Relationship between Microorganisms Inhabiting Alkaline Siliceous Hot Spring Mat Communities and Overflowing Water

In flowing aquatic systems, cell erosion and deposition are important to the dispersal of cells from one location to another. Very little is known about microbial dispersal and the physical processes that underlie it. This study demonstrates its importance to colonization of downstream surfaces and especially to the recolonization and functioning of disturbed sites. Ecological systems in flowing environments are often, roughly speaking, pseudosteady, in that nutrients enter the system and by-products leave at relatively steady rates. Over time, material inputs and outputs must balance. Measurements of input fluxes (e.g., growth rates and proxies, such as photosynthesis rates) are frequent. However, erosion and deposition of cells are seldom measured and ecological significance is sometimes neglected. The importance of these parameters is immediately evident in any attempt to construct a model of long-time community behavior, as spatial ecological structure is significantly impacted and can be dominated by migration of organisms, even in small numbers.

ABSTRACT

The compositions of Octopus Spring and Mushroom Spring (Yellowstone National Park, Wyoming, USA) microbial mats have been thoroughly studied, but the compositions of the effluent waters that flow above the mats have not. In this study, cells in the mats and overflowing waters of both springs were investigated at multiple sites where Synechococcus spp. are the dominant cyanobacteria (ca. 72°C to ca. 50°C), and on several dates. In addition to microscopic analyses of stained and autofluorescent cells, 16S rRNA gene sequencing was used to characterize the major taxa present and a protein-encoding gene (psaA) was sequenced and analyzed by ecotype simulation to predict species of Synechococcus. The mats of both springs were similar in terms of the downstream distribution of predominant taxa detected previously. However, waters above these mats were predominated by taxa that reside in upstream mats or communities above the upper-temperature limit of the mat. A disturbance/recolonization study was performed at a site normally predominated by Synechococcus species adapted to low temperatures. After removing indigenous Synechococcus cells, Synechococcus species adapted to higher temperatures, which were predominant in the water overflowing this site, colonized the newly forming mat. Differences in recolonization under reduced and UV-screened irradiance suggested that, in addition to physical transport, environmental conditions likely select for species that are better adapted to these different conditions and can influence mat recovery. A transport model was developed and used to predict that, in Mushroom Spring, erosion predominates in the narrower and deeper upstream effluents and deposition predominates over erosion in wider and shallower downstream effluents.

IMPORTANCE In flowing aquatic systems, cell erosion and deposition are important to the dispersal of cells from one location to another. Very little is known about microbial dispersal and the physical processes that underlie it. This study demonstrates its importance to colonization of downstream surfaces and especially to the recolonization and functioning of disturbed sites. Ecological systems in flowing environments are often, roughly speaking, pseudosteady, in that nutrients enter the system and by-products leave at relatively steady rates. Over time, material inputs and outputs must balance. Measurements of input fluxes (e.g., growth rates and proxies, such as photosynthesis rates) are frequent. However, erosion and deposition of cells are seldom measured and ecological significance is sometimes neglected. The importance of these parameters is immediately evident in any attempt to construct a model of long-time community behavior, as spatial ecological structure is significantly impacted and can be dominated by migration of organisms, even in small numbers.

INTRODUCTION

The microbial communities of alkaline siliceous hot springs have been studied for many decades (1–6). Chemolithotrophic streamer communities are sometimes observed in waters near the source pools. For instance, in Octopus Spring (Yellowstone National Park, Wyoming, USA), pink streamers are formed by members of the Aquificales, such as Thermocrinis spp. (7, 8) and Thermotogales (3). As the effluent water flows away from the spring, it cools, such that at temperatures below ∼72 to 74°C, photosynthetic mat communities are formed. These mats are associated with unicellular cyanobacteria (Synechococcus) at higher temperatures, together with filamentous cyanobacteria, such as Leptolyngbya and Fischerella, below about 58°C, and Calothrix below ∼40°C (2).

With the advent of 16S rRNA analyses, it was realized that many distinct Synechococcus populations are distributed at different temperatures along the flow path (9, 10), likely due at least in part to adaptation to different temperatures (11, 12). A progression of closely related Synechococcus 16S rRNA variants, termed A'', A', A, B', and B, was shown using denaturing gradient gel electrophoresis to be distributed from the mat’s upper temperature limit to 50°C. Closely related 16S rRNA variants of Roseiflexus and green sulfur-like bacteria (later described as “Candidatus Thermochlorobacter”) (13) were also observed to be distributed differently along the thermal gradient (9, 10). Miller et al. (14) observed similar, though not identical, temperature progressions in 16S rRNA gene barcode analyses of mats inhabiting White Creek and Rabbit Creek in Yellowstone National Park. In addition to observing thermotypes of Synechococcus and Roseiflexus, thermotypes of Chloroflexus and Chloracidobacterium were also observed. Because the conserved nature of 16S rRNA sequences prevented demarcation of all Synechococcus species populations (15–17), Becraft et al. (18) conducted barcode analyses of Synechococcus spp. based on the less-conserved photosystem I reaction center gene, psaA. When psaA variants were analyzed using an evolutionary simulation model called Ecotype Simulation (19), it was possible to predict putative ecotype (PE) populations, which are equivalent to ecological species so long as it is possible to demonstrate the ecological distinctions of PEs and the ecological interchangeability of individuals within each PE. Becraft et al. (18) demonstrated these properties for several Synechococcus PEs, the members of which had distinct vertical distributions in the mat, and this was consistent with different light adaptations of isolates representative of these PEs (20, 21).

In contrast, surprisingly little attention has been given to the cells in the effluent water flowing over these communities and to the microbiology of the water in relation to these benthic communities. As a follow up to experiments on estimating in situ growth rates of the Synechococcus spp. in Mushroom Spring (22), Brock (1) reported counts of Synechococcus of 8.1 × 10³/ml at a 68°C site, 1.32 × 10⁴/ml at a 65.2°C site, 2.31 × 10⁴/ml at a 59.1°C site, and 1.88 × 10⁴/ml at a 57.2°C site in the effluent channel. The analysis of 16S rRNA sequences in the 90°C water emerging from Fairy Geyser by Boomer et al. (23) revealed a variety of Firmicutes, Nitrospirae, Planctomycetes, Proteobacteria, and Thermotogae sequences, but sequences related to Thermocrinis and Thermus were most prevalent, constituting 39% and 20% of the recovered sequences, respectively. Molecular studies of cells in the water flowing above the mats of Octopus Spring and Mushroom Spring have not yet been done.

