et al. (2014) reported a higher range, i.e., 58,823 − 70,588 mg m−2 h−1 in the aerated reactor of a conventional wastewater plant, similar to the reactors that were characterized in the present work. In this study, values were observed to range from 154 to 21,603 mg m−2 h−1 (mean of 4,995 mg m−2 h−1) in the coarse-bubble reactor and from 57 to 12,774 mg m−2 h−1 (mean of 865 mg m−2 h−1) in the fine-bubble reactor. The mean emissions observed in the present work are 1–2 magnitude orders lower than those reported by Aboobakar et al. (2014) and closer to those observed in more complex treatment trains reported here above. A potential reason for that is that the treatment plant selected to test our method has a relatively limited removal efficiency (82 ± 8%, in chemical oxygen demand units) compared to 95%, reported in Aboobakar et al. (2014).

About depicted in Figure 5, the trend along the spatial position within the reactor exhibited a different behavior compared to . While experienced a reduction spanning two orders of magnitude, exhibited a continuous decrease along the residence time in both reactors, specifically for t θH1 values greater than 0.2.
Figure 5

measured in the coarse-bubble reactor and the fine-bubble reactor. For clarity, only error bars above each data point are shown (one standard deviation, triplicate measurements). Please note the logarithmic scale.

Figure 5

measured in the coarse-bubble reactor and the fine-bubble reactor. For clarity, only error bars above each data point are shown (one standard deviation, triplicate measurements). Please note the logarithmic scale.

Close modal

This observation aligns with the understanding that, unlike , results from stripping but also from heterotrophic activity. These findings imply that the suggested OFC method could be a valuable tool for estimating metabolic activity within the reactor. Nevertheless, it is important to acknowledge that the relationship between metabolic activity and off-gas measurements is intricate. This complexity arises from the involvement of various factors, including stripping, heterotrophic respiration, and the physicochemical carbonate equilibrium, which has been the subject of previous studies (Flores-Alsina et al. 2011; Lizarralde et al. 2015, 2018; Le et al. 2018).

Strengths, weaknesses, and perspectives of the method

The combination of straightforwardly designed floating chambers and a gas analyzer allowed the segregated determination of the gas flow rate and the concentration of the target gas emitted to the atmosphere. As demonstrated, this method proved successful in determining GHG emissions from aerated reactors, whether operated with fine bubbles or coarse bubbles. This capability is crucial for establishing the GHG budget of these specific process units, making it a potent tool to enhance our current understanding of GHG emissions. Moreover, it supports a migration from Tier 1 to Tier 2 or Tier 3 methods by allowing to incorporate, in a simpler manner, unique characteristics of process units, sites, facilities, and regions. Our research underscores the method's accuracy in determining methane and carbon dioxide emissions from aerated reactors, making it potentially applicable to other gases emitted in wastewater treatment plants. Its cost-effectiveness, versatility, and simplified measurement process position it as a valuable tool for assessing and managing greenhouse gas emissions in wastewater treatment plants.

Moreover, the OFC method allowed us to determine the spatial distribution of the gas flow rate. It might be a profitable tool to optimize or verify the effectiveness of the aeration system. A key advantage of our method, compared to previously reported OFC methods, is eliminating the need for additional equipment beyond an OFC and a gas analyzer, i.e., flowmeters, tracer gases, pressure measurements, or gas conditioning units. Notably, the need for a gas conditioning unit is a characteristic inherent to the gas analyzer rather than a specificity of our method. Thus, it cannot be considered a potential benefit of our proposed approach.

Regarding cost, assuming no significant differences among the gas analyzers used, our approach would be more economical compared to more standard methods using flowmeters, with or without pressure transducers, or tracer gases. In terms of accuracy and precision, no notable differences should exist between our method and standard ones in terms of the concentration CG, as this is a variable that is analyzer-specific. However, notable differences might exist between measuring gas flow rates through a flowmeter or adjustment of the CG data, as in our method. However, elucidating differences in the accuracy of gas flow rate would require a focused experimental demonstration.

