Which physical factors affect depth-dependence of microbial cell abundance?
Both the porosity and the permeability decrease with increasing effective pressure (Fig. 4), and the correlation between the initial porosity and the initial permeability for fine-grained sediments is well defined (Fig. 7a). The increasing compaction of sediments with depth results in reduced porosity and permeability, which may correspond to lower tortuosity and lower pore connectivity. Increasing compaction may thus constrain the habitability of deep sediments for microbial communities (Prasad 2003; Rebata-Landa and Santamarina 2006). Compaction is also likely to influence the transport of water, nutrients, and secondary products of microbiological activity (Yakimov et al. 1997). Therefore, the positive correlation between cell concentration and in-situ porosity observed at Site C0020 suggests that mechanical sediment compaction was responsible for the low density of microbes at this site (Figs. 2a and 6a). Indeed, the positive correlation between cell concentration and in-situ permeability also supports this notion (Fig. 6b). The influence of sediment compaction on cell abundance would be more apparent in shallower environments than in deeper environments, as the variation in cell concentration is less constrained in the low in-situ porosity and in-situ permeability regimes.
Why do coal-bearing units have anomalously high microbial biomass?
At Site C0020, the coal-bearing sections in unit III and the lower portion of unit IV had relatively high microbial cell concentrations. These units/sections were characterized by having relatively permeable coal and sandstone layers with a relatively large pore size (Fig. 5b, c). The changes that occur with increasing depth, which are mainly due to variations in the lithology, are larger for the permeability and the pore size than for the porosity. Consequently, because of the large variation in lithology at Site C0020, there was no clear correlation between the porosity and the permeability below 1000 mbsf (Fig. 7a).
The large pore size determined by MICP for coal likely reflects the existence of voids in the fractures (cleats). Consequently, the high permeability of coal at low effective pressure likely has a marked influence on both vertical and bedding parallel flow through cleats. The permeability of coal is very sensitive to the effective pressure, and the in-situ coal permeability can be very different to the initial permeability. Therefore, the in-situ pore size (aperture size) for coal is expected to be much smaller than the measured pore size (Fig. 7b) Nevertheless, voids in fractures may form an important pathway for fluid flow through coalbeds, because the in-situ permeability of fractured coal is much higher than that of the intact coal matrix or neighboring siltstone/shale formations (Fig. 5b, Ijiri et al. 2017). Therefore, it is considered that nutrients and energy sources were mainly released from the coal layers through the cleats, where large amounts of energy sources are stored, and that these compounds seeped into the overlying permeable sandy sediments. However, the impermeable shale and siltstone layers, which overlie the permeable sandstone layers, may act as barriers to vertical fluid flow. As shown in Fig. 5b, the permeability change between permeable (sandstone, coal) and impermeable (siltstone, shale) layers was two to four orders of magnitude in coal-bearing units. Since almost no nutrient flow occurs through the impermeable layer, the very low cell densities at 1500 to 1800 mbsf and from 2000 to 2400 mbsf (< 100 cells/cm3) can be explained by the effectiveness of the siltstone barrier for preventing the vertical transport and dispersal of nutrients and cells. Permeable layers may sustain more abundant active microbial populations than low-permeability layers, and may be one of the reasons for the higher cell numbers in coalbed layers (Inagaki et al. 2015). Therefore, in deeper sections, pore size and permeability, rather than porosity, are considered to have a greater influence on cell abundance (Phadnis and Santamarina 2011).
The in-situ permeability was estimated from the effective pressure dependence of the permeability, assuming that the pore pressure is hydrostatic in deep regions. However, overpressure (pore pressures higher than the hydrostatic pressure) occurs in deeper regions of most thick sedimentary/petroleum basins (Bredehoeft and Hanshaw 1968; Swarbrick and Osborne 1998) and in coal sediments (Law 1984; Su et al. 2003). Since siltstone and shale at Site C0020 form an impermeable barrier to fluid migration, overpressure in and below impermeable formations can easily occur due to inhibition of vertical fluid/gas transfer and hydrocarbon generation in coal units (Hunt et al. 1994, 1998; Osborne and Swarbrick 1997; Zhao et al. 2018). The permeability of coal is much more sensitive to the effective pressure than that of other rock types in the Sanriku-oki subbasin. Therefore, overpressure would be expected to increase the in-situ permeability and pore size for coal, so allowing an abundant microbial population to be sustained. Overpressure can also increase the methane sorption and storage capacity of coal (Scott 2002).
