PERMEABLE REACTIVE INTERCEPTOR

Since a denitrifying bioreactor is only designed to remediate a single contaminant, the current authors propose the term ‘permeable reactive interceptor’ (PRI) where the aim is to achieve agricultural sustainability. A PRI is a modified denitrifying bioreactor with additional remediation cells for specific contaminants in the form of solutes, particles or gases, and the capacity to deal with high temporal variation in contaminant fluxes.

During the operational phase of a PRI, practitioners will aim to achieve what the current authors term ‘a mixed-contaminant blockade’. This is the ability of a PRI to achieve complete remediation of mixed-contaminant solutions while blocking the leaks in the ‘hole in the pipe model’ of Firestone & Davidson (Reference Firestone, Davidson, Andreae and Schimel1989) (i.e. preventing pollution swapping). This may be possible in theory but may never happen in reality. Another more realistic term would therefore be a ‘mixed-contaminant blockade balance’ where both the mixed-contaminant remediation and pollution swapping occur but at acceptable rates defined by specific environmental legislation to the area the bioreactor is installed in. This balance is dynamic and will change over the lifetime of the PRI as legislative targets change. In addition, temporal changes to the characteristics of the PRI due to functional dynamics of and interactions within the microbial community, changes to organic C bioavailability, saturation of reactive media within a PRI or changes in contaminant inputs to the bioreactor may be required. Therefore, the PRI design will need to be flexible during its operational lifetime.

Contaminant blockade or contaminant blockade balance?

A simple methodology is used to determine if the mixed-contaminant blockade or contaminant blockade balance, as defined above, is achieved through a holistic analysis of the PRI in its development and operational phases. Once the state is known, intervention can occur to improve the system. The following procedure may be undertaken for any dissolved contaminants (e.g. NH₄ ⁺, NO₃ ⁻, dissolved reactive P and heavy metals), greenhouse gases (GHGs;CO₂, N₂O and CH₄) and NH₃ gas. Consider, for example, N₂O:

(1) $$F_{{\rm N}_{\rm 2} {\rm O(IN)}} - (F_{{\rm N}_{\rm 2} {\rm O(OUT)}} + F_{{\rm N}_{\rm 2} {\rm O(ATM)}} ) = B_{{\rm N}_{\rm 2} {\rm O}} $$

where $F_{{\rm N}_{\rm 2} {\rm O(IN)}} $ and $F_{{\rm N}_{\rm 2} {\rm O(OUT)}} $ are the dissolved flux (F) of N₂O at the inlet (IN) and outlet (OUT) of a PRI, respectively. The term $F_{{\rm N}_{\rm 2} {\rm O(ATM)}} $ is the N₂O lost from the PRI to the atmosphere (ATM) (Fig. 1), and $B_{{\rm N}_{\rm 2} {\rm O}} $ is the balance of N₂O between these three fluxes. If $B_{{\rm N}_{\rm 2} {\rm O}} $ >0, remediation within the PRI has occurred; if $B_{{\rm N}_{\rm 2} {\rm O}} = 0$ , the compound is conserved during transport through the PRI (or remediation and production of the compound are equal), and if $B_{{\rm N}_{\rm 2} {\rm O}} $ <0, then pollution swapping has occurred. Even if such calculations are made by considering the PRI in isolation, they could be extended to any connected water body, by accounting for pre- and post-PRI contaminant fluxes in such a water body.

Balances calculated using Eqn (1) can be used to derive a sustainability index (x) and a weighting system can be applied, whereby the weighting of trade-offs between different loss pathways can be judged (Eqn 2). The advantage of this is that different regions can apply different weightings depending on the environmental policy emphasis. An example here is where aeration is a problem in a system, resulting in a trade-off between CH₄ and N₂O:

(2) $$\eqalign {x = \tab a(B_{{\rm N_{2}O}} ) + b\left( {B_{{\rm NO}_{\rm 3}^{\rm -}}} \right) + c\left( {B_{{\rm CH}_{\rm 4}}} \right) \cr \tab {+ d\left( {B_{{\rm CO}_{\rm 2}}} \right) + {\cdots}} $$

where a–d and so on are weighting factors and B terms here are gathered from Eqn (1). Other contaminants in gaseous (e.g. NH₃ and H₂S), dissolved (e.g. NH₄ ⁺, metals) and particulate (e.g. particulate P) forms may also be added.