In this study, two distinct chamber designs were developed through iterative testing until proven effective in both reactor types. The chambers, detailed in the supplementary material, were self-made using readily available materials, and their design is easily modifiable if needed, according to specific requirements. For instance, a higher ratio would enhance the determination of flow rate by extending the transient period and providing a more precise determination of θG. However, it may pose challenges in terms of operational convenience due to a heavier weight. This aspect introduces a trade-off that operators must carefully consider based on their specific objectives. Importantly, it is noteworthy that the volume/area ratio does not impact the final concentration observed in the chamber.

Importantly, although the focus of this study centered on CH4 and CO2, it is crucial to note that the developed method is not confined to these gases. Pending experimental confirmation, by integrating an appropriate detector, the methodology could extend its applicability to measure any target gas. This expanded gas list encompasses nitrous oxide, a greenhouse gas that has garnered increasing attention in wastewater plants (Czepiel et al. 1995; Foley et al. 2010; Daelman et al. 2013; Mampaey et al. 2015; Hwang et al. 2016; de Haas & Andrews 2022). Moreover, the scope may encompass ammonia and hydrogen sulfide – compounds known for their odoriferous, toxic, and corrosive properties – or volatile organic compounds (Mihelcic et al. 1993; Tata et al. 2003; Kurwadkar et al. 2022), as long as real-time analyzers for these compounds are available. Finally, off-gas oxygen measurements have been suggested (Pasini et al. 2021). The consideration of oxygen is distinctive from other gases, as off-gas measurements provide insights into the fraction of oxygen unconsumed within the reactor rather than indicating a proper emission.

Meanwhile, there has been a significant increase in studies investigating polyfluoroalkyl substances (PFAs) originating from wastewater treatment plants and their persistence in the hydrological cycle. Although, to the best of our current knowledge, real-time monitors for volatile PFAs are not available. Nonetheless, adaptations to this method should be explored and experimentally validated for this particular purpose.

Furthermore, like previously described OFC methods, it is important to highlight that the method presented in this study may serve objectives beyond emission reporting. Among its various applications, it has the potential to yield valuable insights into the metabolic activity of the reactor. This is, for instance, the case of off-gas oxygen measurements, since the difference from the injected oxygen informs the operator about the oxygen uptake rate of the treatment plant, the oxygen uptake efficiency, and the point-specific respiration process, which have been the object of several previous works (Weissenbacher et al. 2007; Amaral et al. 2017; Lizarralde et al. 2018). Moreover, if the dissolved oxygen concentration is known, the oxygen mass balance transfer can be established offering a valuable tool for the experimental determination of the oxygen mass transfer coefficient (KLa), and the oxygen transfer capacity (Schwarz et al. 2021, 2022; Le et al. 2022). The same strategy can be used with other gases, gaining valuable insights on production/uptake processes.

Notably, the method presented here also exhibited a notable limitation. The method worked well for CO2 emissions in any reactor cell. However, its effectiveness in quantifying could have been more robust with CH4, mostly because was relatively low in the downstream region of the reactor. Under these conditions, the magnitude of the transient state, which is the underlying principle of the method, is reduced, and a more significant error in flux determination is therefore expected. As previously mentioned, in simpler terms, the method works better when emissions are higher. In this regard, it should be emphasized that if multiple gases are measured, as in the present work, determining the gas flow rate can be done with any measured gas, thus reducing the risk of not getting results.

In terms of the LOD for the simplified OFC method, employing a typical arbitrary limit set at three times the gas analyzer's LOD (i.e., 10 ppb for CH4), we can assume that the method is capable of detecting CH4 concentrations that are 30 ppb above the atmospheric level. This detectable concentration, when multiplied by the mean gas flow rates of 2,561 and 809 m3 h−1 observed in the coarse- and fine-bubble reactors, respectively, translates to a flux detection limit of 12.9 − 41.0 mg m−2 h−1. These values represent 0.3 − 1.5% of the mean observed in the present study. Consequently, we confidently conclude that the LOD of the present method would have a negligible impact on the accuracy of the total emission measurements. As for CO2, the emissions consistently exceeded any LOD and were, therefore, not a concern.