Characterizing the coupling between organisms in the mats and overflowing water is important for understanding dispersal and the recolonization of disturbed sites. For instance, molecular analyses of physically disturbed sites showed recolonization by Synechococcus 16S rRNA genotypes typically found at higher-temperature upstream sites (24), suggesting the importance of cell dispersal in the effluent waters, but no analyses of cells in the water were performed in that study. The opportunity for viable cells of different species, once released from the mat in some manner, to settle and reattach downstream also introduces potential for impact on mat ecology and productivity. As an example, during a 40-day recovery period following disturbance of regions of the mat, near-surface oxygenic photosynthesis rates, as measured using microsensor analyses, only reached about two-thirds of the rate prior to disturbance (24). This was likely due to recolonization by A'-like Synechococcus, which are adapted to higher temperatures than occur at the downstream recolonization site (24).

Little is known about the erosion of cells to the overflowing water and the deposition of cells onto the mat. In the study mentioned above, Brock (1) reported settling rates of 3.3 × 10² to 2.3 × 10³ Synechococcus cells/cm²/day onto glass slides placed atop the mat. He compared this to estimated productivity rates of 1.6 to 8.1 × 10⁸ cells/cm²/day also reported in that study and inferred that settling did not appear to be a major contributor to mat productivity. The relationship between mats and cells in the flow is more complex than settling alone, as erosion of cells from the mat surface is also expected to occur. These processes are dependent on effluent fluid dynamics and the complexity of flow (e.g., flow velocity, advection, and horizontal and vertical turbulence), and thus necessitates a modeling approach (25). While growth rates and proxies (e.g., photosynthesis rates) have been measured (1, 26, 27), the ultimate fate of carbon fixed in these communities is less certain. In this study, we developed a model that takes into account the underlying physical phenomena and used it to examine the importance of inputs and outputs of cells to and from the mat to overall mat productivity. While consideration of effluent, or bulk, fluid dynamics has sometimes been made in existing models of microbial communities in the context of substrate transport, less attention has been given to its direct connection to long-term ecology. The methodology presented here provides a means to include ecological effects of fluid transport in broader microbial community models.

Our aims in this study were to (i) better characterize the cells in the water flowing above the microbial mats, (ii) explore the consequences of downstream dispersal of Synechococcus species to mat recovery after disturbance, and (iii) develop and parameterize physical models of Synechococcus mat communities that could be used to predict the relationship between cells in the mats and water. To these ends, we conducted microscopic analyses of cells in the flow and an in-depth analysis of the 16S rRNA gene and psaA sequences associated with cells in the flow in relation to cells in benthic mats. We studied both Mushroom Spring and Octopus Spring, as the flow regimes (and also temperature regimes) in these springs are quite different. At the time of our study, Mushroom Spring had an ∼68°C source pool that was lined by Synechococcus mats, and flow was largely steady from this pool to effluent channels. In contrast, Octopus Spring had an ∼92°C source pool and was colonized upstream of Synechococcus mats by pink and yellow streamer communities at ∼80 to 90°C and 75 to 80°C, respectively. Flow surges occurred at Octopus Spring on a periodicity of 4 to 7 min. We also extended previous studies by examining recovery of Synechococcus populations after disturbance of the Mushroom Spring mat community under different irradiance regimes in order to evaluate the importance of adaptation, as well as physical transport. Finally, we developed a physical model, which was parameterized using direct measurements of cells and physical features of these systems, and used it to predict the processes of erosion and deposition of cells downstream.

RESULTS

Site establishment and characterization.

Five collection sites were selected along the temperature gradient of one major effluent channel each for both Mushroom Spring (MS0, MS1, MS2, MS3, and MS4) and Octopus Spring (OS0, OS1, OS2, OS3, and OS4) (Fig. 1). Seven sites in Mushroom Spring were selected for velocity calculation (L₀ to L₇; see Fig. 1). Sites L₀, L₁, L₅, and L₇ corresponded to collection sites MS0, MS1, MS3, and MS4, respectively. In order to estimate flux into or out of the collection zones, site L₂ was located between sites MS1 and MS2, sites L₃ and L₄ were located between sites MS2 and MS3, and site L₆ was located between sites MS3 and MS4 (Fig. 1A). During each field visit (2 June, 11 August, and 21 September 2011), mat and water samples were collected from approximately the same spatial locations in the channel (verified by photographs and conspicuous features in the mat or along the edge of the flow channel).

FIG 1.

Sites studied at Mushroom Spring (A) and Octopus Spring (B). Because collection sites (MS or OS) were chosen to sample certain temperatures, not all were suitable for stream velocity measurements. To account for this, additional sites (L) were selected and used only for velocity measurements. In panel B, the vertical arrow points to the source pool and the slanted arrow indicates the effluent channel containing pink streamers; SR, shoulder region.

Temperatures fluctuated widely over a diel cycle at sites in Octopus Spring, but they were relatively stable at the Mushroom Spring sites (Fig. S1A in the supplemental material); time-averaged temperatures remained relatively constant throughout the study period (Fig. S1B and C). Occasional failure of the iButton monitoring equipment (see the Materials and Methods) prevented continuous data collection at some sites in Octopus Spring. During the study period, the intensity of downwelling light at both sites was relatively constant in June and declined through July and August to about 85% of the June daytime average (Fig. S2A and B). A failure of the light meter at Mushroom Spring prevented acquisition of data between mid-August and mid-September.

Stream velocity at Mushroom Spring ranged from approximately 7 cm/s to 19 cm/s depending on location (Table 1). The stream started with a low average velocity of 7.67 to 8.95 cm/s near the source pool (site L₁) and sped up to 13.4 to 18.5 cm/s as it flowed through a relatively narrow (18.5 to 19.4 cm), but relatively deep (4.25 to 6.41 cm) channel and descended a small hill (sites L₂ to L₄). The channel turned a corner shortly past site L₅, then spread out into a relatively wider (32.5 to 43.39 cm) and shallower (1.64 to 2.81 cm) flow path (sites L₆ and L₇), with little change in average velocity, which ranged from 11.0 to 18.5 cm/s. The overall average of measurements taken at five locations on two different dates was 13.8 cm/s with a standard deviation of 4.2 cm/s. This corresponds to an average flux (using channel geometry data) of 1 liter/s. Measurements in Octopus Spring were not possible due to the relatively fast time scale and magnitude of changing flow during surges (Fig. S1A).

TABLE 1.

