Figure 3.6 - Temporal diversity and physicochemistry of geothermal features across the pH scale.
(a) The microbial communities of 12 geothermal springs were measured six times over one year by amplicon sequencing of the 16S rRNA gene, with variation in alpha diversity between temporal samples indicated by OTU richness (i.e., the number of OTUs per community). (b) Alpha diversity was also measured by the Shannon diversity index to indicate evenness of the communities over time.
(c) Spring pH of the features at the time of sampling. (d) Spring temperature of the geothermal features. (e) Beta diversity of the 12 features is shown by a non-metric multidimensional scaling (NMDS; n=62, stress=0.15) of Bray-Curtis dissimilarities between all temporal samples, with data ellipses generated from multivariate t-distribution with a 95% confidence level. All samples are coloured according to feature name, with alpha diversity and physicochemistry further grouped by pH group.
While analysis of microbial community response to both natural and anthropogenic disturbances corroborated fluctuation in abundances of taxa irrespective of measured environmental change, variation to both taxa composition and relative abundance were noted in geothermal features in the immediate vicinity of perturbations. The influence of Inferno Crater Lake, as evidenced by the dominant population of Thermoproteota (family Sulfolobaceae) and physicochemical conditions during overflow (Figure 3.1), was clearly apparent in Waimangu Stream sites downstream of the pulse disturbance. The pH of the stream was more readily affected than temperature during this coalescence (pH 6.7 to 3.0, and 49.9 to 56.6 °C; Waimangu Feature 4; Figure C.4), indicating minimal buffering capacity of the stream water; with the volumetric rate of Inferno Crater Lake overflow (~79 l s-1) only marginally increasing the temperature of the stream. However, these physicochemical changes to Waimangu Stream are not conserved as the Inferno Crater Lake overflow is short- lived (2-3 days)139. With microbial communities in the stream experiencing this disturbance every 30-40 days, we now know resident populations quickly revert back to pre-disturbed states; the previous overflow event to this study ended eight days prior to sampling (Figure C.2). Waimangu Stream (~104 l s-1) provides a constant reservoir of microbial populations from Frying Pan Lake that can quickly re-colonise downstream sites once the disturbance has ended, thereby providing a short residence time for the disturbed communities266. Frying Pan Lake also undergoes reservoir cycling, albeit less extreme than Inferno Crater Lake259, which could explain the alternating abundances of both Cyanobacteria and Pseudomonadota observed in all stream sites. Similar to reports on freshwater microbial assemblages267,268, these findings suggest geothermal stream communities are resilient and can return to a pre- disturbance state if physicochemical conditions stabilise, and sufficient re-colonisation is attainable. However, microbial variation in Waimangu Stream and Inferno Crater Lake were not clear in alpha and beta diversity measures, with the greatest temporal dissimilarity in beta diversity occurring in the two sites furthest away from the disturbance (Figure C.4). These features could be influenced by other geothermal springs not associated with upstream samples, and/or proximity to the stream outflow into Lake Rotomahana likely facilitates lacustrine water inputs to these ecosystems, supported by an observed increase in Bacteroidota which are commonly found in freshwater lake epilimnia269. Nevertheless, even though phylum-level changes occurred in stream communities immediately downstream of the coalescence disturbance point, these short-term variances were not reflected in diversity analyses.