Working through Eqns (1) and (2) for all contaminants creates an overall balance, identifies contaminants of concern and remediation sequences, which can be implemented to form a PRI (See Fig. 1 for a N₂O example). This process should be repeated at different intervals during the development and operational phases of the PRI.

To illustrate the use of Eqn (1), a dataset summarizing the balances of selected GHGs and dissolved contaminants for a series of laboratory scale denitrifying bioreactors is presented in Table 1 after Healy et al. (Reference Healy, Ibrahim, Lanigan, Serrenho and Fenton2012). In this case study, NO₃ ⁻ reduction and pollution swapping were tested for four different C substrates (lodgepole pine woodchips (LPW), cardboard, lodgepole pine needles (LPN) and BBS) mixed with soil (C source-to-soil volume ratio of 1). Nitrate spiked water (19·5–32·5 mg/l as NO₃-N) was circulated into the columns at a hydraulic loading rate of 30 mm/day for up to 460 days. This low hydraulic loading rate, combined with a source of OC in excess, was aimed to represent shallow groundwater conditions and promote heterotrophic activity (see Healy et al. (Reference Healy, Ibrahim, Lanigan, Serrenho and Fenton2012) for further details). Ammonia concentrations were negligible in the study and were therefore omitted from the balance for this case study. For example, NH₃ losses associated with dirty water and lagoon systems ranged from 0 to 0·6 g/m²/day NH₃-N (Hill et al. Reference Hill, Smith, Russell, Misselbrook and Brookman2008).

Table 1.

Balances (in g/m/day) of major dissolved compounds and greenhouse gases from laboratory scale denitrifying bioreactors (see Healy et al. (Reference Healy, Ibrahim, Lanigan, Serrenho and Fenton2012)) as calculated using Eqn (1). Four carbon sources are used: lodgepole pine woodchips (LPW), cardboard, lodgepole pine needles (LPN) and barley straw (BBS). Negative and positive balances indicate remediation and production of the compound, respectively

In order to implement adequate weighting factors as in Eqn (2) for such a system, one has to identify prominent environmental legislation or issues at stake in the geographic area where the PRI is installed. Current concentration thresholds or maximum admissible GHG emissions can be used to attribute weighting factors. For instance, in Ireland, the maximum admissible concentrations (MAC) for molybdate reactive phosphorus (MRP=PO₄ ³⁻−P in the present case study) and NH₄ ⁺ in rivers are set at 35 and 65 μg/l, respectively, while NO₃ ⁻ in groundwater should not exceed 8·5 mg/l (the current threshold, whereas 11·3 mg/l as NO₃-N is the MAC). Hence, the weighting factor for PO₄-P should be 1·86 times higher than for NH₄-N (65/35=1·86). Similarly, the weighting factor for PO₄-P should be 242·03 times higher than for NO₃-N (8471/35=242·03). Similarly, global warming potential (GWP) is a measure of radiative forcing attributable to an individual GHG relative to that of CO₂, which has a GWP of 1. This variable, which is 25 and 296 on a 100-year basis for CH₄ and N₂O, respectively (IPCC Reference Eggleston, Buendia, Miwa, Ngara and Tanabe2006), could be used in addition to others to determine the weighting factors. This exercise is illustrated using the three cases given below.

Case 1

Case 1 (e.g. USA, intensive feedlot pastoral, pig and poultry systems, legislative instruments are focused on water quality): NO₃ ⁻ removal is the most important environmental concern, while GHGs emissions to the atmosphere and other contaminant losses to water are perceived as secondary. In this case, NO₃-N balances are attributed significantly highest weighting factors in Eqn (2) than other contaminants. An example of this is given in Table 2 (Case 1), where the weighting factor for NO₃-N is set to 1, and all other weighting factors to <1. The results of Eqn (2) are then expressed in g/m²/day as NO₃-N. In this case, the ranking of the different bioreactors is as follow: 1-LPW; 2-LPN; 3-BBS; 4-Cardboard.

