Introduction
The later stages of the Sun's main sequence evolution (2–3 Gyr from the present) will result in much higher surface temperatures on the future Earth and therefore, much more extreme environments for the last life able to grow and survive on the planet. This work builds upon the earlier work (O'Malley-James et al. Reference O'Malley-James, Greaves, Raven and Cockell2013) in which potential refuge environments for life on the future Earth were identified, in order to evaluate what remotely detectable biosignatures the Earth would produce as the Sun's main sequence evolution moves the inner edge of the habitable zone (HZ) outwards driving extreme climate change. The habitable lifetime of any planet is limited, the exact duration depending on the type of star (or stars) hosting that planet; hence, knowing the likely biosignature evolution of the dying Earth can inform us about the potential remote signature appearances of Earth-like exoplanets whose biospheres are dying. A habitable planet can be considered to become uninhabitable when it crosses the inner edge of the HZ, which is defined by the runaway greenhouse limit – the point at which the stellar flux reaching the top of a planet's atmosphere crosses a threshold value that triggers runaway heating. The most recent estimate places the inner edge of the Sun's HZ at 0.99 au (Kopparapu et al. Reference Kopparapu, Ramirez, Kasting, Eymet, Robinson, Mahadevan, Terrien, Domagal-Goldman, Meadows and Deshpande2013), corresponding to a threshold flux of 1.06 S0 (where S0 is the present solar flux). However, other work places higher limits on this threshold of up to 1.25 S0 (Goldblatt & Watson Reference Goldblatt and Watson2012; Goldblatt et al. Reference Goldblatt, Robinson, Zahnle and Crisp2013).
The death of the biosphere as we know it today begins with the extinction of higher plant species. Rising temperatures cause silicate weathering rates to increase, increasing CO2 draw-down, lowering CO2 levels in the atmosphere (see appendix). This results in conditions that are increasingly unsuited to (higher) plant life (Lovelock & Whitfield Reference Lovelock and Whitfield1982; Caldeira & Kasting Reference Caldeira and Kasting1992). During the CO2 decline, rapid ocean evaporation would not yet have begun. From Henry's Law, a reduction in atmospheric CO2 would lead to a reduction in the CO2 levels in the surface ocean, while increased silicate weathering could potentially lead to increased carbonate deposition. A reduced CO2 concentration would lead to an increase in ocean pH, although organisms should be able to tolerate more alkaline oceans. Cyanobacteria have developed CO2-concentrating mechanisms in response to the low affinity for CO2, and the low selectivity of ribulose-1,5-bisphosphate carboxylase–oxygenase (Rubisco), the enzyme involved in the first step of the carbon fixation process for CO2 over O2, which involves importing/accumulating inorganic carbon in the form of bicarbonates in the cytoplasm and converting this to CO2 (Koropatkin et al. Reference Koropatkin, Koppenaal, Pakrasi and Smith2007; Zarzycki et al. Reference Zarzycki, Axen, Kinney and Kerfeld2013). Those that live in environments with variable inorganic carbon (Synechocystis PCC 6803, for example) have high-affinity bicarbonate transporters that are used under low CO2 conditions using energy, ultimately from light, to accumulate CO2 inside the cells to values much higher than those in the external medium (Koropatkin et al. Reference Koropatkin, Koppenaal, Pakrasi and Smith2007). Such mechanisms, although with lower concentration factors (inside concentration/outside concentration) for CO2, occur in many eukaryotic algae (Raven et al. Reference Raven, Giordano, Beardall and Maberly2012). Mechanisms like this could allow oceanic photosynthesis to survive for longer than on land (Raven et al. Reference Raven, Giordano, Beardall and Maberly2012).
The continual decrease in CO2 levels eventually renders photosynthesis impossible for higher plant species, bringing an end to the age of plants. A corresponding decrease in atmospheric oxygen levels, coupled with the loss of primary food sources, would lead to the concurrent, sequential extinction of animal species, from large vertebrates to smaller ones, with invertebrates having the longest stay of execution (O'Malley-James et al. Reference O'Malley-James, Greaves, Raven and Cockell2013).
Life on Earth will then once again become microbial. Initially, while there is still sufficient atmospheric O2 and CO2 to fuel microbial metabolisms, this global microbial biosphere would likely be diverse, with maximum productivity values similar to those of the pre-photosynthetic Earth; 180–560 × 1012 mol C yr−1 (Canfield Reference Canfield2005). The actual productivity will probably be lower as primary productivity in the absence of photosynthesis involves chemolithotrophy, which would be subject to similar constraints on the supply of inorganic carbon as photosynthesis. When atmospheric oxygen reaches negligible levels, a rapid shift (within a few Myr) towards anaerobic metabolizers will occur. The possible microbial functional types that could grow under these conditions are discussed in Table 1. Chemolithotrophs, although they use a wider range of carboxylases than photosynthetic organisms, would be subject to similar carbon constraints to cyanobacteria because none of the carboxylases used have much higher affinity for inorganic carbon than the highest-affinity members of the family of photosynthetic carboxylases Rubisco (Raven et al. Reference Raven, Giordano, Beardall and Maberly2012).