In conclusion, the proposed simplified OFC method for determining the air flow rate and composition was successfully applied in aerated reactors with fine and coarse bubbles. The method's strength lies in its uncomplicated and versatile design using straightforward floating chamber designs, eliminating the need for a gas flowmeter and additional peripheral equipment. Considering the cost, the simplified method ensures that equipment expenses include only the essential gas analyzer and the chambers made from readily available materials. The simplicity of operation is enhanced not only by measuring just one variable (gas concentration) but also by the connection of the chamber to the gas analyzer, which requires only small-diameter flexible tubing to transfer a low sampled air flow rate. This eliminates the need to transport the complete off-gas flow rate to the reactor's perimeter path, where the gas analyzer and flowmeter are usually located. A notable feature of this method is its ability to determine point-specific flow rates and their spatial distribution, which is crucial for optimizing and verifying aeration system effectiveness. Regarding gas emissions determination, the method performed well for CO2 under all conditions. In the case of CH4, some limitations were observed in sections with low emissions, which are of minimal significance as they pertain to very low or negligible emissions. This showcases the method's robustness in accurately measuring emissions when detectable. Theoretically, the method could be extended to encompass other gases such as ammonia, hydrogen sulfide, volatile organic compounds, and oxygen, among others, pending experimental confirmation. Additionally, if coupled with the determination of dissolved oxygen concentration, the method could serve as a valuable tool for determining oxygen uptake rate, efficiency, mass transfer coefficient, and transfer capacity. This same strategy can be applied to other gases, providing valuable insights into production and uptake processes. While these features are not inherent to the simplified OFC method presented here, they become more accessible through a more straightforward, cheaper, and more versatile approach.

We thank the ‘Consejo Nacional de Ciencia y Tecnología (Conacyt)’ for the financial support received (project A3-S-75824), as well as for the support to Pablo Morales-Rico, Jessica Ramos-Diaz, Estefani Mendoza-León, (#703565, #1008375 and #721344, respectively). The authors thank Mr David Flores Rojas, M.Sc. Victoria Teresita Velázquez Martínez and Mr Aaron Vazquez Martínez for their technical support. The authors acknowledge the field support received and the personal of the ‘San Juan de Aragón’ WWTP. The authors declare that they have no conflict of interest.

All relevant data are included in the paper or its Supplementary Information.

The authors declare there is no conflict.

