3.1. Pre-Treatment Tests
The pH values were measured in samples withdrawn from both columns at three sampling ports (5, 28, and 50 cm) on days 3 and 7 (see Supplementary Materials
, Table S3
). After the first week, pH stabilized at the influent value of about 8 for the whole duration of the tests.
Ammonium, chloride, and COD concentrations were measured in MLL samples withdrawn at the outlet of the columns. The results, which are the average of two replicates, are summarized in Table 4
for the column tests carried out with the ZVI/lapillus and ZVI/GAC granular reactive media, respectively.
In general, the removal of ammonium and chloride was basically negligible for both granular media. Ammonium content decreased in the first days of operation in both pre-treated MLLs, with the removal efficiency for the ZVI/GAC granular medium being better than that for the ZVI/lapillus mixture (83% and 57%, respectively). From day 8 onward, the amount of ammonium in the MLLs was close to the influent concentration. GAC was not expected to help with ammonium removal [15
] due to its nonpolar surface, which results in poor interactions with polar adsorbates [45
]. Nevertheless, it was noticeable that ZVI mixed with GAC showed a better performance compared to the ZVI/lapillus mixture (RE
= 57%), even if only for a short period of time. Chloride was also basically unaffected by pre-treatment.
COD concentration measured after 2 days witnessed a decrease in both tests (RE
of 59% and 65% for ZVI/lapillus and ZVI/GAC column, respectively), whereas during the tests, it was slightly removed only by the ZVI/GAC granular mixture due to the key role played by the presence of GAC. Indeed, GAC is known as an effective adsorbent for organic matter [18
], and its application is common in leachate treatment research [16
]. In general, as [49
] asserted, the application of activated carbon adsorption is effective for the removal of organic compounds from MLL but not for ammonia nitrogen.
On the contrary, lapillus was found not to be a suitable material for ammonium, chloride, and COD removal. This proves that lapillus has no interactions with these tested contaminants, although they were not the specific targets of the pre-treatment.
Metal removal, at the outlet of the columns, is shown in Figure 1
. After 2 days of operation, metal removal in the ZVI/lapillus filter was almost complete with an efficiency close to 100% for the three metals. Subsequently, the removal efficiency toward Cu was slightly reduced assuming values between 84% and 93% until the end of the test. The reductions in Ni and Zn were more noticeable; for the former, it was in the range 55–67%, while for the latter, it gradually decreased from 61% on day 8 to 36% on day 38.
Referring to the column filled with the ZVI/GAC mixture, Cu was completely removed at the beginning of the test (RE of about 100%). Its removal efficiency was reduced only in the last days of the test, assuming the lowest value of 75% on day 38. Ni removal was scarce until day 2 (RE = 27%) but significant on day 4 (RE = 93%). From this point, the removal efficiency toward Ni progressively decreased to 57% on day 38. Lastly, Zn removal efficiency increased to almost 100% until day 4, before gradually decreasing from 84% on day 8 to 66% on day 38.
In general, the results showed that both granular filters were efficient in removing the three contaminants according to the following sequences for ZVI/lapillus and ZVI/GAC granular mixtures, respectively: Cu > Ni > Zn and Cu > Zn > Ni.
ZVI was expected to remove HMs through a mechanism of reduction, adsorption onto its corrosion products, and precipitation and coprecipitation with its precipitating oxides [50
]. Cu was removed via the well-described cementation process, whereas Ni and Zn removal relied on adsorption onto ZVI corrosion products and coprecipitation with ZVI oxides. Thus, the processes were facilitated by low pH values. Because, in these tests, an alkaline solution was used, it can be supposed that ZVI corrosion was slowed. However, this may imply the functionality of the filter for a longer period of time.
The role of GAC in HMs removal from leachate is well known [19
]. For instance, its removal efficiency was tested by [52
] for solutions with pH ranging from 6.0 to 7.7 with the removal of several HMs (Cd(II)
, and Zn(II)
) in the range of 80–96% at an initial concentration of 184 mg/L. GAC can remove HMs through sorption on its porous surface, and through both surface and pore precipitation [53
]. Moreover, sorption can also occur with organic matter because HMs in landfill leachate are thought to form complexes with it [55
Similarly, lapillus is expected to remove HMs ions from aqueous solution via adsorption, as demonstrated in previous studies [17
Cu was, as expected, the metal most easily removed by both filters (Figure 1
). The ZVI/GAC filter performed slightly better than the ZVI/lapillus-filled column with RE
values of 91% and 85% on average, respectively. Similar results in terms of removal efficiency were reported by [15
] for a ZVI/GAC mixture (weight ratio 30:70 and 225 g of ZVI) but considering an acid leachate (pH = 5) and a longer residence time (flow rate of 0.1 mL/min).