Conversely, the long-term disturbance at the Waikite geothermal field was most evident in the beta diversity of temporal microbial communities at the outlet to the wetland, which corresponded with the greatest variation in physicochemistry of all sites studied in this area (Figure 3.2). Even though there was a temperature decrease of 20 °C at the outlet, diversity only correlated with a decrease in pH from 8.6 to 7.6, indicating that a larger magnitude of change in temperature may be required than pH to influence microbial populations in geothermal ecosystems. Interestingly, alpha diversity of all stream sites around the wetland showed increases in diversity in the first timepoint post-disturbance, but these had returned to pre-disturbance levels by the final timepoint (Figure 3.2). These observations advocate the importance of using multiple diversity metrics to investigate microbial communities over time so that a cohesive picture of ecosystem behaviour is obtained270. Both relative abundances and beta diversity of microbial communities within (Features 5 and 6) and outside the wetland (‘Pig Scorcher’ and ‘Big Spring’) fluctuated irrespectively of the perturbation (Figure 3.4), with no correlation to pH or temperature (Figure C.5). Features 5 and 6 have additional geothermal inputs to Otamakokore Stream, and are relatively low volume features, suggesting they are more susceptible to fluctuation in environmental conditions86. A notable temperature difference (93.8 to 87.2 °C) was observed for the ‘Pig Scorcher’ spring after the weir installation, however, this physicochemical change was not attributed to wetland dynamics260. It should be noted that both sampling timepoints after the disturbance occurred outside of the austral summer, which could partially explain the decrease observed in Cyanobacteria at all wetland sites. Even though limited Cyanobacterial change was reported with or without ultraviolet radiation in geothermal microbial mats249, seasonal variation was detected in other geothermally-associated Cyanobacteria populations between dry and rainy seasons251. Greater temporal variation can occur in the microbial communities of geothermal sediments than associated water columns175, and with shorter residence times for taxa in stream sites than other geothermal features, physical processes enabling community assembly through time would conceivably alternate across sample types.
Overall, these findings suggest the abundances of planktonic microbial communities acclimatise to sustained physicochemical change in the local environment, but there is a general instability or hysteresis in geothermal habitats, even when physicochemistry appears relatively stable.
Similar to the previous categories of geothermal features, relative abundances of microbial communities from all pH groups varied over the one-year sampling timeframe (Figure 3.5),
with temperature fluctuating more than pH in these habitats (Figure 3.6). The pH 7 group had the most change in alpha diversity, particularly for Shannon diversity index (Figure 3.6), which could be a result of the increased number of taxa that prefer circumneutral conditions153. In contrast, the greatest changes in beta diversity were detected in pH 3, 5 and 9 groups (Figure 3.6), which corresponded with the variation observed in relative abundances of taxa in these features. Like microbial community analysis of Inferno Crater Lake, these results suggest alpha diversity fails to adequately measure temporal variation outside of circumneutral pH. This could be due in part to reduced diversity potential to fill any voids created by changing communities at either ends of the pH spectrum for geothermal springs, with the magnitude change that occurs with varying pH outside of circumneutral being more impactful. Previous longitudinal research has suggested alpha diversity can fail to present change despite highly divergent microbial communities270, again reinforcing the use of multiple diversity metrics when investigating temporal scales. Whakarewarewa Feature 51 (pH 3 group) was the only feature that significantly correlated with pH for both alpha and beta diversity metrics (Figure C.8), with a corresponding pH increase of 3.3 to 4.8 observed.
This feature also had a positive correlation between diversity and temperature, along with six other features from across the pH range (Figure C.8), suggesting temperature is more likely to fluctuate over time than pH in geothermal features. At least eight of the features in this category are shallow (<2 m in depth), and so are more susceptible to meteoric and groundwater fluctuations that occur seasonally in Aotearoa-New Zealand271.
Comparable to the microbial community of Inferno Crater Lake, Thermoproteota populations (family Sulfolobaceae) decreased in Whakarewarewa Feature 51 as the temperature decreased and pH increased over the study period (Figure 3.6 & Figure C.6). This population decrease, characteristic to Sulfolobaceae when pH increases272, was counteracted with increased Aquificota, and indeed, Aquificota featured prominently in geothermal community fluctuations from across the pH scale. This could be a result of the widespread abundance of this phylum found throughout the TVZ195, but temporal changes to Aquificota populations have also been observed in other geothermal systems worldwide72,139,169, suggesting hypersensitivity in this phylum to physicochemical change. Interestingly, Ohaaki Feature 2 (pH 9 group) had an almost complete community shift from Aquificota to Deinococcota (Figure 3.5), which correlated with both groundwater and hydrogen sulfide levels. Hydrogen sulfide gas frequently percolates through geothermally-heated soil in the TVZ273, and consequently is common to geothermal springs that are influenced by groundwater.
Therefore, this observed shift indicates water table levels directly affect the microbiology of this spring. We also know that historic modification, stemming from local geothermal power generation, influenced both temperature and water levels of the spring274, so anthropogenic impacts could explain the community variation observed in this study. Regardless, geothermal features from a range of source fluids across the pH scale exhibited both stochastic and deterministic fluctuation in the relative abundances of microbial communities, proposing that all geothermal features do not adhere to the same processes of community assembly through time.