Earth's surface becomes largely uninhabitable between 1.2 and 1.85 Gyr from present, depending on the latitude (O'Malley-James et al. Reference O'Malley-James, Greaves, Raven and Cockell2013), consistent with current estimates of 1.75 Gyr for Earth's habitable lifetime by Rushby et al. (Reference Rushby, Claire, Osborn and Watson2013). However, as surface temperatures across the planet continue to increase, lower temperature refuges where liquid water is still available will still exist beyond this time. In O'Malley-James et al. (Reference O'Malley-James, Greaves, Raven and Cockell2013) potential refuge environments in the subsurface, caves and at high altitudes were discussed, which could enable a biosphere to exist for up to 2.8 Gyr from present. This biosphere would favour unicellular, anaerobic organisms with a tolerance for one or more extreme conditions.
In this paper, changes in biosignatures as the biosphere declines, from the extinction of animals and plants to an increasingly less diverse microbial world, are modelled and their potential for remote detection discussed.
Methods
A simple atmosphere–biosphere chemical interaction model was used, combining the metabolic reactions listed in Table 1, abiotic gas fluxes and simple atmospheric chemistry. The model is linked to a surface temperature evolution model (shown in Fig. 1), described in O'Malley-James et al. (Reference O'Malley-James, Greaves, Raven and Cockell2013), where the surface temperature (T) evolution with latitude and altitude is modelled using
F in and F out represent the incoming and outgoing energy fluxes to/from the planet and C p is the heat capacity of the planet (at constant pressure). The model was run for 3 Gyr accounting for increased solar luminosity (L) using,
where t is the time elapsed on the main sequence, t e is the current age of the Sun and L s is the present-day solar luminosity. Long-term variations of eccentricity and obliquity are accounted for over the modelled time period as described in O'Malley-James et al. (Reference O'Malley-James, Greaves, Raven and Cockell2013). The modelled time period begins during the period in which all plants and animals become extinct (∼1 Gyr from the present; point A in Fig. 1). A global microbial biosphere is assumed thereafter until those organisms can no longer survive.
This biosphere is assumed to consist of microbial species using the metabolic processes listed in Table 1. An initial microbial abundance similar to the total of present estimates for the terrestrial subsurface (in this case defined as habitats below 8 m in depth) (2.5 × 1029 cells), soils (2.6 × 1029) and subseafloor sediment (2.9 × 1029) (Whitman et al. Reference Whitman, Coleman and Wiebe1998; Kallmeyer et al. Reference Kallmeyer, Pockalny, Adhokari, Smith and D'Hondt2012) was assumed. This microbial biosphere is initially evenly distributed over the surface of the Earth, giving an initial cell count of 1.5 × 1015 cells m−2. Splitting the planet into three latitude zones (polar (±60–90°), mid-latitude (±30–60°) and equatorial (0 − ±30°)) and assuming ∼1% of the surface area is at an elevation ⩾5 km gives the microbial abundance within the latitude zones assumed in the temperature model.
The microbial abundance in a particular region is then linked to the surface temperature evolution such that the count drops to zero when the temperature in a particular zone exceeds 420 K; an upper temperature bound for life (set to allow some increase over the currently known upper temperature tolerance of thermophiles) at which biological chemical processes break down. The total modelled cell abundance is shown in Fig. 2.
The chemical reactions occurring in a single cell for each type of microbial metabolism (see Table 1) are modelled by calculating the per cell rate of the gas production/consumption (see Table 2) for species representing the different microbial metabolic types using data from studies in the recent literature. By combining this information with the estimated microbial cell abundance, reactants are drawn from an atmospheric reservoir of gases and added to that reservoir, assuming the reactant use rates/product production rates described in Table 2. These rates are used to determine the production rates/reactant use rates in the metabolic reactions using the individual molar ratios of reactants to products in a given reaction. The limiting reactant (the one not in excess) determines the yield of products in each case.
The number of reactions taking place is determined by the temperature-dependent cell abundance calculated previously. The number of organisms using particular metabolisms is estimated based on the likely abundance of those metabolizers in the far-future environment. This is calculated from the total initial cell abundance estimate (8 × 1029 cells), assuming 1% are found at high altitude (>5 km) and an order of magnitude fewer aerobic metabolizers than anaerobic metabolizers due to the low oxygen levels on the far-future Earth (based on abundance differences between aerobes and anaerobes in low oxygen environments, e.g. Tartakovsky et al. Reference Tartakovsky, Manuel and Guiot2005; Ulloa et al. Reference Ulloa, Canfield, DeLong, Letelier and Stewart2012), giving initial total cell counts of 8 × 1028 cells and 7.2 × 1029 cells for aerobes and anaerobes, respectively. A ‘per-habitat’ cell count for anaerobes and aerobes is calculated from these values (assuming 1% at high altitude and an even split between the cave and subsurface environments for the remainder in each case) and the ‘likely habitat’ estimates from Table 1 are used to calculate an initial cell abundance for each of the modelled metabolisms (see Table 2). Other factors accounted for in the model (see Fig. 3) are abiotic gas fluxes (see Table 3), ocean evaporation rates (1.32 × 1015 g yr−1 after the onset of rapid ocean evaporation) and hydrogen escape into space. While hydrogen escape on Earth is presently limited by diffusion through the homopause, during the rapid ocean evaporation phase the temperature–pressure structure of the atmosphere is altered, allowing large amounts of water vapour to reach the stratosphere where it photodissociates, resulting in a hydrogen-rich atmosphere. During this phase, hydrogen escape is limited only by the solar extreme ultraviolet (UV) heating rate (which, for an ageing sun-like star, would be similar in intensity to the present-day flux from the Sun and would not vary significantly over the modelled time period – Ribas et al. Reference Ribas, Guinan, Gúdel and Audard2005); hence, a hydrodynamic escape rate of 3.06 × 1030 H atoms s−1 (Caldeira & Kasting Reference Caldeira and Kasting1992) is assumed for the hydrogen-rich phase. At other times a diffusion-limited rate is assumed.