Aboobakar
A.
,
Jones
M.
,
Vale
P.
,
Cartmell
E.
&
Dotro
G.
2014
Methane emissions from aerated zones in a full-scale nitrifying activated sludge treatment plant
.
Water, Air, & Soil Pollution
225
(
1
),
1814
.
doi:10.1007/s11270-013-1814-8
.
Amaral
A.
,
Gillot
S.
,
Garrido-Baserba
M.
,
Filali
A.
,
Karpinska
A. M.
,
Plósz
B. G.
,
De Groot
C.
,
Bellandi
G.
,
Nopens
I.
,
Takács
I.
,
Lizarralde
I.
,
Jimenez
J. A.
,
Fiat
J.
,
Rieger
L.
,
Arnell
M.
,
Andersen
M.
,
Jeppsson
U.
,
Rehman
U.
,
Fayolle
Y.
,
Amerlinck
Y.
&
Rosso
D.
2017
Towards advanced aeration modelling: From blower to bubbles to bulk
.
Water Science and Technology
75
(
3
),
507
517
.
doi:10.2166/wst.2016.365
.
Amaral
A.
,
Schraa
O.
,
Rieger
L.
,
Gillot
S.
,
Fayolle
Y.
,
Bellandi
G.
,
Amerlinck
Y.
,
Mortier
S. T. F. C.
,
Gori
R.
,
Neves
R.
&
Nopens
I.
2019
Modelling gas–liquid mass transfer in wastewater treatment: When current knowledge needs to encounter engineering practice and vice versa
.
Water Science and Technology
80
(
4
),
607
619
.
doi:10.2166/wst.2019.253
.
Bao
Z.
,
Sun
S.
&
Sun
D.
2016
Assessment of greenhouse gas emission from A/O and SBR wastewater treatment plants in Beijing, China
.
International Biodeterioration & Biodegradation
108
,
108
114
.
doi:10.1016/j.ibiod.2015.11.028
.
Chandran
K.
2011
Protocol for the measurement of nitrous oxide fluxes from biological wastewater treatment plants
. In:
Methods in Enzymology
.
Elsevier
, pp.
369
385
.
doi:10.1016/B978-0-12-381294-0.00016-X
.
Czepiel
P.
,
Crill
P.
&
Harriss
R.
1995
Nitrous oxide emissions from municipal wastewater treatment
.
Environmental Science & Technology
29
(
9
),
2352
2356
.
doi:10.1021/es00009a030
.
Daelman
M. R. J.
,
van Voorthuizen
E. M.
,
van Dongen
L. G. J. M.
,
Volcke
E. I. P.
&
van Loosdrecht
M. C. M.
2013
Methane and nitrous oxide emissions from municipal wastewater treatment – results from a long-term study
.
Water Science and Technology
67
(
10
),
2350
2355
.
doi:10.2166/wst.2013.109
.
de Haas
D.
&
Andrews
J.
2022
Nitrous oxide emissions from wastewater treatment – revisiting the IPCC 2019 refinement guidelines
.
Environmental Challenges
8
,
100557
.
doi:10.1016/j.envc.2022.100557
.
Flores-Alsina
X.
,
Corominas
L.
,
Snip
L.
&
Vanrolleghem
P. A.
2011
Including greenhouse gas emissions during benchmarking of wastewater treatment plant control strategies
.
Water Research
45
(
16
),
4700
4710
.
doi:10.1016/j.watres.2011.04.040
.
Foley
J.
,
de Haas
D.
,
Yuan
Z.
&
Lant
P.
2010
Nitrous oxide generation in full-scale biological nutrient removal wastewater treatment plants
.
Water Research
44
(
3
),
831
844
.
doi:10.1016/j.watres.2009.10.033
.
Hwang
K.-L.
,
Bang
C.-H.
&
Zoh
K.-D.
2016
Characteristics of methane and nitrous oxide emissions from the wastewater treatment plant
.
Bioresource Technology
214
,
881
884
.
doi:10.1016/j.biortech.2016.05.047
.
IPCC
2006
‘2006 IPCC guidelines for national greenhouse gas inventories’, 5. Available from: https://www.ipcc-nggip.iges.or.jp/public/2006gl/vol5.html.
IPCC
2019
‘2019 Refinement to the 2006 IPCC Guidelines for National Greenhouse Gas Inventories – IPCC’
. .
Kurwadkar
S.
,
Dane
J.
,
Kanel
S. R.
,
Nadagouda
M. N.
,
Cawdrey
R. W.
,
Ambade
B.
,
Struckhoff
G. C.
&
Wilkin
R.
2022
Per- and polyfluoroalkyl substances in water and wastewater: A critical review of their global occurrence and distribution
.
Science of the Total Environment
809
,
151003
.
doi:10.1016/j.scitotenv.2021.151003
.
Le
Q. H.
,
Verheijen
P. J. T.
,
van Loosdrecht
M. C. M.
&
Volcke
E. I. P.
