5 Explaining the Microbial Communities of the WWTPs

5.1 The influence of plant design on the microbial communities

To investigate possible causes of the differences between the WWTPs other than the influent wastewater, the influence of plant design of the WWTPs is interesting to investigate. Whether differences in plant design have an influence on the microbial communites of the activated sludge of a full scale WWTP is nearly impossible to test experimentally. This would require two identical WWTPs with the exact same influent wastewater, weather conditions, dimensions etc, where one WWTP is designed with one feature which the other WWTP does not have, and the microbial communities of each WWTP then compared. By using ordination it is possible to examine the correlation between the microbial community and environmental variables like plant design as an alternative approach. This can be done by either plotting the environmental variables onto an ordination plot creating a triplot (Figure 5.1), where the relative positions of the labels represent the probabilities of the variables to correspond to samples nearby, or it can be done by constrained/canonical ordination of the variables individually (Figure 5.2), as already described. A table of the design characteristics of each WWTP which will be investigated can be found in Appendix C.

A CCA triplot (Figure 5.1) shows that the AS samples seem to be correlated with most of the design characteristics. The four characteristics plotted are whether the WWTPs have a digester or not (“DigesterYes” and “DigesterNo”), whether they have primary setling or not (“PrimarySetlingYes”, and “PrimarySetlingNo”), whether they utilise Enhanced Biological Phosphorous Removal (“DesignEBPR”) or Biological Nutrient Removal (“DesignBNR”), and lastly whether there are alternating aerobic/anaerobic conditions or not (“ConfigurationAlternating” or “ConfigurationRecirculation”). The correlation of the characteristics (or factors) are significant as all four have a P-value below 0.001 with 999 random permutations, so the design of the WWTPs seems to be correlated with the composition of the microbial communities. The fact that the positions of the labels are not all exactly at the center (0,0) supports this. However, by interpreting the positions of the labels, the magnitude of the correlation seems small and sometimes misleading when interpreting the individual WWTPs. Take for example Aars (-0.5,-0.8), which is positioned very close to PrimarySetlingYes, but it does not feature it (see Appendix C). Similarly Kerteminde (0.5,0.0) is positioned very close to ConfigurationAlternating, but is expected to be positioned near ConfigurationRecirculation. There are several more examples of this kind, which suggests that there must be additional, unknown factors having a larger direct influence on the microbial communities than the ones investigated. When doing analyses involving correlation, it is thus important to remember that correlation does not imply causation, and the observed correlations can only provide a hint of what could be interesting to investigate and confirm experimentally.

Supporting the CCA triplot (Figure 5.1) with constrained ordination of some of the individual characteristics (Figure 5.2), it is again clear that the differences between the WWTPs can be partly explained by their design characteristics. It is important to note that both axes can be interpreted in Figure 5.2(D), but only the first axis in Figure 5.2(A+B+C) (due to the number of axes obtained with constrained ordination are always one less than the number of different possibilities of the particular variable).

Figure 5.1: Canonical Correspondence Analysis of samples from the 32 WWTPs. This figure is identical to Figure 4.3, except that only the labels of the WWTPs have been plotted for clarity. Labels of the correlation between samples and different plant designs are plotted at their center of gravity, where their relative position represents their significance.

In general, the percentages of the first axes are low in all four plots (between 1.8% and 2.4%), which indicates that the differences are small.