Ni removal had an almost identical efficiency (66% and 67% on average) for both filters. However, the ZVI/lapillus-filled column removed Ni with a more regular trend compared to the ZVI/GAC filter (Figure 1
). Ni removal was far more efficient in [15
] where a ZVI/GAC granular mixture was tested toward an acid leachate (removal efficiency ranged from 93% to 80%). This behavior can be attributed to the aforementioned more favorable conditions for iron corrosion.
Lastly, the main difference between the two columns was exhibited on account of Zn removal. In fact, the ZVI/GAC granular mixture removed Zn better than the ZVI/lapillus filter with efficiencies of 75% and 58% on average, respectively. As in the case of Ni, Zn removal performance was considerably better for the ZVI/GAC filter studied by [15
ranged from 98% to 95% using an acid leachate) confirming the better aptitude of the filter for the treatment of acidic solutions due to iron corrosion. Nevertheless, the metal removal sequence of this test (Cu > Ni > Zn) was in accordance with that of [15
All results considered, the ZVI/GAC granular filter had a slightly better metal removal performance than the ZVI/lapillus granular mixture for MLL treatment. The aptitude of GAC for effective metal removal in alkaline conditions supported the ZVI, which performs better with acidic leachate. On the contrary, lapillus seemed less efficient in sufficiently supporting ZVI with metal removal; however, it is far cheaper because it is widely available and often produced as a by-product of pumice extraction, which is a significant advantage for its practical use.
3.2. Semicontinuous AD Tests
As shown in Table 2
, the composite samples of MLL pre-treated by ZVI/lapillus and ZVI/GAC mixtures had very similar compositions in terms of pH, in addition to ammonium, chloride, Cu, and Ni concentrations. Conversely, COD and Zn concentrations were lower in the pre-treated MLL derived from the ZVI/GAC column.
Maximum concentrations of HMs and chloride were calculated for each reactor, according to the MLLs and water volumes added from the second week onward (Table 5
), under the reasonable hypothesis that HMs and chloride were conserved (i.e., unaffected by the process).
In the case of HMs and chloride, concentrations in the AD tests were far lower than the inhibitory thresholds, which were found to be 500, 100, and 50 mg/L for Cu, Ni, and Zn [25
], respectively, and 4–9 g/L for chloride [57
]. For reactors B and partially D the HMs maximum expected concentrations already complied with the Italian discharge limits for soil (Legislative Decree 152/2006; Cu < 0.1 mg/L; Ni < 0.2 mg/L; Zn < 0.5 mg/L). In a full-scale application, an appropriate dosage and/or and earlier replacement of the filter material used in the pretreatment can be considered in order to fully comply with the standard required for the intended use of digestate. The use of a cheap admixing agent for the ZVI (i.e., the lapillus) may, in this case, reduce the cost of the replacement. Similarly, it can be stated that the low amounts added, in addition to the dilution with water and inoculum, would significantly reduce the concentration of persistent contaminants eventually present in AD reactors (e.g., pharmaceuticals and microplastics).
The pH trends of each semicontinuous reactor during the AD processes are depicted in Figure 2
. During the semicontinuous AD tests, the pH of all reactors decreased from an initial value of around 7.8, showing similar trends to those during the AD tests. The lowest value (6.2) was reached in reactor A on day 20, probably due to an accumulation of VFAs. On days 20 and 27, 4 g of NaHCO3
was added to all reactors in order to buffer the pH. As a result, a peak of 7.7 was measured on day 30 in reactor A. At the end of the tests, all reactors exhibited a pH of 6.9, except for reactor A (6.7). Despite the decreasing trends, the pH did not reach inhibitory values for methanogenic bacteria. However, the very irregular pH trend in reactor A could possibly hint at process imbalance.
Methane yields and cumulative methane productions for each reactor are depicted in Figure 2
. Reactor A exhibited a gradually growing methane production up to 0.3 NL/gVSadded
(days 16–20) when methane production abruptly stopped. Reactor B reached the highest methane yield (0.37 L of methane per grams of VS added) after the first week of operation. From days 10 to 30, methane yield was fairly stable, remaining between 0.22 and 0.30 NL/gVSadded
on average). During the last week of operation, methane production sharply decreased to 0.06 NL/gVSadded
on the last day of the test. Reactor C showed a trend similar to reactor B but with lower methane yield values. The maximum (0.27 NL/gVSadded
) was measured on day 13, before production gradually decreased. In reactor D, no considerable methane yield variations were measured. Indeed, methane yield remained close to 0.1 NL/gVSadded
for almost the entire test duration.
Cumulative methane productions of the four reactors were compared to that of the ideal AD of pure cellulose. The latter was determined by accounting for the cellulose biomethanation potential of 335 NmL/gVS (±25%) reported on the basis of the aforementioned UNI/TS 11703:2018 norm. After 38 days, the complete digestion of pure cellulose would have generated 7.2 NL of methane on average.