Summary of stream geometry and average velocity measurements (± standard deviation) for Mushroom Spring on 2 days in 2012

Date	Parameter	Siteb
		L1	L2	L3	L4	L5	L6	L7
15 June 2012	Width (cm)	19.4 ± 2.2	18.8 ± 2.0	-	-	43.39 ± 9.53	41.7 ± 7.85	-
	Depth (cm)	6.41 ± 0.41	4.25 ± 1.09	-	-	2.81 ± 0.81	1.69 ± 0.46	-
	Velocitya (cm/s)	7.67 ± 0.17	15.0 ± 2.0	-	-	11.0 ± 0.21	17.8 ± 4.0	-
21 Sep 2012	Width (cm)	18.5 ± 2.4	17.2 ± 2.6	-	37.5 ± 2.1	-	32.5 ± 0.95	-
	Depth (cm)	4.89 ± 0.7	4.4 ± 1.68	-	2.1 ± 1.0	-	1.64 ± 0.1	-
	Velocitya (cm/s)	8.95 ± 0.7	18.5 ± 4.1	-	13.8 ± 1.6	-	18.5 ± 4.0	-

^aVelocity measurements were calculated from 18 to 66 trials, depending on time and location.

^bSymbol -, not measured.

Cells in water overflowing sites.

Figure 2 shows representative results of microscopic analyses of cells filtered from water samples at sites in Mushroom Spring on 21 September 2011. The images shown were selected from a larger set of five images for each sample. In addition to red-autofluorescent Synechococcus cells, numerous small rod-shaped cells were observed. Similar results were observed for samples collected on 11 August 2011 (data not shown) and also for Octopus Spring samples, where Synechococcus cells were rarer (Fig. S3), and filaments were observed in the upstream sites (Fig. 3).

FIG 2.

Representative fluorescence photomicrograph images of cells filtered from water samples above Mushroom Spring sites collected on 21 September 2011, using SYBR gold and FITC (left) and rhodamine (right) filters. Red-autofluorescent Synechococcus cells average 7.35 μm in length. White bar indicates 10 μm.

FIG 3.

(A) Pink streamers found in the Octopus Spring effluent channel at 85 to 88°C. (B and C) Photomicrographs of cells filtered from water collected at sites OS0 and OS1 on 11 August 2011 and stained with SYBR gold. Small stained cells average 2.89 μm in length. White bar indicates 10 μm.

As shown in Fig. 4A, at Mushroom Spring an increase in total cell counts was observed in upstream water samples (MS0 through MS2) collected on 11 August and 21 September 2011. A similar trend was noted for Synechococcus cell counts in August and September (Fig. 4C). Total cell counts and Synechococcus cell counts were lower at sites MS3 and MS4. The total cell count ranged from 5 × 10³ to 2.5 × 10⁴ cells/ml and the proportion of Synechococcus cells increased as water began to flow rapidly over the Synechococcus mat, from 0.73 to 2.15% at MS0 to as high as 34.8 to 42.3% of the total cell count at site MS2 in August and September. The percentage of Synechococcus cells dropped at sites MS3 and MS4, possibly indicating settling of cells onto downstream mat regions.

FIG 4.

Densities of total cells (A and B) and Synechococcus cells (C and D) in water samples above sites as a function of month and sampling site in Mushroom Spring (A and C) and Octopus Spring (B and D) during the study period. Bars represent standard error (n = 5, the number of images sampled from each water sample). Note scale differences between total cell and Synechococcus cell counts.

In Octopus Spring, total cell counts were more similar across sites (Fig. 4B), possibly due to the flow surges. Although Synechococcus cells were absent or very low at site OS0 (above the upper temperature limit of the mat) and site OS1 (just at the upper temperature limit of the mat), Synechococcus cell counts increased at sites OS2, OS3, and OS4 (Fig. 4D). Synechococcus cell densities were lower in Octopus Spring water, ranging from 0 to 17.1% of the total cell count.

Molecular analyses of cells in the mat and in the flow at collection sites.

DNA was extracted from mat and water samples collected on 2 June 2011. Portions of the 16S rRNA gene and the psaA gene were PCR amplified and sequenced using high-throughput methods. This yielded between 1,248 and 3,715 partial 16S rRNA gene sequences per sample, and between 384 and 3,174 partial psaA sequences per sample (see Table S1). These results are a subset of a much larger matrix of samples involving sequencing of these two loci in mat and water samples collected at five stations in both springs at monthly intervals over an entire year. The scale of this effort prohibited the rote replication of samples for distribution studies. Nevertheless, we are confident in our observations based on (i) the observation of similar distributions of Synechococcus PEs in water overflowing the Mushrooom Spring mat in samples collected on 15 December 2011 (data not shown); (ii) the reproducibility among replicates observed in previous molecular analyses of this kind (9, 10, 18, 24, 28), as well as in the disturbance study reported herein; and (iii) the pseudoreplication of making similar observations along thermal gradients in two different, yet chemically and biologically similar, hot spring effluents.

(i) 16S rRNA-defined taxa.

(a) Mushroom Spring. As shown in Fig. 5, the mat in Mushroom Spring contained 16S rRNA gene sequences of predominantly Synechococcus-like and Chloroflexi community members. The relative contribution of A'- and A-like Synechococcus was greater at high-temperature sites, whereas at lower-temperature sites, B'- and B-like Synechococcus were more abundant. The relative contribution of Chloroflexus-like 16S rRNA genes was greater than that of Roseiflexus-like 16S rRNA genes at higher temperature sites, but lower at sites MS3 and MS4. Armatimonadetes sequences were detected at site MS2. “Ca. Thermochlorobacter”-like sequences were found at sites MS3 and MS4.

FIG 5.

16S rRNA gene sequence profiles for bacterial phyla in Mushroom Spring water (A) and mat (B) collected at sites along the flow path on 2 June 2011.

The water overflowing the MS mat contained predominantly Thermus-like sequences. Synechococcus-like sequences were next highest in relative abundance in water from the source pool (MS1), which is lined with a Synechococcus mat. The contribution of Synechococcus-like sequences was higher in water flowing above downstream sites. The relative abundances of A'-like Synechococcus were highest at all sites and inputs of A-like, B'-like, and B-like sequences increased at progressively lower temperature sites. Chloroflexus-, Roseiflexus-, Armatimonadetes-, and Meiothermus-like sequences were found in much lower relative abundance in all sites.