The gas reservoir is then sampled at different time points (A, B and C in Fig. 1) and biosignature gases are compared with their present-day abundances and their detectability evaluated.
Results and discussion
Atmospheric biosignatures caused by the end of photosynthesis and the extinction of animals
Plants
The end of the higher plants will probably come at the end of a gradual decline in plant diversity as species become less able to survive under changing environmental conditions. The net primary productivity (essentially the difference between photosynthesis and plant respiration) is 104.9 Pg C yr−1, the terrestrial component of which is 56.4 Pg C yr−1 (Field et al. Reference Field, Behrenfeld, Randerson and Falkowski1998). At present, 500 Pg of carbon is locked up in terrestrial plant biomass (Mahli Reference Mahli2002); hence, at present rates of primary productivity, this biomass could be produced in <10 years. However, other loss processes result in the net increase in plant biomass being ∼1.4 Pg C yr−1 (Mahli Reference Mahli2002), giving an accumulation time of <500 years at present rates.
Decaying plant matter initially transfers carbon to the soil carbon pool (1500 Pg C; Mahli Reference Mahli2002), microorganisms and, in some cases, to animals. From here it is mostly released to the atmosphere via decomposition or metabolism (CO2 aerobically and CH4 anaerobically), with other gases produced as a result of converting some of the energy in plant biomass into a form usable in cell synthesis and maintenance, increasing decomposer biomass (Allison Reference Allison2006). Some carbon is transferred to oceans and geological carbon pools. Assuming similar carbon storage in the far-future biosphere, the death of all plants could release up to 500 Pg of carbon into the atmosphere via decomposition or animal metabolism over the time taken for plant species to decline. Higher temperatures are known to increase the rate of decay of dead plant matter into the soil carbon pool, with higher temperatures increasing the rate at which organic matter is broken down (Mahli Reference Mahli2002).
At present, 80% of the methane flux to the atmosphere comes from the decay of organic matter, estimated as 529–825 × 1012 g CH4 yr−1 (Ehhalt Reference Ehhalt1974). The total dead phytomass (not including soil organic carbon) from which this is produced is ∼184 × 1015 g; 15% of the estimated global total living phytomass, 1243.9 × 1015 g (Ajtay et al. Reference Ajtay, Ketner, Duvigneaud, Bolin, Degens, Kempe and Ketner1979). In the extreme case, in which the decline in plant life is very rapid, there would be an order of magnitude increase in methane flux until all dead phytomass has decayed. As methane has a relatively short atmospheric lifetime of ∼10 years (Kaltenegger et al. Reference Kaltenegger, Traub and Jucks2007) due to reactions with hydroxyl radicals and O(1D), there would be a more intense methane signature in Earth's atmospheric spectrum associated with this event (see Fig. 4), but only for a short time period. This period may be shortened by increased production of hydroxyl radicals as a result of the increased atmospheric water vapour content.
However, the plant extinction sequence is likely to be less abrupt. For example, larger species tend to be less tolerant to increased temperatures compared with smaller species. The CO2 uptake also varies for different plant types (Mooney Reference Mooney1972), suggesting that those with higher CO2 uptake rates would be more vulnerable to extinction in a declining CO2-environment.
Plants using the C3 pathway to fix carbon (the dominant pathway in higher plants) would be able to survive until atmospheric CO2 levels drop to 150 ppm; 0.5 Gyr from now (Lovelock & Whitfield Reference Lovelock and Whitfield1982; Caldeira & Kasting Reference Caldeira and Kasting1992; Raven et al. Reference Raven, Giordano, Beardall and Maberly2012).
Plant species may be able to continue to exist by becoming able to use lower concentrations of CO2 for photosynthesis (Raven et al. Reference Raven, Giordano, Beardall and Maberly2012) and to tolerate arid environments or by implementing other nutritional strategies, such as carnivory (Givnish et al. Reference Givnish, Burkhardt, Happel and Weintraub1984; Rice Reference Rice2002; Porembski et al. Reference Porembski, Theisen and Barthlott2006; Raven et al. Reference Raven, Beardall, Flynn and Maberly2009; Albert et al. Reference Albert, Jobson, Michael and Taylor2010; Flynn et al. Reference Flynn, Stoecker, Mitra, Raven, Glibert, Hansen, Granéli and Burkholder2013; Schmidt et al. Reference Schmidt, Raven and Paungfoo-Lonhienne2013) or mycoheterotrophy (Leake & Cameron Reference Leake and Cameron2010; Schmidt et al. Reference Schmidt, Raven and Paungfoo-Lonhienne2013). As a survival strategy, carnivory may be useful in certain low-nutrient ecosystems with intense solar radiation where liquid surface water is still abundant (Givnish et al. Reference Givnish, Burkhardt, Happel and Weintraub1984; Rice Reference Rice2002). These conditions would be more common near the beginning of the moist greenhouse stage as the planet begins heating up, e.g. loss of soil nitrogen in wet, high temperature conditions (although carnivory may be generally more related to phosphorus deficiency: Wakefield et al. Reference Wakefield, Gotelli, Wittman and Ellison2005). However, CO2 is still fixed from the atmosphere or freshwater bodies in carnivorous plants today with the smallest fraction of carbon obtained from photosynthesis rather than carnivory in terrestrial Utricularia species (Porembski et al. Reference Porembski, Theisen and Barthlott2006; Albert et al. Reference Albert, Jobson, Michael and Taylor2010) and almost all use the C3 pathway to fix carbon, which does not permit rapid photosynthesis in very low CO2 concentration, high-temperature environments. Some bromeliads (a few of which may be carnivorous: Givnish et al. Reference Givnish, Burkhardt, Happel and Weintraub1984) use crassulacean acid metabolism (CAM) to fix carbon, which helps to acquire and conserve carbon in arid environments. However, ultimately, as a strategy, carnivory would only give a temporary advantage as prey ultimately depend on plants for food. So as plant species decline, so too would prey.