Alternation vs recirculation (Figure 5.2(A)) seems to be the characteristic with the least influence on the OTUs as they are not clearly separated on the first axis and the two groups both overlap the OTUs on the first axis. In Figure 5.2(B+C) some of the OTUs are positioned on a vertical line directly through the center of the two groups, which is a clear indication that these OTUs are representative of the particular group. Any OTUs between these two vertical lines are shared among both groups. This is most evident in Figure 5.2(B), where there are many OTUs positioned on a wide line at x=-0.5 corresponding to EBPR, and similarly also OTUs on a line at x=2 corresponding to BNR. It is worth noting that only 9 WWTPs utilise BNR, while the remaining 23 WWTPs utilise EBPR. This explains why most of the OTU points near the center (0,0) are closer to the EBPR group and not BNR. This is also evident in Figure 5.1, where the DesignEBPR label is close to (0,0) and DesignBNR at (0,-1). The microbial communities of WWTPs that are utilising Enhanced Biological Phosphorous Removal (EBPR) have been studied extensively in recent years and several key bacteria are confirmed to play an important role in the process, for example the Phosphate Accumulating Organisms (PAO) Tetrasphaera and Accumulibacter (Kristiansen et al., 2013; Mino, Loosdrecht, & Heijnen, 1998; P. H. Nielsen et al., 2010; R. Seviour & Nielsen, 2010). In Figure 5.2(B) these bacteria are positioned to the left near the EBPR label (unfortunately it is not possible to search in the plots), which confirms that the WWTPs utilising EBPR may indeed have a different microbial community composition. With primary setling (Figure 5.2(C)) the positions of the OTUs are not as clearly separated on the first axis, and there do not seem to be particularly unique OTUs corresponding to each group, a large majority of the OTUs seem to be shared. Lastly, the amount of industrial wastewater in the influent wastewater (Figure 5.2(D)) also seems to be influencing the microbial communities of the WWTPs.

Figure 5.2: Canonical Correspondence Analysis with the constraints (A): Alternation vs Recirculation, (B): Enhanced Biological Phosphorous Removal (EBPR) vs Biological Nutrient Removal (BNR), (C): Primary Setling, and (D): the amount of industrial wastewater content, where Low is 5%<10%, Avg. is 10%<35% and High is 35%<100%. The points represent samples colored by the particular constraint, OTUs are marked as grey circles.

As expected, all three groups (Low, Avg., and High) have many OTUs in common, but there are many OTUs which seem to correspond to a high industrial wastewater content, and likewise several OTUs that correspond only to the WWTPs having a low amount of industrial wastewater content, however not as many. The influence of the amount of industrial wastewater in the influent have not been extensively studied previously, however, but one study by Ibarbalz, Figuerola, & Erijman (2013) used constrained ordination to analyse the differences between industrial WWTPs, and concluded that their microbial communities were clearly distinct from those of municipal WWTPs.

5.2 General differences related to sampling time

The microbial communities of the 32 WWTPs appear to be changing slightly over time as revealed by CCA (Figure 5.3(A)). It is clear that the year 2014 seems strikingly different from the rest of the years, but this is believed to be the result of a different way of preparing the samples from this particular year. When taking a closer look into which OTUs seemed unique to the year 2014, many of the OTUs have only been classified to a kingdom (bacteria) or a class level, which indicates that something went wrong during one of the experimental steps, either during DNA extraction, PCR, or DNA sequencing. It would simply seem impossible that all 32 WWTPs in one year differed from other years in the exact same way, when the WWTPs have nothing in common except that they are all located in Denmark. It is important to mention that the samples from 2014 have not been filtered in the analyses presented in this report, however doing so did not result in any noticable differences. A CCA plot without the 2014 samples can be found in Figure A.10 in Appendix A. Besides 2014, there seem to be a small difference between each year. It is not easily distinguishable exactly which OTUs could be representative of each year, it seems more like the differences are generally caused by differences in the OTU read abundances.

Figure 5.3: Canonical Correspondence Analysis constrained to which year (A) or seasonal period (B) the samples were taken. The labels of each year or seasonal period have been marked at the approximate centroid of the corresponding samples.

Furthermore, it is not possible to conclude whether the observed clusters per year are actually the result of the samples being taken within the same year, or whether they simply are the result of identical experimental processing of the samples from each year. However, the differences are small, as the eigenvalue percentage of the main axis is relatively small (3.8%). It is expected that samples taken at different seasonal periods of the year (summer/winter) would be different as a result of differences in temperature, which is perhaps the environmental factor with the largest impact on the thriving of microbes (J Farrell & A Rose, 1967). As seen in Figure 5.3(B), this seems to generally be the case in all the WWTPs, however there do not seem to be distinct clusters of samples per period, they are largely overlapping.