(b) Octopus Spring. Like Mushroom Spring mat samples, Synechococcus and Chloroflexi-like sequences predominated at all sites (Fig. 6). As in Mushroom Spring mats, a progression from A'-like, A-like, and B'-like Synechococcus sequences was noted from sites OS1 to OS4, and a greater predominance of Chloroflexus over Roseiflexus at high-temperature sites and of Roseiflexus over Chloroflexus at low-temperature sites was observed. Thermus sequences were found in the mat at OS site 1; Armatimonadetes-like sequences were found at sites OS1, OS2, and OS3; and “Ca. Thermochlorobacter”-like and Chloracidobacterium-like sequences were found at sites OS3 and OS4.

FIG 6.

16S rRNA profiles for bacterial phyla in Octopus Spring water (A) and mat (B) collected at sites along the flow path on 2 June 2011.

In contrast to Mushroom Spring water samples, Thermocrinis-like sequences were in greatest relative abundance in water collected at sites OS0, OS1, and OS2. Hydrogenobacter-like sequences were also abundant at these sites. Thermus-like sequences were present in much lower relative abundance. These populations reflect the presence of communities positioned upstream of the mat community, though it is interesting to note that Thermocrinis streamers were not observed in the effluent channel studied, despite their regular presence in the effluent channel closest to the source pool (Fig. 1B [slanted arrow] and 3A). Both Thermocrinis and Thermus also declined in relative abundance in downstream samples. Ralstonia-like sequences were also detected in the water samples at all sites. The presence of small amounts of Synechococcus-like (predominantly A'- and A-like) and Chloroflexus-, Roseiflexus-, and Armatimonadetes-like sequences in water samples from sites OS0 and OS1 suggest inputs from mats that occur in the shoulder region of the large source pool (see Fig. 1B), which is near the effluent channel we studied and is flushed regularly due to flow surges. The rise in relative abundance of Synechococcus sequences in downstream samples was consistent with the increase in Synechococcus cell counts as water flowed over the mat (Fig. 4D), progressing downstream from A' to A-like sequences.

(ii) Synechococcus putative ecotypes.

(a) Mushroom Spring. Results for psaA-based Synechococcus putative ecotype (PE) populations, which closely paralleled 16S rRNA gene distributions, are shown in Fig. 7A. The mats at sites MS1 and MS2 were predominated by A'-like Synechococcus PEs, whereas both A-like and B'-like PEs were abundant at site MS3, and B'-like PEs were the only PEs detected at site MS4. Although water samples collected at sites MS0 and MS1 typically contained low counts of Synechococcus cells (Fig. 4), PCR amplification yielded psaA sequences representative of those inhabiting the site MS1 mat (e.g., predominantly A'-like PEs) (Fig. 7A). The presence of some psaA populations that are more typically found at lower-temperature sites in the mat (e.g., PEs A1, A5, and B'8) is consistent with the inference made above that there had also been some input from mats in cooler side pools. At lower-temperature sites MS2, MS3, and MS4, the inputs to the overflowing water of Synechococcus cells residing in the mat beneath the flow could be clearly observed, especially in the case of B'-like populations at sites MS3 and MS4. However, A-like PEs remained in greater relative abundance in the overflowing water through site MS3. This indicates displacement of Synechococcus cells from the mat to the water as it flows rapidly through the narrow effluent channel, with the cells from the mat adding to the composition of cells at upstream sites.

FIG 7.

Relative abundances of Synechococcus putative ecotypes (PEs) predicted by ecotype simulation analysis of psaA sequences detected in mat and water samples at sites along the flow path in Mushroom Spring (A) and Octopus Spring (B) collected on 2 June 2011. PEs are color coded, as indicated by the key (dark blue, light blue, and green for A'-, A-, and B'-like PEs, respectively). PCR amplification of the mat sample collected at site OS3 failed.

(b) Octopus Spring. A similar trend was noted in Octopus Spring (Fig. 7B). The mat showed a progression of A', A, and B' PEs at sites OS1, OS2, and OS4, respectively. The OS0 water sample was collected from a site just upstream of the upper temperature limit of the mat in the effluent channel. However, the presence of Synechococcus PEs, and higher relative amounts of PEs normally found at lower temperature (e.g., A1, A5, A7, and B'9) is consistent with inputs from the mat in the shoulder region mentioned above. Otherwise, as in Mushroom Spring, the site OS0 sample contained mainly A'-like PEs, whose relative contributions declined as the water flowed over cooler mat sites, at which the relative abundance of A-like and then B'-like PEs increased.

Molecular analyses of postdisturbance recolonization.

In order to investigate the significance of water above the mat being biased toward upstream Synechococcus species, a disturbance experiment was conducted on a 58 to 62°C site in the Mushroom Spring mat. This experiment was conducted in July 1996, but samples had remained frozen and unanalyzed until 2011. The study site was similar to sites MS3 and MS4, as judged by similarity in taxonomic composition (i.e., dominance of B'-like and A-like Synechococcus PEs; compare Fig. 8 control and Fig. 7A mat sites MS3 and MS4). As shown in Fig. S4, an undisturbed control site was compared with sites disturbed by removal of the upper green layer of the mat containing Synechococcus cells. Recovery of Synechococcus populations at both undisturbed and disturbed sites was monitored by analyzing psaA sequences (and PEs predicted by ecotype simulation). In order to evaluate the role of selection of differently adapted species, recovery was monitored not only under ambient irradiance, but also under conditions in which UV light had been removed, or in which irradiance was reduced by 92%. The composition of the undisturbed control site was relatively stable over a 2-day period during which recovery of disturbed sites was monitored (Fig. 8A to C). However, in all disturbed sites there was a dramatic shift toward A'-like Synechococcus PEs, including PEs A'4, A'5, A'7, A'9, and A'16, none of which were detected in abundance in the undisturbed mat. Small numbers of A'-like sequences at day 0 and 1 are likely due to their presence in overlying water of collected mat samples. All of these A'-like PEs, except PE A'4, were also detected in the water overflowing the MS3 site in 2011, suggesting that erosion, dispersal, and deposition of Synechococcus cells that enter the flow from upstream sites are important to recolonization. Only PEs A'4 and A'9 were detected in samples collected from sites covered by a UV screen, and PE A'5 was only detected in samples incubated under a 92% light-reduction screen, possibly indicating that the fitness of different Synechococcus species is influenced by the light environment.

FIG 8.