Similar considerations apply to mycoheterotrophs, which are the plants obtaining some or all of their organic carbon from symbiotic mycorrhizal fungi, which in turn obtain their organic carbon from soil organic carbon or from fully photosynthetic plants with which the fungi are also symbiotic (Leake & Cameron Reference Leake and Cameron2010). About 10% of vascular plants have a juvenile mycotrophic and subterranean phase growing from (usually) very small spores or seeds which depends on mycoheterotrophy (Leake & Cameron Reference Leake and Cameron2010). While almost all of these ultimately become photosynthetic autotrophs, about 450 species never develop photosynthesis and live as mycoheterotrophs through their life cycle (Leake & Cameron Reference Leake and Cameron2010). Such plants would only have a short time to live after the last autotrophic plants in a habitat died, using soil organic carbon that is no longer being replenished.
Adaptations that conserve water such as the lower surface area per unit volume (as seen in cacti and other succulents, noting that this adaptation relates mainly to restricting water loss when photosynthesis using external CO2 is not possible) or increased reflectivity via the use of epicuticular wax, such as in the case of Dudleya brittonii, may be more widely adopted as temperatures increase (Jones Reference Jones2013; Sage & Zhu Reference Sage and Zhu2011). The drivers for these adaptations are not known, but the cause may have been a prolonged global dry spell which coincided with a global temperature drop and a fall in atmospheric CO2 (Arakaki et al. Reference Arakaki, Christin, Nyffeler, Lendei, Eggli, Matthew Ogburn, Spriggs, Moore and Edwards2011). Present-day C4 plants are more rapid photosynthesizers in dry, high temperature, CO2- and nitrogen-limited environments, suggesting adaptations like this may be employed as conditions for plants worsen.
The decline in plant species would lead to a lowering of atmospheric O2 and O3, which eventually fall to trace levels in only a few Myr once all plant life has disappeared (Walker Reference Walker, Schneider and Boston1991). Decaying plant life would produce methanethiol (CH3SH) (see later). Animal and plant extinction sequences will inevitably overlap. Those animals that are in food chains dependent on live plants will begin declining shortly after the plants at the base of their food chains disappear. Larger animal species could survive for longer in the oceans where phytoplankton would outlast land plants. Animals that are not dependent on live plants could survive for longer. Termites, for example, are able to digest dead wood due to a symbiotic relationship with microorganisms in their guts (Geib et al. Reference Geib, Filley, Hatcher, Hoover, Carlson, Jimenez-Gasco, Nakagawa-Izumi, Sleighter and Tien2008), whereas some isopod crustaceans are able to digest dead wood without the aid of symbiotic microbes (King et al. Reference King, Cragg, Li, Dymond, Guille, Bowles, Bruce, Graham and McQueen-Mason2010). Chemolithotrophically (chemosynthetically) symbiotic vetimentiferan worms (Riftia) found near hydrothermal vents may also be well suited to surviving longer than other animals, although these too would face extinction when the oxygen content of deep waters is depleted.
The decrease in atmospheric ozone would lead to less attenuation of DNA damaging UV radiation, particularly UVC radiation (200–280 nm). The increased UV surface flux would be further enhanced by the decrease in atmospheric CO2 levels, which, in the present-day atmosphere, scatter shorter wavelengths in the UVC range. This would lead to surface UV irradiation similar to that of the Archaean Earth, reaching ∼1 × 10−1 W m−2 nm−1 for longer UVC wavelengths (Cnossen et al. Reference Cnossen, Sanz-Forcada, Favata, Witasse, Zegers and Arnold2007) and a DNA-weighted dose up to a thousand times higher than on present-day Earth (Cockell & Raven Reference Cockell and Raven2007). If a similar irradiation environment to the Archaean is assumed, there are likely to still be environments where organisms could survive using DNA repair mechanisms or in shielded environments (Cockell & Raven Reference Cockell and Raven2007).
Land plants and oceanic phytoplankton produce significant amounts of the volatile compound isoprene (Seager et al. Reference Seager, Schrenk and Bains2012). Isoprene has a very short atmospheric lifetime, being rapidly oxidized. However, in an increasingly anoxic atmosphere, isoprene may be a plausible atmospheric biosignature. Oceanic phytoplankton would be better able to survive for longer than terrestrial photosynthetic organisms due to their low CO2 requirements and ability to extract carbon from bicarbonate (Raven et al. Reference Raven, Giordano, Beardall and Maberly2012). Therefore, if these organisms release isoprene at a rate similar to the present-day rate (0.31–1.09 Tg C yr−1 – Gantt et al. Reference Gantt, Meskhidze and Kamykowski2009), this could allow a detectable level of isoprene to build up in the atmosphere, although the photodissociation of the increased atmospheric water vapour content could provide enough free oxygen to keep the lifetime of isoprene short (it has a lifetime of <1 day in an oxygen-rich atmosphere – Palmer et al. Reference Palmer, Jacob, Fiore and Martin2003).