Table 2.

Ranking of the four carbon sources. Case 1 is expressed in g/NO₃-N/m/day, Case 2 in g/PO₄-P-eq/m/day and Case 3 in g/CO₂-eq/m/day. Smaller values within the same column indicate lower environmental impact. (X) indicates ranking for individual cases. For Case 1 all weighting factors are set to zero except for NO₃-N, for Case 2 all weighting factors for greenhouse gases are set to zero and for Case 3 all weighting factors for dissolved contaminants are set to zero

Case 2

Case 2 (e.g. EU – Ireland, extensive pasture-based livestock production, legislative instruments are focused on water quality (EU Water Framework Directive 2000/60/EC (CEC 2000) and Nitrates Directive, Directive 91/676/EEC (CEC 1991), GHG (EU 2020 Effort Sharing Decision, Decision No. 406/2009/EC (CEC 2009)) and transboundary pollutants (National Emissions Ceilings Directive, Directive 2001/81/EC (CEC 2001)): contaminant losses to water are perceived as a more important issue than GHG and transboundary emissions to the atmosphere. When only accounting for NO₃ ⁻-N, NH₄ ⁺-N and PO₄ ³⁻-P, the weighting factors are set to 1, 0·538 (35/65) and 0·004 (35/8471), respectively. The results of Eqn (2) are then expressed in g PO₄ ³⁻-P-eq/m²/day. In this case, the ranking of the different bioreactors is: 1-BBS; 2-LPN; 3-Cardboard; 4-LPW.

When accounting only for GHG and NH₃ emissions, one must refer to national legislation targets. In Ireland, the national target of GHG reduction is set to 20% in 2020, but there is no legislative limit set on individual farmers. In addition, while transboundary gases are limited to 100 ktonnes NH₃, national emissions are 10% under this ceiling. Hence, the potential weighting factors for GHG and NH₃ are 0·2 and 0·1, respectively.

Research should now focus on ways to amalgamate weighting factors for dissolved and particulate contaminants, and GHG and NH₃ emissions.

Case 3

Case 3: For example, in New Zealand (intensive pastoral system, legislative instruments are focused on GHG without NH₃) gaseous emissions to the atmosphere are perceived as a more important issue than contaminant losses to water. Currently, New Zealand is committed to reducing emissions by 5% by 2020 (10–20% in the case of a global agreement). Under the Climate Change Regulations (Agricultural Sector) 2010 (New Zealand Government 2010), agricultural emission targets are set for processors and live exporters with a subsequent ‘trickle down’, which is an impetus on farmers, particularly in terms of reducing farm N surpluses. In this case, CH₄, CO₂ and N₂O balances are attributed significantly higher weighting factors than for N deposition or dissolved and particulate contaminants. An example of this is given in Table 2 (Case 3), where the weighting factor for CO₂, CH₄ and N₂O are set at 1, 25 and 296, respectively, while all dissolved contaminants are given a weighting factor of <1. The results of Eqn (2) are then expressed in g CO₂-eq/m²/day. In this case, the ranking of the different bioreactors is: 1-LPW; 2-LPN; 3-BBS; 4-Cardboard. One key point of the New Zealand approach is that agricultural emissions are quantified by ‘emissions intensity’ or emissions per unit product basis. Therefore in this case, the emissions generated from the PRI should be expressed in terms of the production activity (yearly milk or beef output, for example) from the farm.

Based on these three case studies, an overall ranking can be determined, as presented in Table 2. Attributing weighting factors to both dissolved contaminants and GHGs in the same case study is more complex and would require assessing their respective environmental costs and benefits. Stevens & Quinton (Reference Stevens and Quinton2009) propose economics through cost-effectiveness analysis as a tool which could potentially enable the inclusion of pollution swapping in policy. Schipper et al. (Reference Schipper, Robertson, Gold, Jaynes and Cameron2010) have already shown using a preliminary cost–benefit analysis that such systems are synergistic to other management practices that try to minimize N losses, but this needs to be assessed at PRI implementation level. Multi-criteria analysis may offer a solution and incorporates any number of variables including cost effectiveness to assess a system in a holistic way, thereby taking account of pollution swapping (Balana et al. Reference Balana, Vinten and Slee2011).