Changes over a 2-day period (21 to 23 July 1996) under different light regimes of abundant Synechococcus putative ecotypes (PEs) in samples from an ∼58°C section of mat at Mushroom Spring. Compared are undisturbed sites (A to C) versus sites disturbed by removing the top green Synechococcus layer (D to F). Light regimes: ambient (A and D), removal of UV irradiance (B and E), and 92% light reduction (C and F). All A-like PEs were <1 to 2% in samples A to C and are not shown. Samples D to F show B-like PEs above (green) and A-like PEs below (blue). PEs are color coded, as indicated by the key (dark blue and green for A'- and B'-like PEs, respectively). Error bars represent the standard error of duplicate samples.

Modeling.

The mathematical model outlined schematically in Fig. S5 was developed to describe the transport of cells via conservation of mass in a two-dimensional domain. It computes cell concentration in the downstream and vertical directions given empirically estimated stream velocity, cell counts, and measured cell dimensions. The model balances accumulation and the net mass transport of cells by advective transport in the stream-wise direction, gravitational settling (or buoyancy) in the vertical dimension, and turbulence in both the vertical and stream-wise directions. A more detailed description is provided in the Materials and Methods section and in the supplemental material.

(i) Application of the model to field data. Based on data in Table 1, we divided the Mushroom Spring flow channel into two regions, one upstream between sites MS0 and MS2, and another downstream between sites MS2 and MS4. With cell count data at MS0, MS2, and MS4, these two stream segments were modeled. Parameter values for modeling these segments are found in Table 2.

TABLE 2.

Model parameters computed for the Mushroom Spring effluent channel between the source, L₀ to L₃ upstream region, and the L₃ to L₇ downstream region^a

Parameter	Definition	L0 to L3 region	L3 to L7 region	Units
w (r = 10 μm)	Settling velocity of non-Synechococcus cells	−0.0005	−0.006	cm/s
w (r = 7.4 μm)	Depositional velocity of Synechococcus cells	−0.0033	−0.0405	cm/s
Kz	Vertical turbulent diffusivity	0.0352	0.0251	cm2/s
Kx	Horizontal turbulent diffusivity	0.0789	0.0562	cm2/s
v	Avg. stream velocityb	12.5	15.1	cm/s
H	Avg. stream depth	5	2.1	cm
L	Stream section length	380	930	cm
Width	Avg. stream width	18.5	38.8	cm
A1	Influent cell concn (non-Synechococcus)	72,848	396,997	cells/cm2
B1	Effluent cell concn (non-Synechococcus)	163,552	136,203	cells/cm2
A2	Influent cell concn (Synechococcus)	2,256	236,210	cells/cm2
B2	Effluent cell concn (Synechococcus)	97,312	11,651	cells/cm2

^aStream velocity and geometry (depth and width) are averaged from Table 1 over June and September. Values for K_x and K_z were computed using Fisher's formulas using the given H. Note that the turbulent diffusivities are different because they depend on v and H per equations, which have changed on this segment of the stream.

^bVelocity measurements were calculated from 18 to 66 trials, depending on time and location.

(a) Upstream region. By averaging the channel velocity and geometry data shown in Table 1 across June and September for sites L₀ to L₃, the channel could be approximated between sites MS0 and MS2 by a rectangular prism with a width of 18.5 cm, a length of 380 cm, and a depth of 5 cm, with an average stream velocity of 12.5 cm/s. Furthermore, at each end of this segment we have both total cell counts and counts of Synechococcus cells available, enabling the model to be solved for erosional rate E for Synechococcus or for the smaller cells predominating the total cell count. Note that to convert the measured cells/ml to the square units (cells/cm²) needed by the two-dimensional model, it was necessary to multiply this value by the average measured width of the stream.

(b) Downstream region. Similarly, by averaging the velocity and geometry data for sites L₃ to L₇ over June and September, the lower half of the channel between sites MS3 and MS4 could be approximated as a rectangle prism with a width of 38.8 cm, a length of 930 cm, and a depth of 2.1 cm, with an average stream velocity of 15.1 cm/s. Again, applying measurements for both types of cells allowed the model to be solved for two more erosion values.

(ii) Model predictions. Substituting these values into the model to be solved for erosion rate E as described above, cell concentration plots for each of these cases were generated and are shown in Fig. 9. Total cells and Synechococcus cells both increased along the upstream segment of the stream (Fig. 9A and B, respectively), which suggests that the net erosional value E should be positive as cells are eroded into the flow. Indeed, in this case E = 2.65 × 10³ cells/(cm s) for non-Synechococcus cells and E = 3.09 × 10³ cells/(cm s) for Synechococcus. In the downstream segment of the stream, both total counts and Synechococcus cell counts decreased between sites L₃ and L₇ (Fig. 9C and D, respectively). Thus, a negative erosional value E is expected on this stretch where there is net deposition of cells. Indeed, here E = −8.84 × 10³ cells/(cm s) for non-Synechococcus and E = −7.78 × 10³ cells/(cm s) for Synechococcus.

FIG 9.

Model output showing vertical distributions of non-Synechococcus cells (A and C) or Synechococcus cells (B and D) between sites L₀ and L₃ (A and B) and sites L₃ and L₇ (C and D), which are stretches of the Mushroom Spring effluent channel. Color bar indicates cells/ml.

DISCUSSION

This study brings together a variety of observations made in two similar well-studied Yellowstone hot spring cyanobacterial mats (Octopus Spring and Mushroom Spring) that were made during summer months in 1968 (see reference 1 and the supplemental material), 1995 (see reference 24), and 1996 and several months during the summer of 2011 (this study). Numerous kinds of empirical observations were made and a physical model was developed and employed. All of these observations point toward the progressive erosion of Synechococcus species in upstream mat regions and their deposition in downstream regions as water flows above the mat community. A resultant bias in the water toward species adapted to higher temperatures leads to their deposition in downstream regions. The recolonization of disturbed sites by species adapted to high temperatures causes a lowered rate of photosynthesis, as these predominant recolonizing species are less active at lower temperatures.

The estimates of Synechococcus cell counts presented here were of similar order of magnitude, though somewhat lower, than those reported by Brock (1). We believe this is likely due to differences in sites at which these measurements were made (see the supplemental material). Our observations were extended by consideration of all cells in the flow, which were several fold more abundant than Synechococcus cells, and which showed similar trends (see below).