During the first 2 weeks of operation, the methane amount produced by reactor B process was consistent with the expected production from the complete conversion of cellulose, whereas at the end of the test, the final cumulative methane production was about 5.3 NL; in fact, this reactor was the best performing, with a final production equal to about 75% of the expected one. Reactor C was the second-best process in terms of the cumulative volume of methane produced, with a final value of about 4.1 NL and a production close to that of reactor B for the first 4 weeks. Reactor A reached a final methane production of 2.0 NL on day 16, as, from that point onward, methane generation was null. Conversely, reactor D generated methane throughout the experiment, despite maintaining low yields, achieving 1.6 NL of total methane production.
Both TS and VS trends were similar among digestates and decreased during the tests in each reactor. Considering the initial TS value of each reactor mixture (4%), solid matter was scarcely consumed by the microbial community during the first week of operation, whereas the solid content decreased from the second week onward. At the end of the tests, TS reached values of 0.6–0.8%. VS remained almost constant for the first 3 weeks of tests in all reactors before their presence sharply decreased. The abrupt and abnormal reduction in VS/TS from week 4 onward hints at a reduction in microbial activity in the reactors as discussed below.
During the first and second weeks of the tests, the highest contents of VFAs (852 and 975 mg/L, respectively) were recorded in reactor A, whereas those in reactors B, C, and D were similar (data for reactor C from the second week were not recorded because of a sampling problem). From week 3 onward, a continuous increase in VFAs was recorded in reactor A, again suggesting a probable imbalance of the process. The VFA concentration reached inhibitory levels from week 4 onward in this reactor. VFAs also increased to a lesser extent in reactors C and D in weeks 4 and 5. The drop in the level of VFAs in week 6 when methane production was already scarce, even in the best-performing reactor (i.e., B), may have been due to a possible inhibition of hydrolysis.
In the centrifuged composite samples from the first week, a COD of about 3200 mg/L was measured in each reactor. Despite the aforementioned sampling problem for reactor C, in the second week, slight decreases in COD content were recorded for all reactors. From the third week onward, COD concentration in reactor A started increasing again, reaching a peak of 4935 mg/L at the end of the experiment. The COD concentration in reactors B and C increased from the fourth week until the end of the tests, but with lower values compared to reactor A. Conversely, the COD concentration in reactor D smoothly decreased during the digestion, with the lowest value of 680 mg/L recorded in the fifth week, only increasing in the last week to 1755 mg/L.
The highest NH4
-N concentrations were recorded in the first week of tests for reactors B, C, and D (519, 565, and 521 mg/L, respectively) and in the second week for reactor A (527 mg/L). In reactor A, ammonium content decreased to 306 in the fourth week, before increasing until the end of the experiment. In other reactors, analogous trends were recorded but with NH4
-N concentration values, which were lower in reactors B and D and similar in reactor C compared to reactor A. As with HMs and chloride (Table 5
), the NH4
-N content in the AD reactors was lower than the inhibitory threshold (1.7 g/L) found in the literature [37
Methane yields and cumulative methane productions depicted in Figure 2
clearly show that, with the exception of reactor D, methane production was not steady; moreover, after an initial production quite close to the “ideal” expected for cellulose, the process showed clear signs of inhibition. This was early and abrupt for reactor A, where after day 16, methane production totally stopped, whereas it was more delayed for reactor B, i.e., the best-performing system, where it was significant and quite steep from day 30 onward. By comparison, inhibition was more gradual in reactor C, with a steady and slow decline beginning on day 13. The behavior of reactor D was very peculiar, whereby the methane production was very slight but regular throughout the test, equal to approximately 25% of the “ideal” production expected for pure cellulose. The analysis of trends for pH, VFAs, and NH4
-N (Figure 3
) did not justify this observation. This behavior could be classified as an “inhibited steady state”, i.e., a condition where the process ran stably but with methane yield lower than expected due to limited bacterial activity [37
]. The identified inhibition of methanogenesis is most probably not attributable to an excess of metals or chloride (Table 5
) or NH4
-N (Figure 3
) because, as already shown, the concentrations of those parameters were well below their respective inhibition thresholds. Ref. [29
] reported the inhibition of anaerobic processes (particularly hydrolysis and methanogenesis) due to the presence, as in this case, of noticeable amounts of HSs; moreover, they also stated that the inhibition was irreversible and exacerbated by an increase in the HS/VS ratio. Because the biomass could not adapt to the presence of HSs, the abundance of microorganisms decreased, which is in agreement with the reduction in VS measured for all reactors. According to the same paper, HSs interfere with hydrolytic and methanogenic enzymes.
It is interesting to highlight that the two reactors where methane production was more noticeable (reactor B), or more regular albeit quite scarce (reactor D), were those where the addition of pre-treated synthetic MLL and, thus, of HSs was lower.
On the contrary, reactor A, supplemented with a higher amount of COD-rich pre-treated leachate, had a more significant presence of HSs and showed a more abrupt inhibition of AD.