In the first instance, pressure or weighting will be towards water quality as agriculture is currently outside of the GHG emissions trading scheme (ETS). As a result, there is no mechanism for individual farmers to gain or lose under such schemes. While this may change in the future, medium-term pressure for farmers to reduce emissions will be from producers and retailers focusing on marketing low C footprint agri-food products. Reducing the C footprint of agricultural produce is currently viewed as a principle (and easily measurable) metric of ‘sustainable’ production, and is viewed as a key factor in distinguishing quality produce and exploiting emerging markets by retailers and processors. This, rather than official trading schemes, is likely to put pressure on farmers to reduce GHGs. The efficacy of incorporation into trading schemes will be dependent on: (1) the point of obligation (farmer or processor) and (2) the price of C. This is exemplified in New Zealand where agriculture has been brought into the national ETS. Here the point of obligation is the processor, which in turn puts pressure on the farmer to reduce emissions.

Permeable reactive interceptor performance will be affected by inherent variability of contaminant fluxes and base flow/storm flow conditions between sites and PRI type (groundwater v. drainage). Using a natural drainage spring system as an example, an adaptable system will need to be designed in both cases to cope with variations in contaminant fluxes (e.g. base flow contamination from point source pollution or storm flow contamination from diffuse source pollution). In Ireland, large areas of intensive grassland is underlain by karst limestone, and spring discharges tend to range from a few litres per second for most to >10 litres/s in 200 cases and >25 litres/s for 50 cases with catchments of several km² (Drew Reference Drew2011). Using an average denitrification rate of 7·6 g N/m³/day calculated by Warnecke et al. (Reference Warnecke, Schipper, Bruesewitz, Mcdonald and Cameron2011) in a denitrification bed with a flow rate of 1·7 litres/s (equivalent to 145 m³/day), which is similar to that of a likely karst spring, the volume and dimensions of a PRI primarily targeting NO₃ ⁻ remediation can be estimated for different input and output NO₃ ⁻ concentrations. For example, to attenuate an input concentration of double the present EU-MAC (22·6 mg/l NO₃ ⁻-N) to the actual MAC, a bioreactor would need a volume of 218·4 m³, while to achieve total removal of N the volume increases further to 436·8 m³.

Several aspects here are interesting: (1) the denitrification rate needs to be estimated preferably at a field site, ensuring that real conditions prevail (e.g. temperature), using the selected organic C media or alternatively in control rooms under laboratory conditions (Healy et al. Reference Healy, Ibrahim, Lanigan, Serrenho and Fenton2012); (2) where land availability is scarce, the size of the PRI may be a limiting factor and may control the output concentration of NO₃ ⁻; and (3) an adaptive system might be needed to divert a proportion of the inlet flow to additional remediation units during times of excessive flow rates, e.g. storms.

THE WAY FORWARD: USING MOLECULAR ECOLOGY AS A PREDICTOR OF POLLUTION SWAPPING IN A PERMEABLE REACTIVE INTERCEPTOR

For the purposes of this special issue, the present paper focuses on the impact of molecular ecology on N cycling. Equivalent technologies should be explored for other elements. It will be expensive to assess a PRI in a holistic way. In the future, useful in situ tools such as biosensors should be developed for on-going monitoring and improvement of PRI performances. Microorganisms drive the majority of biogeochemical processes in natural and engineered systems, and have a major role in the cycling of N. As such, an understanding of the prevalence, diversity and activity of functionally important microbial groups can offer critical insights into important biogeochemical cycles in systems such as PRIs, and offer tools for optimization of system functioning. While numerous approaches are currently available to measure transformation processes, many suffer from limitations making their use undesirable or impractical (Groffman et al. Reference Groffman, Altabet, Böhlke, Butterbach-Bahl, David, Firestone, Giblin, Kana, Nielsen and Voytek2006). A particular methodological challenge in the context of the pollution swapping is accurate measurement of N₂ gas fluxes against a background of high N₂ concentrations in the environment (Groffman et al. Reference Groffman, Altabet, Böhlke, Butterbach-Bahl, David, Firestone, Giblin, Kana, Nielsen and Voytek2006), and the resulting calculation of N₂O : N₂ ratios, a critical component in the assessment of the environmental sustainability of PRI systems. Molecular microbial ecology approaches that focus on the populations, genes, transcripts and enzymes driving these processes offer new opportunities to overcoming these challenges (Philippot et al. Reference Philippot, Hallin and Schloter2007).