Molecular analyses provided a better understanding of the nature of the cells above the mat at different sites along the effluent channel. In general, the prevalence of Thermocrinis and Thermus resembled the results of Boomer et al. (23) for Fairy Spring. The presence of Thermocrinis-like and Hydrogenobacter-like 16S rRNA gene sequences in Octopus Spring water correlates with the hotter source pool and the presence of pink streamers comprised of these bacteria in this spring (3, 5). In contrast, the prevalence of Thermus-like 16S rRNA gene sequences in Mushroom Spring water suggests that these organisms may be in greater abundance at higher temperatures in the mats that line its 68°C source pool and/or in the vent itself. Although the Mushroom Spring source-pool mat was not analyzed in our study, it is interesting that site OS1 mat samples contained the highest relative abundance of Thermus-like sequences. Nold and Ward (29) reported Thermus species with different temperature adaptations as low-abundance inhabitants of the Octopus Spring Synechococcus mat. This was confirmed in the present analysis, as Thermus was present at high temperatures and low-temperature adapted Thermus relatives, which have been reassigned to the genus Meiothermus (30), were detected in downstream Mushroom Spring water samples. The observation of Ralstonia in water samples from all Octopus Spring sites was novel. The erratic pattern of its distribution along the thermal gradient may suggest it was the result of contaminated PCR reagents (31).

The influence of Synechococcus mats on water flowing above them is clear. Synechococcus cell counts increased in samples collected below the upper temperature limit of these mats. Furthermore, the A'-, A-, B'-, and B-like Synechococcus species detected in water overflowing the mat also reflected the downstream progression of species of Synechococcus inhabiting the mat itself. This led to a biased representation of cells from upstream regions in the water overflowing a particular site in the downstream mat. Similar results were observed for samples collected on 15 December 2011. This pattern must occur because of erosional processes as water flows over the mat. To understand this, we developed a model that comprehensively took into account the physical processes that control erosion of cells from the mat, as well as the flux of cells and their redeposition downstream. Indeed, the model predicted net erosion of cells at upstream sites and decreased Synechococcus cell counts at downstream sites. Furthermore, these observations are consistent with the results of recolonization experiments conducted in late-July 1996 at a 58 to 62°C site in Mushroom Spring, as observed in this study, as well as with results in which A'-like Synechococcus 16S rRNA genotypes normally found at higher temperatures were found to predominate in the mat recovering from disturbance at 55 to 62°C sites in experiments conducted in June and July 1995 at Octopus Spring (24). The results presented here are more extensive because the use of a more rapidly evolving gene (psaA) and the ecotype simulation model permitted us to examine recolonization in a more refined way. As mentioned above, the consequence of a shift in PEs optimized for higher temperatures to lower-temperature downstream sites is a significant reduction in productivity. The use of psaA gene sequences and ecotype simulation also permitted us to examine the importance of different light conditions on selection of PEs adapted to different light regimes, along with cell transport in recolonization. Lowered light levels were included because the existence of low-light adapted Synechococcus populations has been reported (15, 18, 32). The detection of PE A'5 only under low-light conditions might suggest that this PE has enhanced fitness at low light. We also tested the effect of lowering UV light with screens, since UV-sensitive Synechococcus populations have been reported (33). These conditions had greater impact on which populations were most important in recolonization, as only PEs A'4 and A'9 were detected under these conditions. This might suggest that the fitness of these species was enhanced by UV removal, consistent with increased photosynthesis under these conditions observed by Miller et al. (33). In that study, Synechococcus populations derived from 70°C Octopus Spring mat samples were shown to be either resistant to or sensitive to UV exposure, thus suggesting the existence of A'-like populations that have such adaptations. Interestingly, these same populations were able to conduct photosynthesis at 55°C, albeit at a lower rate than at 65°C or 70°C. We observed some evidence of a bias toward surface-associated populations in overflowing water in Mushroom Spring, where vertical distributions of PEs were measured at sites of temperatures comparable to those at sites MS2 and MS3 (18). In that study, PEs A1 and A7 were relatively abundant A-like PE surface populations in the mat at a site comparable to MS2, and PEs B'2, B'8, B'9, and B'12 were observed to be relatively abundant and surface-oriented at a site comparable to MS3 (34).

The net influx of cells from upstream does not appear to disrupt the distribution of predominant taxa and species in undisturbed mats. This is presumably because the lower fitness of cells normally predominating in upstream regions reduces their ability to establish in lower-temperature downstream regions, which are densely populated by cells adapted to those regions. For instance, growth rates of A'-like and A-like Synechococcus isolates are optimized for higher temperatures, whereas growth rates of B'-like and B-like Synechococcus isolates are optimized for lower temperatures at downstream sites (11, 12). At these downstream sites, B'- and B-like Synechococcus have higher growth rates than A'- and A-like Synechococcus, which are forced to grow at suboptimal temperatures and are thus less fit. Although upstream cells are likely to have difficulty competing with downstream cells, they are apparently still present at low levels in downstream mats. This likely explains why shifting samples from downstream (50°C) to upstream (65°C) sites led to a rise in A-like populations normally predominating at higher temperatures (35). For this to happen, A-like Synechococcus had to be present in low-temperature samples. In contrast, samples from the upstream site did not shift composition when incubated at the downstream site. This was presumably due to the lack of transport of cells from low-temperature regions to high temperature sites (i.e., against the stream flow) and a lower chance of survival for organisms adapted to low temperatures if distributed to temperatures above their upper limit. That is, the microbial “seed rain” is directional.

Modeling efforts on the dispersion of cells due to physical characteristics of the channel and flow showed that stream velocity is a crucial parameter for predicting cell deposition or erosion (B. Jackson and I. Klapper, unpublished). Of course, velocity is itself dependent on the flow rate of the hot spring and cross-sectional area of the stream if flow is conserved. Thus, the width and height of the channel are also important model parameters. Other parameters of importance, such as vertical mixing, critically govern how quickly cell concentrations in the flow homogenized the effects of erosion and deposition. However, vertical mixing is here modeled as turbulence-induced diffusion and thus also depends directly on velocity.

We were able to apply our model to upstream stations used by Brock (1) (see supplemental material). The outcome was similar to the model output for our upstream stations (Fig. 9B), with net erosion of cells (Fig. S10). Model estimates for settling rates were 5.9 × 10⁶ and 4.8 × 10⁶ cells/cm² at these upstream stations, similar to estimates in our study (see Table S3), and 1.7 × 10⁴ and 2.1 × 10³ times greater than Brock’s reported settling rates. The difference is presumably due to the fact that erosion had not been taken into account in that study. The higher estimated settling rates were still lower that Brock’s estimated productivity rates, but suggest that cellular inputs could be on the order of 0.7 to 3.7% of mat production rates. Although there was not sufficient information about channel dimensions to model Brock’s downstream regions, his Synechococcus cell counts continued to rise (Fig. S9), consistent with greater flow velocity and likely narrower downstream channels.