In the past, the shale layer may have supplied organic matter that was consumed by microbes in the adjacent sandstone layers (Fredrickson et al. 1997). However, because the shale deposits at Site C0020 currently have a much lower TOC content and a lower diffusion rate than those for coal, the nutrients that support indigenous microbial communities in sandstone and coal are considered to be derived from lignite coal. Cleats and fractures in coal increase the substrate surface area available for microbial action (Strąpoć et al. 2011). Since young lignite coal is susceptible to microbial attack, microbial coalbed methane would have been readily produced in the coalbeds of the Sanriku-oki basin.
Influence of pore size on sedimentary microbial biomass
Pore size can restrict both the size of microbial populations and the cellular volume in porous sediments (Fredrickson et al. 1997). At Site C0020, the modal pore size for shale (0.028 ± 0.035 μm) and siltstone (0.044 ± 0.009 μm) adjacent to coal units is less than 0.2 μm (Fig. 5c), whereas the pore size for sandstone and coal is > 2 μm (fracture aperture size of > 1 μm). A pore size of 0.2 μm for lithologies with low in-situ permeability is consistent with the proposed threshold size for microbial metabolic activity (Fredrickson et al. 1997). The pore size for shale and siltstone is probably smaller than the majority of in situ microbial cells at C0020, since microbial cells in deep marine sediments are approximately 0.5 μm in size (Kallmeyer et al. 2012; Braun et al. 2016), and the microbial cell size estimated after incubating coal and shale samples from C0020 ranged from 0.3 to 1.2 μm (Trembath-Reichert et al. 2017). If the pore size is smaller than the cell size, then cell dispersal will not occur despite the presence of intense hydrological fluid flows, and microbial activity in such sediments will be low.
At Site C0020, it is conceivable that microbial cells were buried and trapped in shale and siltstone, promoting the agglutination of cells, and limiting the physical transport and/or migration of cells and eukaryotic spores from sandstone and coal layers to shale and siltstone layers (Inagaki et al. 2015; Gross et al. 2015; Liu et al. 2017).
Other potential geophysical constraints on deep subseafloor microbial biomass
In addition to the physical properties of sedimentary rock, other environmental factors also affect the microbial habitability of the deep subseafloor biosphere, e.g., temperature, pressure, water content, Aw, pore fluid chemistry, and pH (Beuchat 1974; Fernández Pinto et al. 1991; Zhang et al. 1998; Jenkins et al. 2002; Musslewhite et al. 2003; Grant 2004; Williams and Hallsworth 2009; Stevenson et al. 2015). These different geophysical parameters control the availability of nutrients and energy substrates required for the maintenance of essential cellular functions (e.g., biomolecule repair) over geological time (Hoehler and Jørgensen 2013; Lever et al. 2015).
In general, soil bacteria exhibit optimal growth at high Aw, but several bacteria have a high tolerance to desiccation (Aw < 0.8, Griffin 1972; Lavelle and Spain 2001) and highly saline water (Aw ≈ 0.61, Grant 2004). The Aw values at Site C0020A range from 0.954 to 0.983 and are almost unaffected by depth or lithology (Fig. 2e, Tanikawa et al. 2018). Since the Aw values are considerably higher than the Aw limit for the biosphere, water activity is not considered to affect the microbial population size or activity in the Sanriku-oki subbasin.
Although the temperature at Site C0020 increases almost linearly with depth (24.4 °C/km, Tanikawa et al. 2016), it is not clear whether such temperature increases limit the population size, activity, or microbial community structure in the deep subseafloor biosphere (Hinrichs and Inagaki 2012).
Permeability estimates based on pore characteristics
Our results indicate that differences in pore size are mainly responsible for the observed variation in initial permeability (Fig. 7b). To date, numerous attempts have been made to relate the pore size distribution and the porosity to the fluid permeability (Dullien 1992). The relationship between the initial porosity and initial permeability also indicates that the permeability characteristics of fine-grained sediments are governed by the traditional Kozeny-Carman equation (Fig. 7a). This equation is widely applied to mathematical models of microbiological processes (Taylor et al. 1990; Vandevivere et al. 1995) or prediction of permeability in well logging tools (Prasad 2003; Osterman et al. 2016). The formula describes the relationship between the permeability and the pore size, porosity, and geometry of porous sediments based on the assumption that the pore structure consists of a bundle of cylindrical capillaries (Dullien 1992). The Kozeny-Carman equation is expressed as follows:
$$ k=\frac{\varPhi {d}^2}{16f{\tau}^2}, $$
(5)
where k is the permeability, ϕ is the porosity, d is the diameter of a volume-equivalent spherical pore, f is a factor that accounts for pores with different cross-sectional shapes, τ is the tortuosity factor, and fτ2 is the Kozeny-Carman coefficient. In this study, the mercury intrusion porosity and representative pore size evaluated from the MICP measurements were used to calculate the permeability based on Eq. (5), where f was assumed to have a value of 2.