Animals
Decaying animals (as well as plants) release CH3SH, a gas that can be spectrally inferred and has no known abiotic source. The extreme case of the extinction of all animals (vertebrates ⩾ invertebrates) over a few Myr timescale could lead to a large release of CH3SH into the atmosphere. Assuming a biomass of plants and animals equivalent to present day values (560 × 109 tonnes C – Groombridge & Jenkins Reference Groombridge and Jenkins2002) and cell sulphur contents of 0.3–1% (Pilcher Reference Pilcher2003), this allows for the potential release of 1.12 × 1015 g of CH3SH (via conversion from methionine). If this were released at the current biological production rate (3 × 1012 g yr−1) it would all be released to the atmosphere within 350 years. CH3SH itself would not be a useful biosignature as it has a very short atmospheric lifetime, being readily photodissociated. However, the dissociated methyl groups combine to form Ethane, which would have a much longer atmospheric lifetime under the increasingly anoxic conditions at the end of the age of animals, making it a potential biomarker under these circumstances (Domagal-Goldman et al. Reference Domagal-Goldman, Meadows, Claire and Kasting2011). However, as with plant extinctions, the extinction of animal species is likely to occur over a much longer (10–100 s Myr) timescale, which could lead to a much weaker biosignature associated with the event, e.g. a 100 Myr extinction time would lead to a smaller flux of ∼107 g CH3SH yr−1.
Chemolithotrophs in far-future refuge environments
High altitude
Temperatures at high altitudes could remain low enough to permit liquid water pools for longer than at surface levels, providing refuges for life as surface temperatures increase. Additionally, any hot spot volcanoes (up-welling plumes from the deep mantle that occur in the middle of plates rather than at plate boundaries) that are present on the far-future Earth would provide refuges for simple microbial communities. Recent findings from a site at high altitude (>6000 m) in the Atacama Desert show that life in such environments uses the oxidation of carbon monoxide, ammonia or dimethyl sulphides to obtain energy (Lynch et al. Reference Lynch, King, Farías, Sowell, Vitry and Schmidt2012), producing various by-products such as CO2, N2, SO2, methane sulphonate and sulphates (see Table 1 for reactions). For oxidation reactions to be feasible, localized O2 sources would be needed for such communities to exist in far-future refuge environments. A lesser energy gain per mol of substrate oxidized could also be obtained with other oxidants such as ferric iron and nitrate, but the availability of these tends to decrease with oxygen availability. As O2 levels are reduced, iron will tend to be in a reduced state, making iron oxidation more favourable in a declining biosphere. Life is also known to exist around volcanic steam vents, for example, McMillan & Rushforth (Reference McMillan and Rushforth1985) reported the presence of many species of diatom living in condensing steam around a steam vent near Kilauea Crater, Hawaii (as these are photosynthetic species, a sufficient CO2 flux from the volcanic vent would be needed to support similar communities on the far-future Earth). Water vapour is the most abundant gas released via volcanic activity, which may mean that high altitude volcanic vents provide oases on the dry future Earth as steam condenses onto lower temperature surfaces.
There are over 1500 subaerial volcanoes that have been recently active, i.e. within the past 10 000 years (Siebert et al. Reference Siebert, Simkin and Kimberly2010). Only ∼49 volcanic sites are found at volcanic hotspots (Courtillot et al. Reference Courtillot, Davaille, Besse and Stock2003). Comparing hypsometric curves for Earth, Venus and Mars shows that, while there are clear differences in the surface area–elevation relationships for each planet, in all cases only <1% of the surface area is found at elevation above 5 km.
Caves
Microbial life in cave environments on Earth today consists of chemolithotrophically based metabolizers and those that make use of organic carbon made available from photosynthesis, often being found in microbial mat communities. Life in far-future cave refuges is more likely to be chemolithotrophic due to the absence of photosynthetic carbon sources. An example of a cave ecosystem entirely based on chemosynthesis is found in Movile Cave, Romania. Here, life has been separated from the outside world for 5.5 Myr and has similarities to hydrothermal vent communities (Chen et al. Reference Chen, Wu, Boden, Hillebrand, Kumaresan, Moussard, Baciu, Lu and Murrell2009). The cave receives a high flow rate of hydrothermal water, which is rich in sulphides and methane; hence, life in the cave is mainly based on sulphide and sulphur oxidation and methane oxidation. This produces by-products such as sulphuric acid, iron III (ferric) sulphate and CO2 (see Table 1).
However, sulphur, sulphide and methane oxidation would only be possible to a limited extent on the largely anoxic far-future Earth. Sulphate and anaerobic methanotrophs have been found together in anoxic marine sediment environments (Orphan et al. Reference Orphan, Hinrichs, Ussler, Paull, Taylor, Sylva, Hayes and Delong2001), suggesting a potential alternative for far-future cave ecosystems, although this would not provide as much energy as oxidation reactions. Anammox bacteria, known to be high-temperature tolerant (Byrne et al. Reference Byrne2009), could potentially produce a hydrazine biosignature (assuming gas exchange with the surface) as hydrazine is a free intermediate in the anammox reaction (Strous et al. Reference Strous2006).