The lack of a decline in total cell counts and the much lower Synechococcus cell counts in Octopus Spring water might be consequences of the ebb and flow of the effluent, which may even cause temporary upstream flow. This might be a mechanism for resuspension of cells. Also, periodic decrease in depth associated with ebbing might promote deposition of cells, especially the larger Synechococcus cells.

In conclusion, our results demonstrate the importance of physical realities associated with flow to both erosional and depositional processes in the mat. The rise and fall of Synechococcus cell counts in water flowing over the mats is consistent with model predictions of erosion at upstream sites and deposition at downstream sites. The bias toward the prevalence of upstream Synechococcus species in the overflowing water is also consistent with upstream erosion and with the colonization of disturbed downstream sites by Synechococcus species normally found upstream. Although the net deposition of cells adapted to high temperatures in downstream regions is small in comparison to mat productivity, it does contribute to community complexity in the form of “seed rain.” It becomes of paramount importance in affecting the downstream colonization of disturbed sites, as it leads to significant decreases in mat productivity during recovery.

MATERIALS AND METHODS

Sampling sites and collection times.

Five collection sites were selected along the temperature gradient of one major effluent channel each for both Mushroom Spring and Octopus Spring (Fig. 1). The sites at Mushroom Spring were located in the main channel at approximately 0, 0.25, 3, 6.85, and 13.1 m down channel from the source and were named MS0 to MS4, respectively. The sites at Octopus Spring were located in the southern-most effluent channel at approximately 0, 1.1, 3.8, 5.5, and 13.1 m down channel from the large pool and were named OS0 to OS4, respectively. Because sampling sites were chosen to sample certain temperatures, not all were suitable for stream velocity measurements. To compensate for this, additional sites for velocity calculation (L₀ to L₇) were selected in Mushroom Spring, some selected to bracket the sampling sites, in order to estimate flux into or out of the sampling zones. Specifically, site L₂ was located between sites MS1 and MS2 at approximately 2.5 m from the source; site L₄ was located between sites MS2 and MS3 at approximately 4.7 m from the source; and site L₆ was located between sites MS3 and MS4 at approximately 9.3 m from the source pool.

Parameters measured in the field.

(i) Temperature. Temperature data were recorded using DS1922T Temperature Logger iButtons with 8KB memory (Maxim Integrated, San Jose, CA). An iButton was left in the channel at each collection site, and an additional iButton was used to monitor air temperature. Each iButton was programmed to record a point temperature reading every 20 min between 1 June 2011 and 30 September 2011. To monitor the temperature on a much finer scale, from 18 to 21 September 2011, one iButton was left in each channel (near sites MS1 and OS2, respectively) and was programmed to take a temperature reading every minute during that 3-day period.

(ii) Irradiance. Light (mol photons m⁻² s⁻¹) was measured with an LI-190SA quantum sensor (which measures photosynthetic active radiation [PAR] in the 400 to 700 nm wavelength range) and data were recorded with the LI-1400 Datalogger (LI-COR, Lincoln, NE). Light measurements were taken every 5 min and the average of these measurements over every 30-minute period was recorded from 1 June 2011 to 30 September 2011. To support the aforementioned fine-scale temperature readings, light was also recorded every minute over the same 3-day period from 18 to 21 September 2011.

(iii) Flow velocity. Estimates of flow velocity were made at 5 locations in Mushroom Spring in regions that exhibited a fairly regular channel shape and, as much as possible, uniform bed depth (Table 1). Selected regions were demarcated with string stretched across the channel and secured at an upper and lower location a known distance apart. Brightly colored fishing floats of semineutral buoyancy were introduced into the channel above the first string, and the transit of the demarcated region by the float was filmed with a tripod-mounted digital camera. As the float left the measurement region it was captured for redeployment. The resulting movies were analyzed frame by frame. For each transit, the float’s entrant and exit frame number was tabulated and the total transit frame count was calculated. This was then multiplied by the known frame rate (1/30 s) for a total transit time. Dividing the measured distance between the upper and lower strings by this time gave an average transit velocity.

Channel cross-sectional areas at each site were measured to compute flux using averages of multiple stream width and depth measurements (Table 1).

Sample collection.

During each field visit (2 June, 11 August, and 21 September 2011) mat and water samples were collected from approximately the same spatial locations in the channel (verified by photographs and conspicuous features in the mat or along the edge of the flow channel).

(i) Water samples. Two sets of water samples were collected at all sites. Syringes (60-ml) were used to pass 100 ml of channel water through a 0.2-μm Whatman filter using a Swinnex apparatus (Millipore, Bedford, MA) to hold the filter. The filters (in the Swinnex) were immediately frozen on dry ice and stored in a −80°C freezer until analyzed. One set of filters was used to obtain microscopic cell counts and the other set was used for molecular analyses.

(ii) Mat samples. Mat samples were collected at sites 1 to 4 in each channel with a number 2 cork borer (19.6 mm²), immediately frozen on dry ice, and then stored at –80°C until analyzed.

Microscopic analyses.

(i) Staining. Filters contained in Swinnex devices were allowed to defrost for 30 min at room temperature. Then 10 ml of a 1× SYBR Gold stain solution (Invitrogen, Eugene, OR), prepared by diluting a 10,000× stock solution in autoclaved dH₂O, was placed above the filter and allowed to stand for 30 min. The staining solution was then passed through the filter, which was dried by passing air through the filter and the dried filter was mounted on a slide for microscopy.

(ii) Microscopy. A Zeiss microscope (Axioskop 2 plus with an HBO 100 UV lamp) was used to obtain data for cell counts. A rhodamine optical filter combination was used in order to distinguish Synechococcus cells, by exploiting their red autofluorescence due to the presence of chlorophyll a. All cells containing DNA, and hence stained by the SYBR Gold stain, were excited using a fluorescein isothiocyanate (FITC) filter combination. For each filter, five viewing fields at 100× magnification were randomly chosen and photomicrographs were taken and later used to obtain digital cell counts. The raw microscope image data were imported into ImageJ (36) for analysis.

(iii) Cell counts. Images taken using the rhodamine filter, showing only Synechococcus cells in a given viewing field, were counted manually using a grid imposed over each image and ImageJ's “cell counting” plugin. The resulting counts were exported to a spreadsheet and converted to cells per ml of filtered water. Images taken using the FITC filter, showing all cells in a given viewing field, generally contained large numbers of cells and thus made manual counts impractical. Instead, an ImageJ script was applied to each image. This script applied a threshold to each image using the Li algorithm (37) and then measured the signal area using the automated “analyze particles” plugin. Total signal pixel counts were recorded and exported to a spreadsheet and again converted to cells/ml. In the case of the image taken using the FITC filter, an estimation of 432 sq pixels per cell was used to convert signal area into a cell count estimate (see the supplemental material).