2018
Experimental design for evaluating WWTP data by linear mass balances
.
Water Research
142
,
415
425
.
doi:10.1016/j.watres.2018.05.026
.
Le
Q. H.
,
Verheijen
P. J. T.
,
van Loosdrecht
M. C. M.
&
Volcke
E. I. P.
2022
Application of data reconciliation to a dynamically operated wastewater treatment process with off-gas measurements
.
Environmental Science: Water Research & Technology
.
10.1039.D2EW00006G. doi:10.1039/D2EW00006G
.
Lizarralde
I.
,
Fernández-Arévalo
T.
,
Beltrán
S.
,
Ayesa
E.
&
Grau
P.
2015
A new general methodology for incorporating physico-chemical transformations into multi-phase wastewater treatment process models
.
Water Research
74
,
239
256
.
doi:10.1016/j.watres.2015.01.031
.
Lizarralde
I.
,
Fernández-Arévalo
T.
,
Brouckaert
C.
,
Vanrolleghem
P.
,
Ikumi
D. S.
,
Ekama
G. A.
,
Ayesa
E.
&
Grau
P.
2018
Validation of a multi-phase plant-wide model for the description of the aeration process in a WWTP
.
Water Research
129
,
305
318
.
doi:10.1016/j.watres.2017.10.061
.
Mampaey
K. E.
,
van Dongen
U. G. J. M.
,
van Loosdrecht
M. C. M.
&
Volcke
E. I. P.
2015
Novel method for online monitoring of dissolved N2O concentrations through a gas stripping device
.
Environmental Technology
36
(
13
),
1680
1690
.
doi:10.1080/09593330.2015.1005029
.
Masuda
S.
,
Suzuki
S.
,
Sano
I.
,
Li
Y.-Y.
&
Nishimura
O.
2015
The seasonal variation of emission of greenhouse gases from a full-scale sewage treatment plant
.
Chemosphere
140
,
167
173
.
doi:10.1016/j.chemosphere.2014.09.042
.
Mihelcic
J. R.
,
Baillod
C. R.
,
Crittenden
J. C.
&
Rogers
T. N.
1993
Estimation of VOC emissions from wastewater facilities by volatilization and stripping
.
Air & Waste
43
(
1
),
97
105
.
doi:10.1080/1073161X.1993.10467120
.
Nguyen
T. K. L.
,
Ngo
H. H.
,
Guo
W.
,
Chang
S. W.
,
Nguyen
D. D.
,
Nghiem
L. D.
,
Liu
Y.
,
Ni
B.
&
Hai
F. I.
2019
Insight into greenhouse gases emissions from the two popular treatment technologies in municipal wastewater treatment processes
.
Science of the Total Environment
671
,
1302
1313
.
doi:10.1016/j.scitotenv.2019.03.386
.
Pan
Y.
,
van den Akker
B.
,
Ye
L.
,
Ni
B.-J.
,
Watts
S.
,
Reid
K.
&
Yuan
Z.
2016
Unravelling the spatial variation of nitrous oxide emissions from a step-feed plug-flow full scale wastewater treatment plant
.
Scientific Reports
6
(
1
),
20792
.
doi:10.1038/srep20792
.
Parravicini
V.
,
Filali
A.
,
Delre
A.
,
Gutierrez
O.
&
Duan
H.
2022
Full-scale quantification of N2O and CH4 emissions from urban water systems
. In:
Quantification and Modelling of Fugitive Greenhouse Gas Emissions From Urban Water Systems
(
Ye
L.
,
Porro
J.
&
Nopens
I.
, eds).
IWA Publishing
, pp.
91
131
.
doi:10.2166/9781789060461_0091
.
Pasini
F.
,
Garrido-Baserba
M.
,
Ahmed
A.
,
Nakhla
G.
,
Santoro
D.
&
Rosso
D.
2021
Oxygen transfer and plant-wide energy assessment of primary screening in WRRFs
.
Water Environment Research
93
(
5
),
677
692
.
doi:10.1002/wer.1349
.
Pavelka
M.
,
Acosta
M.
,
Kiese
R.
,
Altimir
N.
,
Brümmer
C.
,
Crill
P.
,
Darenova
E.
,
Fuß
R.
,
Gielen
B.
,
Graf
A.
,
Klemedtsson
L.
,
Lohila
A.
,
Longdoz
B.
,
Lindroth
A.
,
Nilsson
M.
,
Marañon-Jimenez
S.
,
Merbold
L.
,
Montagnani
L.
,
Peichl
M.
,
Pihlatie
M.
,
Pumpanen
J.
,
Serrano Ortiz
P.
,
Silvennoinen
H.
,
Skiba
U.
,
Vestin
P.
,
Weslien
P.
,
Janouš
D.
&
Kutsch
W.
2018
Standardisation of chamber technique for CO2, N2O and CH4 fluxes measurements from terrestrial ecosystems
.
International Agrophysics
32
(
4
),
569
587
.
doi:10.1515/intag-2017-0045
.
Ren
Y. G.
,
Wang
J. H.
,