The deep subsurface
The deep subsurface generally refers to depths of 50 m or more below the continental surface or ocean floor. The maximum inhabited depth is, at present, unknown – the deepest microbial life found to date was found at a depth of 5.3 km in igneous rock aquifers. Microbial subsurface life could use anaerobic methane oxidation, ferric iron reduction or sulphate reduction to metabolize (see Table 1). Hydrogen-driven communities may also be present, although there is some debate as to whether enough abiotic hydrogen (produced by the radiolysis of water (Blair et al. Reference Blair, D'Hondt, Spivack and Kingsley2007) or reactions between dissolved gases in magma, for example) could be produced to support them (Reith Reference Reith2012). Originally it was assumed that subsurface life would be predominantly methanogenic; however, recent work suggests that this may not be the case due to the higher than expected counts of sulphate reducers found, suggesting a syntrophic biosphere of chemolithoautotrophs or organotrophs with fermenting bacteria (Teske Reference Teske2005). Life in the deep subsurface would experience higher temperatures than surface dwelling life, as temperature increases with depth due to the geothermal temperature gradient of the Earth's crust. This means it is likely that this would be the first group of microorganisms on the dying Earth to disappear.
Detecting the presence of subsurface life may be more challenging than detecting life in other habitats. While life in the shallow subsurface could produce waste gases such as methane, which could percolate out into the atmosphere through rock or soil, deeper life would require the presence of deep faults in rocks for biomarker gases to escape into the atmosphere (Stevens Reference Stevens1997). However, in some cases it could be possible to detect subsurface life indirectly by looking at the surface features of the environments that support it (Hegde & Kaltenegger Reference Hegde and Kaltenegger2013). Temperature below the surface is determined by the surface temperature and increases with depth. The Earth's geothermal temperature gradient varies depending on rock type, but on average it increases by ∼30° km−1 depth (Bohlen Reference Bohlen1987). Although this mean temperature gradient would preclude life once surface temperatures exceed the maximum temperature tolerance for life, the variable nature of the crust's temperature gradient could allow life to persist for longer than in any other refuge (O'Malley-James et al. Reference O'Malley-James, Greaves, Raven and Cockell2013).
Remote detectability
For atmospheric abundances similar to present-day levels, CO2, H2O and O2/O3 would all be remotely detectable with low resolution spectrographs on an extrasolar Earth-like planet (Selsis Reference Selsis2004; Kaltenegger et al. Reference Kaltenegger2010). The only case where atmospheric O2 and O3 would be detectable is in the beginning of the time period covered in (A) (see Fig. 5). H2O would be detectable in cases (A) and (B) (Figs. 5 and 6). Estimated CO2 levels are lower than current atmospheric levels, making the remote detection of this less likely in all three cases. Ethane related to the large-scale decay of plant and animal species, in the extreme case discussed in (A), is theoretically remotely detectable, but water vapour and other atmospheric aerosols could interfere with its detection (Domagal-Goldman et al. Reference Domagal-Goldman, Meadows, Claire and Kasting2011). Although not as biologically relevant, H2 should be detectable at the levels predicted in both cases (A) and (B), producing a particularly strong signal in case (B) (Miller-Ricci & Fortney Reference Miller-Ricci and Fortney2010). N2, despite its abundance, is spectrally inactive. It lacks any marked electronic transition (except at extreme UV wavelengths), but its presence can be inferred from collisions between O2 and N2 molecules when observing the transmission spectra of planetary atmospheres (Pallé et al. Reference Pallé, Osorio, Barrena, Montañés-Rodríguez and Martín2009). Oxygen collision complexes and molecular oxygen bands can then be used to derive an average atmospheric N2 column density. In low O2 atmospheres, detecting N2 may therefore be more challenging. From the minimum abundance limits for detection described in Selsis (Reference Selsis2004) only CH4 would be potentially detectable in cases (A), (B) and (C) – Figs. 5–7 – with NH3 being detectable in case (A). All other atmospheric species fall below the minimum abundance thresholds in each case. At current CH4 levels, the main CH4 spectral feature is found in the mid-infrared (IR) wavelength region (at ∼7 μm), but this is largely obscured by adjacent H2O and N2O features (Des Marais et al. Reference Des Marais, Harwitt, Jucks, Kasting, Lin, Lunine, Schneider, Seager, Traub and Woolf2002). The predicted far-future CH4 levels are approximately an order of magnitude greater than present-day levels, which, combined with negligible atmospheric H2O and N2O levels, would make the 7 μm feature clearer and more readily remotely detectable. Upcoming and planned future missions, such as the James Webb Space Telescope (Gardner et al. Reference Gardner2006), Terrestrial Planet Finder (Lawson et al. Reference Lawson2008) and the Exoplanet Characterisation Observatory (EChO) (Tinetti & EChO Team Reference Tinetti2012) would all be looking in appropriate wavelength ranges to detect such a feature (although EChO would be limited to super-Earth (>10 Earth-mass) planets), with follow-up observations being possible with future ground-based instruments such as the European Extremely Large Telescope (Gilmozzi & Spyromilio Reference Gilmozzi and Spyromilio2007). However, a positive CH4 detection may not necessarily indicate the presence of life. CH4 can be released to the atmosphere abiotically via volcanic activity and serpentinization, for example (Oze et al. Reference Oze, Jones, Goldsmith and Rosenbauer2012; Burton et al. Reference Burton, Sawyer and Granieri2013). Building a case for a biological CH4 source on a late-habitable-stage Earth-like planet may be easier than for younger habitable planets due to the slowing of carbon recycling as a result of the slowing of plate tectonics and a lower abundance of water to drive serpentinization reactions. Recent experimental work also suggests that a low atmospheric H2/CH4 ratio would support the case that CH4 is being produced by life rather than via serpentinization (Oze et al. Reference Oze, Jones, Goldsmith and Rosenbauer2012). The potentially detectable biosignature gases are summarized in Table 4.