The capability for denitrification is widely dispersed across environments and microbial phylogenetic groups, with greater than 60 genera known thus far to contain denitrifiers (Throbäck et al. Reference Throbäck, Enwall, Jarvis and Hallin2004; Philippot et al. Reference Philippot, Hallin and Schloter2007). These include representatives from bacterial, fungal and archaeal groups (Groffman et al. Reference Groffman, Altabet, Böhlke, Butterbach-Bahl, David, Firestone, Giblin, Kana, Nielsen and Voytek2006). As such, commonly used molecular taxonomic markers, such as 16S rRNA, are of limited use for identifying denitrifiers. Molecular techniques since the late 1990s have instead focused on the use of functional genes and enzymes involved in N transformations within the denitrification pathway (Groffman et al. Reference Groffman, Altabet, Böhlke, Butterbach-Bahl, David, Firestone, Giblin, Kana, Nielsen and Voytek2006; Philippot et al. Reference Philippot, Hallin and Schloter2007). Seven enzymes have since been identified which catalyse the four major transformations within the pathway (Fig. 2) (Philippot & Hallin Reference Philippot and Hallin2006; Philippot et al. Reference Philippot, Hallin and Schloter2007). The initial transformation of NO₃ ⁻ to nitrite is catalysed by NO₃ ⁻ reductase. This can be of two types: membrane-bound (Nar) or periplasmic (Nap), and denitrifers can have either or both forms (Philippot & Hallin Reference Philippot and Hallin2006). Dissimilatory NO₂-reduction is catalysed by enzymes encoded by two genes, nirS and nirK. These genes are functionally and physiologically similar, yet structurally different (Throbäck et al. Reference Throbäck, Enwall, Jarvis and Hallin2004). In most cases, only one of the two types of nitrite reductase is observed within individual denitrifers, but cases of isolates possessing both have also been reported (Philippot & Hallin Reference Philippot and Hallin2006). Reduction of NO is also performed by two types of enzymes, cNor and qNor. Nitrous oxide reductase performs the final step in denitrification, the reduction of N₂O to N₂, and the catalytic subunit nosZ is used as a molecular marker for this process in Gram-negative bacteria (Philippot et al. Reference Philippot, Hallin and Schloter2007). Other functional gene targets have been identified within other N-cycling pathways, such as within N fixation and nitrification. The quantification of these functional genes enables potential rates of different N transformations to be estimated and the effect of media/environmental changes on potential rates to be deduced. In particular for denitrification, the quantification of nosZ and nirK/nirS within PRIs enables potential rates of complete and incomplete denitrification, respectively, to be estimated. Functional gene targets for other processes that may be important with respect to pollution swapping within these systems, such as methanogenesis (mcrA) and sulphate-reduction (dsrAB), are also available (Nercessian et al. Reference Nercessian, Bienvenu, Moreira, Prieur and Jeanthon2005). While microbial ecology tools are now being employed with increasing frequency to better understand processes underlying system functionality in a large array of other engineered systems, such as water infiltration systems, their use in PRIs is rare. Recent investigations of denitrifiers in such systems have used culture-based methods (Moorman et al. Reference Moorman, Parkin, Kaspar and Jaynes2010), which may underestimate numbers due to difficulty in culturing denitrifiers (Wallenstein et al. Reference Wallenstein, Myrold, Firestone and Voytek2006; Philippot et al. Reference Philippot, Hallin and Schloter2007) or non-quantitative molecular approaches (Elliot Reference Elliot2009). A notable exception was a study carried out by Warnecke et al. (Reference Warnecke, Schipper, Bruesewitz, Mcdonald and Cameron2011), who investigated the microbial ecology of laboratory denitrification beds filled with different C media incubated at different temperatures and receiving variable NO₃ ⁻ inputs. NirS, nirK and nosZ genes were quantitatively investigated over time and compared with measured rates of N transformations. Results showed that the abundance of denitrifying genes was similar in replicate barrels under cold incubation and this changed under warmer incubation temperatures. Warm incubation enhanced growth of nirS-containing bacteria and bacteria that lacked the nosZ gene. This may have implications for studies undertaken in a warmer laboratory or natural environments, which generally exhibit higher N₂O emissions. Feng et al. (Reference Feng, Xu, Xu, Zhu, Xu, Ding and Luan2012) also recently used molecular fingerprinting methods to deduce the relative dominance of functional microbial populations in a denitrifying bioreactor, according to variations in bioreactor operation and environmental conditions during treatment.