Li
H. F.
,
Zhang
J.
,
Qi
P. Y.
&
Hu
Z.
2013
Nitrous oxide and methane emissions from different treatment processes in full-scale municipal wastewater treatment plants
.
Environmental Technology
34
(
21
),
2917
2927
.
doi:10.1080/09593330.2012.696717
.
Ribera-Guardia
A.
,
Bosch
L.
,
Corominas
L.
&
Pijuan
M.
2019
Nitrous oxide and methane emissions from a plug-flow full-scale bioreactor and assessment of its carbon footprint
.
Journal of Cleaner Production
212
,
162
172
.
doi:10.1016/j.jclepro.2018.11.286
.
Ritter
A.
&
Muñoz-Carpena
R.
2013
Performance evaluation of hydrological models: Statistical significance for reducing subjectivity in goodness-of-fit assessments
.
Journal of Hydrology
480
,
33
45
.
doi:10.1016/j.jhydrol.2012.12.004
.
Rodriguez-Caballero
A.
,
Aymerich
I.
,
Poch
M.
&
Pijuan
M.
2014
Evaluation of process conditions triggering emissions of green-house gases from a biological wastewater treatment system
.
Science of the Total Environment
493
,
384
391
.
doi:10.1016/j.scitotenv.2014.06.015
.
Rosso
D.
,
Stenstrom
M. K.
&
Garrido-Baserba
M.
2023
Aeration and mixing
’, in G. Chen, M. C. M. van Loosdrecht, G. A. Ekama, and D. Brdjanovic (eds) Biological Wastewater Treatment: Principles, Modelling and Design.
IWA Publishing
, pp.
419
474
.
Available at: https://doi.org/10.2166/9781789060362_0419.
Schwarz
M.
,
Behnisch
J.
,
Trippel
J.
,
Engelhart
M.
&
Wagner
M.
2021
Oxygen transfer in two-stage activated sludge wastewater treatment plants
.
Water
13
(
14
),
1964
.
doi:10.3390/w13141964
.
Schwarz
M.
,
Trippel
J.
,
Engelhart
M.
&
Wagner
M.
2022
Determination of alpha factors for monitoring of aeration systems with the ex situ off-gas method: Experience from practical application and estimation of measurement uncertainty
.
Environmental Science and Pollution Research
29
(
58
),
87950
87968
.
doi:10.1007/s11356-022-21915-2
.
Tata
P.
,
Witherspoon
J.
&
Lue-Hing
C.
2003
VOC Emissions From Wastewater Treatment Plants: Characterization, Control and Compliance
.
Boca Raton
,
CRC Press
.
Tauber
J.
,
Parravicini
V.
,
Svardal
K.
&
Krampe
J.
2019
Quantifying methane emissions from anaerobic digesters
.
Water Science and Technology
80
(
9
),
1654
1661
.
doi:10.2166/wst.2019.415
.
van Loosdrecht
M. C. M.
,
Nielsen
P. H.
,
Lopez-Vazquez
C. M.
&
Brdjanovic
D.
2016
Experimental Methods in Wastewater Treatment
.
IWA Publishing
.
Weissenbacher
N.
,
Lenz
K.
,
Mahnik
S.
,
Wett
B.
&
Fuerhacker
M.
2007
Determination of activated sludge biological activity using model corrected CO2 off-gas data
.
Water Research
41
(
7
),
1587
1595
.
doi:10.1016/j.watres.2006.12.037
.
Yan
X.
,
Li
L.
&
Liu
J.
2014
Characteristics of greenhouse gas emission in three full-scale wastewater treatment processes
.
Journal of Environmental Sciences
26
(
2
),
256
263
.
doi:10.1016/S1001-0742(13)60429-5
.
Ye
L.
,
Porro
J.
&
Nopens
I.
2022
Quantification and Modelling of Fugitive Greenhouse Gas Emissions From Urban Water Systems
.
IWA Publishing
.
doi:10.2166/9781789060461
.
Zhan
X.
,
Hu
Z.
&
Wu
G.
2017
Greenhouse gas emission and mitigation in municipal wastewater treatment plants
.
Water Intelligence Online
16
,
9781780406312
.
doi:10.2166/9781780406312
.
This is an Open Access article distributed under the terms of the Creative Commons Attribution Licence (CC BY-NC-ND 4.0), which permits copying and redistribution for non-commercial purposes with no derivatives, provided the original work is properly cited (http://creativecommons.org/licenses/by-nc-nd/4.0/).

Supplementary data