Other biosignatures
Clouds
A potential refuge for microbial life in the far future would be in the atmosphere where temperatures would be cooler than at maximum land elevation. It has been postulated that if life existed on a more clement early Venus, it could have found its way into the Venusian atmosphere and could be making a living in the upper cloud decks where temperatures fall within a habitable range (Schulze-Makuch Reference Schulze-Makuch, Grinspoon, Abbas, Irwin and Bullock2004).
If a similar scenario were to occur when Earth's oceans evaporate, one of the final refuges for life may be in the air. Microbial life is known to exist in the atmosphere today, although it is not yet known whether they are just in transit or whether they are actively metabolizing and reproducing in the atmosphere. Airborne microorganisms can influence cloud formation (see later discussion), suggesting that if biological cloud formation can be distinguished from abiotic cloud formation, clouds could be used as bio-indicators of an aerial biosphere.
Biological particles have little to no effect on cirrus clouds (thin, wispy clouds above altitudes of 5 km) (Möhler et al. Reference Möhler, Demott, Vali and Levin2007). Different types of cloud can be distinguished using information on cloud-top pressure (the atmospheric pressure at the cloud top) and optical thickness. Cirrus clouds (the most common type of cloud on Earth, making up 25% of the total global cloud cover and playing a large role in Earth's radiation budget) have high cloud-top pressures and low optical thickness, whereas convective clouds (responsible for most precipitation in the tropics) have a high cloud-top pressure and high optical thickness (Durieux et al. Reference Durieux, Machado and Laurent2003). It is feasible that these properties could be detected in the atmospheric spectra of exoplanets, although, if the cloud-top is at a deep layer in the atmosphere and the optical thickness of the cloud layer is high, direct detection is not possible, but a lower limit on cloud-top pressure can be determined (Benneke & Seager Reference Benneke and Seager2012).
In a recent study of the effect of airborne microbial populations on cloud formation, DeLeon-Rodriguez et al. (Reference DeLeon-Rodriguez, Lathem, Rodriguez-R, Barazesh, Anderson, Beyersdorf, Ziemba, Bergin, Nenes and Konstantinidis2013) found that 20% of particles sampled in the upper troposphere were bacterial (the remainder being salt, dust etc.). If we assume that generally, 20% of cloud seeding particles are microbial this could make cloud cover itself a biosignature on inhabited, arid planets if the expected abiotic cloud coverage could be estimated. For largely arid, desert planets, high desert dust levels in the atmosphere could influence the predominant cloud type. Saharan desert dust, for example, acts as small cloud condensation nuclei, causing clouds composed of small water droplets (<14 μm radii) to form, reducing coalescence and suppressing precipitation (Rosenfeld et al. Reference Rosenfeld, Rudich and Lahav2001). Cloud water droplet size could be determined via remote sensing of exoplanet atmospheres using the rainbow scattering effect (Bailey Reference Bailey2007); hence, unexpectedly large cloud droplets in the atmospheres of arid planets could provide indirect evidence of microbial cloud seeding.
Leaf area index
A decrease in vegetation during the stage in which plant extinction occurs could produce detectable spectral signatures. Remote sensing can be used to estimate vegetation coverage from the leaf area index (LAI), defined as green leaf area/surface area. Spectral reflectance in the near-IR and visible wavelength regions is used to determine the normalized difference vegetation index (NDVI)
from which an estimate of LAI (locally or globally) can be obtained (see, e.g. Deng et al. (Reference Deng, Chen, Plummer, Chen and Pise2006), Ganguly et al. (Reference Ganguly2012)). As plant extinction rates increase, the LAI should decrease as leaf area decreases, potentially leading to a less intense red edge spectral feature. Filella & Peñuelas (Reference Filella and Peñuelas1994) found that the area of the red edge peak changed with LAI, as LAI determines the ratio of near-IR and red reflectance. C4 plants have been shown to exhibit greater leaf area than C3 plants grown under the same conditions (Anten et al. Reference Anten, Schieving, Medina, Werger and Schufflen1995), suggesting the presence of more rapid photosynthesizers could delay a reduction in the spectral red edge signal. However, as plant species become extinct, vegetation will become increasingly sparse, with LAI values approaching high desert values (<1 – Running et al. Reference Running, Peterson, Spanner and Teuber1986) and the LAI – NDVI relation may no longer hold. When vegetation cover becomes too low, the observed spectrum is dominated by soil and rock, which have varied red-near-IR slopes, which alter the measured vegetation indices (Elvidge & Lyon Reference Elvidge and Lyon1985; Huete et al. Reference Huete, Jackson and Post1985). Additionally, as vegetation becomes more adapted to arid conditions it is likely to incorporate adaptations that decrease the amount of visible light absorbed, making it harder to detect.