Fig. 2.

Selected N biotransformation pathways and common gene targets used in molecular ecology.

While in general terms denitrification will be the most important process in a denitrification bioreactor or PRI, other lesser known processes such as DNRA, nitrification and anammox may play a role in certain sections of the bioreactor where conditions for denitrification cease to be ideal. For example, if NO₃ ⁻ rates decrease at the end of the PRI pathway, other processes such as DNRA may become more important. This may offer opportunities with regard to pollution swapping, for example, when end-of-system NH₄ ⁺ rates are considered unacceptably high. Environmental conditions within a PRI system change both temporally and spatially, with a resulting impact on the distribution of denitrifying microbial communities and their metabolic response to a given set of conditions (Groffman et al. Reference Groffman, Altabet, Böhlke, Butterbach-Bahl, David, Firestone, Giblin, Kana, Nielsen and Voytek2006; Wallenstein et al. Reference Wallenstein, Myrold, Firestone and Voytek2006; Philippot et al. Reference Philippot, Andert, Jones, Bru and Hallin2011). For example, not all denitrifiers can catalyse all steps in the denitrification process, and a truncated pathway is often observed (Philippot & Hallin Reference Philippot and Hallin2006). Most commonly, denitrifiers lack the nosZ enzyme, thereby limiting the genetic capacity to produce N₂ gas and necessarily producing the GHG N₂O as an end product (Braker et al. Reference Braker, Schwarz and Conrad2010; Philippot et al. Reference Philippot, Andert, Jones, Bru and Hallin2011). Thus, the response to environmental regulators of compositionally different denitrifier communities varies and has been shown to result in different N₂O emission rates (Braker et al. Reference Braker, Schwarz and Conrad2010). The genetic composition of the denitrifier population is therefore of importance, when considering the potential of PRIs to reach the contaminant blockade. However, it must be noted that the genetic composition alone of a community is insufficient to predict the transformation rates, as bacterial denitrification is generally a facultative ability. Thus, use of this respiratory pathway is not essential for the survival of members of these communities (Groffman et al. Reference Groffman, Altabet, Böhlke, Butterbach-Bahl, David, Firestone, Giblin, Kana, Nielsen and Voytek2006). Consequently, the targeted genes may not necessarily be expressed under the environmental conditions studied (Philippot Reference Philippot2005). However, molecular characterization of denitrifiers, together with quantification of environmental controllers and denitrification rates, potentially offers a unique opportunity to evaluate and manipulate both abiotic and biotic controls driving observed N transformations (Groffman et al. Reference Groffman, Altabet, Böhlke, Butterbach-Bahl, David, Firestone, Giblin, Kana, Nielsen and Voytek2006; Stark & Richards Reference Stark and Richards2008). In addition to DNA-based insights into the diversity and genetic potential of a community for denitrification, currently evolving methodologies, based either on mRNA or proteins, aim to target active microorganisms to understand the relationship between community diversity and system function (Philippot & Hallin Reference Philippot and Hallin2006; Wallenstein et al. Reference Wallenstein, Myrold, Firestone and Voytek2006).