Poremaster software (Poremaster for Windows® 8.01, Quantachrome Instruments, FL) was used to calculate the tortuosity values from the MICP measurement results. The empirical formula introduced by Carniglia (1986) is based on Fick’s first law and a cylindrical pore structure model, and is expressed as follows:
$$ \tau =\left(2.23-1.13V{\rho}_b\right){\left[0.92\times \frac{4}{S}\sum \left(\frac{v_i}{d_i}\right)\right]}^{1+\varepsilon }, $$
(6)
where V is the total specific pore volume, ρb is the bulk density of the sediment, S is the total surface area of sediment particles, vi is the change in pore volume within a pore size interval i, di is the average pore size within a pore size interval i, and ε is the pore shape factor exponent (0 < ε < 1). A value of ε = 1, which corresponds to cylindrical pores, was used in this study. To estimate the permeability using Eqs. (5) and (6), the mean, modal, and median pore sizes were used (see Additional file 1: Figure S2 and Table S1).
In the low-permeability region (< 10−17 m2), the permeabilities estimated using the mean and modal pore sizes were in reasonable agreement with the measured permeabilities under a low effective pressure of 3 MPa (Fig. 9). In the higher permeability region, use of the mean pore size underestimated the permeability, while use of the modal or median pore size overestimated it (Fig. 9). The modal pore size was therefore considered to be representative of the fluid transport properties over a wide permeability range.
Fracture permeability model
Using the average pore size, the Kozeny-Carman equation (Eq. 5) did not accurately reproduce the measured permeability when the permeability was high (Fig. 9). Microstructural images of the coal material (Fig. 8, Core 15R3) showed that fluids could selectively flow through fractures; consequently, a fracture permeability model seems more appropriate than the capillary-tube-based Kozeny-Carman model. The well-known empirical equation for single fracture flow, known as the cubic law (Witherspoon et al. 1980; Raven and Gale 1985), is expressed as follows:
$$ {Q}_{fr}\kern0.5em =\kern0.5em \frac{b^3}{12R}\frac{\Delta P}{\upmu \mathrm{L}}\kern0.5em \left(1.06\kern0.5em <\kern0.5em R\kern0.5em <\kern0.5em 1.65\right), $$
(7)
where Qfr is the flow rate per unit width normal to the flow direction, b is the mean aperture size of the fracture, R is a fracture surface characteristic factor, μ is the fluid viscosity, and ΔP/L is the pressure gradient in the direction of the fluid flow. Here, we consider the cuboid coal samples that were used in the permeability measurements in this study. Fracture planes develop parallel to the sample length, L, and sample width, W, and normal to the H axis. If multiple parallel fractures developed parallel to the L and W directions, then the flow rate per unit area, q, measured in the permeability tests can be expressed as follows:
$$ q=\frac{Q_{fr}\times W\left({\uprho}_{fr}\mathrm{H}\right)}{W\times H}=\frac{b^3}{12R}\frac{\Delta P}{\mu L}{\uprho}_{fr}, $$
(8)
where ρfr is the fracture density (number of fractures per unit length normal to the fracture surface). Based on Eq. (2), the fracture permeability is then given by:
$$ {k}_f=\frac{b^3{\uprho}_{fr}}{12R}. $$
(9)
The μX-CT data for the coal sample (Fig. 8c) showed that b ranged from 1 to 10 μm, and ρfr was 1000 m−1. Therefore, assuming that R = 1, the coal permeability approaches 10−13 to 10−14 m2, which is more consistent with the experimentally determined permeability than that estimated using the Kozeny-Carman model. This result suggests that fluid transport within coal is mainly affected by the fracture (or cleat) geometry, rather than by the coal matrix, for which the permeability model consists of an assembly of circular pipes.