DNA extraction and molecular methods.

Cells filtered from water samples were lysed directly from the filters by bead-beating as described in Becraft et al. (17). DNA was extracted from these and whole-mat samples with the FastDNA spin kit for soil (SKU 116560200-CF) according to the manufacturer’s instructions. A region of the 16S rRNA gene was amplified using 28F and 519R universal primers, and amplified DNA was sequenced at the Research and Testing Laboratory, Lubbock, TX. All 16S rRNA gene sequences were classified using RDP 11 (38) and CREST (39), and Synechococcus sequences were compared to those of known mat inhabitants for identification of Synechococcus genotypes (e.g., A', A, B', and B). The psaA locus was PCR amplified and sequenced with the ‘centerforward primer’ using Ti454-barcoding technology at either the J. Craig Venter Institute (mat samples) as described in Becraft et al. (18), or at the Research and Testing Laboratory (Lubbock, TX) (filtered water samples). Although deeper sequencing methods have become available more recently, at the time we conducted this study, Ti454 sequencing was the best method available. Sequences were trimmed to 302 bp to obtain the maximum number of nucleotides for analysis, cleaned, and analyzed to identify high-frequency sequences (≥50 identical copies in all samples combined) and associated low-frequency sequences (<50 copies in all samples combined) (18). The site OS3 mat sample failed to PCR amplify and is not included in psaA analyses. Sequence files are available at NCBI (see “Data availability” below).

The Ecotype Simulation model (19) was used to separately demarcate PEs of A-like (fine-scale demarcation method) and B′-like (conservative demarcation method) Synechococcus, as described in Becraft et al. (18). Since the A′-like lineage was more divergent than the A-like lineage, A′ PEs were demarcated using the conservative approach. Percentage of population of each PE in a sample was calculated by dividing the total number of sequences within a PE detected in a sample (PE sequences = all high-frequency sequences + all low-frequency sequences) by the total number of sequences in that sample ([PE sequences in sample/all sequences in sample] ×100). Only PEs that were present at ≥5% of the total in a sample are shown in the figures.

Recovery after disturbance.

On 25 July 1996, a 58.6 to 62.2°C site in the microbial mat of Mushroom Spring was covered with a wire mesh screen supported by a wooden platform placed ∼1 to 2 cm above the water surface to protect the experimental area from possible hail damage (Fig. S4). The area beneath the screen was divided into 6 separate sections, each of which corresponded to a different environmental condition as follows: (i) undisturbed mat (control); (ii) scraped (removal of top green layer); (iii) UV-blocked; (iv) ∼92% light reduction; (v) scraped and UV-blocked; and (vi) scraped and ∼92% light reduction. To examine the effects of light alteration on recolonization, one section was covered with 4 layers of stretched muslin to reduce downwelling photon irradiance by ∼92%, and another section was covered with UV-blocking plexiglass (UF-5; 3.2-mm thick; Plexiglass; Autohaas N. America, Philadelphia, PA, USA). To analyze the recolonization of newly forming mat after disturbance, the upper green mat layer was removed as completely as possible in 3 separate 7 × 10 cm areas using a bent spatula, with minimal removal of the red undermat. For the light alteration experiments, the scraped areas were located beneath the center of each light cover. Samples were collected in duplicate daily from each of the six sections from 21 (day zero) to 23 July 1996 using a number 4 cork borer (38.5 mm²) every day between 1200 and 1800 h. Samples were immediately frozen on dry ice (–20°C) in the field and kept frozen at –80°C until DNA extraction and sequence analysis in 2011.

Cell transport model.

The mathematical model summarized here describes the transport of cells via conservation of mass in a two-dimensional domain (Fig. S5). Cell concentration c is computed in the downstream direction, denoted by x, and in the vertical direction, denoted by z, measured from the bottom of the stream. Accumulation is balanced by the net mass transport of cells by advective transport in the stream-wise direction, gravitational settling (or buoyancy) in the vertical dimension, and turbulence in both the vertical and stream-wise direction. Turbulent transport is treated as turbulent diffusion, which treats turbulent flow as a random walk analogously to Fick’s Law of molecular diffusion (40, 41). A more detailed treatment of turbulence calculations is given in the supplemental material. Using v to represent average stream velocity, w settling velocity due to gravity, and K_x and K_z to represent turbulent diffusion, this balance may be written as:

$$\frac{\partial c}{\partial t}=\frac{\partial }{\partial x}\left(K_x\frac{\partial c}{\partial x}\right)-\frac{\partial }{\partial x}(\mathit{vc})+\frac{\partial }{\partial z}\left(K_z\frac{\partial c}{\partial z}\right)-\frac{\partial }{\partial z}(\mathit{wc})$$ (1)

Influent and effluent boundary conditions can be set to match measured concentrations or, in the case of the effluent boundary, relaxed to specify the rate of change of cells in the flow (typically zero when “no flux” is assumed). Upper and lower boundary conditions are derived by a consideration of flux balance. At the air-bulk fluid interface, cells may settle due to gravity or be drawn down by turbulent transport. However, in this model, no cells enter the region from the air above the stream, and no cells escape into the air. These assumptions generate an upper boundary condition given by:

$$-K_z\frac{\partial c}{\partial z}(x,H,t)+\mathit{wc}(x,H,t)=0$$ (2)

Similarly, at the bottom of the water column, cells settle due to gravity at velocity w and may be transported by turbulence. These fluxes are balanced by E, a sediment erosion rate (42, 43), and the depositional flux. That is, flux to the water column at the bottom of the stream is the sum of contributions from depositional and erosional forces. Using w_d to represents the rate of deposition, or “depositional velocity” of cells (42, 44), this lower boundary condition is written:

$$-K_z\frac{\partial c}{\partial z}(x,0,t)+(w-w_d)c(x,0,t)=E$$ (3)

A common assumption is that w_d = w, but this need not be the case. The parameter E is unknown and is computed by the model. Further details of this calculation, along with simplifying assumptions and a complete statement of the model, are available in the supplemental material.

Data availability.

Sequence files are available at NCBI under accession numbers SAMN15061381 to SAMN15061398 (16S rRNA) and SAMN15086850 to SAMN15086900 (Synechococcus psaA).