Increasingly sparse vegetation will lessen the intensity of the red edge signature, which will eventually disappear completely with the end of all plant life. Warming temperatures lead to a poleward expansion of the boundaries of equatorial Hadley cells (currently at ±30° latitude), which has been linked to the simultaneous poleward expansion of subtropical dry zones (Lu et al. Reference Lu, Vecchi and Reichler2007). The meridional extent of the equatorial Hadley cells can be approximated using
where g is the acceleration due to gravity, H t is the height of the tropical tropopause, Δh is the equator-to-pole difference in radiative-equilibrium potential temperature, Ω is the rate of rotation of the Earth, a is the radius of the Earth and θ 0 is the global mean temperature (Showman et al. Reference Showman, Cho, Menou and Seager2010). Using the model results for temperature evolution and accounting for the slowing of the Earth's rotation due to tidal interactions with the Moon (increasing the day length by 1.70 × 10−5 s yr−1 – Stephenson Reference Stephenson1997), ϕ H can be used to estimate the expansion of the subtropical dry zones, giving an approximate timescale for the loss of abundant vegetation. Within 0.3 Gyr equatorial Hadley cells would have expanded to ∼40° north and south of the equator, increasing the arid fraction of the surface by 25%, doubling the current total area of arid land. By assuming plants in arid regions are too sparsely spread to contribute to the red edge signature, this alone would reduce the strength of the signature. However, Earth is currently half-way through a supercontinent cycle, with another supercontinent expected to be formed within the next 250 Myr (Yoshida & Santosh Reference Yoshida and Santosh2011). The location of this landmass would greatly influence environmental conditions. An equatorial supercontinent would have a more stable climate, but would likely have large dry, arid interiors, whereas polar supercontinents would experience extreme temperature fluctuations, except in coastal regions (Williams & Kasting Reference Williams and Kasting1997). Either case would lead to large areas of land with only sparse vegetation coverage, effectively ending the distinctive vegetation red edge signature. Supercontinent formation and break-up have been linked to past mass extinction events (Santosh Reference Santosh2010). Increased competition between species due to the merging of continental landmasses and the possibility of increased volcanic emissions from deep mantle superplumes, combined with generally less hospitable conditions on the warming Earth could lead to a mass extinction event for plants and animals from which they may not be able to recover.
Biosignatures of dying biospheres on non-Earth-like planets
Habitable planets that are not direct Earth analogues would probably exhibit different biosignatures associated with dying planetary biospheres when compared to the case of Earth analogue planets nearing the end of their habitable lifetimes.
Smaller, colder Mars-like planets would end their habitable lifetimes in a cold, arid state, with the best habitats for any remaining life being in the warmer subsurface regions. Outgassing from the near subsurface could be one of the indicators of ongoing biological activity on such planets.
For example, there is a claim that methane has been detected in plumes emanating from three distinct areas on Mars (Mumma et al. Reference Mumma, Villanueva, Novak, Hewagama, Bonev, DiSanti, Avi, Mandell and Smith2009). As the lifetime of methane in Mars' atmosphere is relatively short (Elwood Madden et al. Reference Elwood Madden, Leeman, Root and Gainey2011), the detection of methane implies continual replenishment of the gas. Methane in Earth's atmosphere is largely biological in origin; hence, it has been suggested that this may be an indicator of ongoing biological processes on the planet (Mumma et al. Reference Mumma, Villanueva, Novak, Hewagama, Bonev, DiSanti, Avi, Mandell and Smith2009). However, there are also abiotic mechanisms, such as the release of ancient methane trapped beneath the surface, that could be responsible for the atmospheric methane (McMahon et al. Reference McMahon, Parnell and Blamey2013) and the actual detection of methane on Mars has been called into question (Zahnle et al. Reference Zahnle, Freedman and Catling2011; Webster et al. Reference Webster, Mahaffy, Atreya, Flesch, Christensen and Farley2013).
Similarly, hotter inner-HZ, Venus-like planets would end their habitable lifetimes in a hotter and potentially geologically active state (the Venus Express spacecraft found evidence of recent volcanism on Venus – Smrekar et al. Reference Smrekar, Stofan, Mueller, Treiman, Elkins-Tanton, Herbert, Piccioni and Drossart2010). The habitable lifetimes of such planets would likely be much shorter than Earth-like planets (if warm oceans were present on early Venus, its habitable lifetime would correspond to the lifetime of liquid surface water on the planet – estimates for this range from 600 Myr (Kasting Reference Kasting1988) to a few Gyr (Grinspoon & Bullock Reference Grinspoon and Bullock2003)). The upper atmosphere would likely be the best final refuge for any life that developed on such planets. For example, the possibility of microbial life in the cloud decks of Venus has long been postulated (Sagan & Morowitz Reference Sagan and Morowitz1967; Sagan & Saltpeter Reference Sagan and Saltpeter1976; Cockell Reference Cockell1999). It may be that any microbial life that developed on the early Venus was able to migrate upwards as the planet's oceans evaporated, perhaps using sulphur allotropes in the atmosphere to survive (Schulze-Makuch et al. Reference Schulze-Makuch, Grinspoon, Abbas, Irwin and Bullock2004), with the presence of such life potentially being inferred from unexpected UV radiation absorption (a UV-transparent biosphere would be less favourable than one in which UV screening and repair mechanisms are developed – Cockell & Airo Reference Cockell and Airo2002) and atmospheric chemical disequilibrium.
Conclusions
The term ‘Earth-like’, when applied to atmospheric spectra or life tends to imply conditions similar to those on the present-day Earth. Earth's atmosphere and life inhabiting the planet have evolved considerably over time and will continue to do so into the future, which means that, for a lot of Earth's habitable lifetime, conditions would not be described as Earth-like when compared to the present environment. The implications for the search for extraterrestrial life are that the search criteria may need to be broadened to improve the chances of successfully detecting life.
The extinction of plants and animals caused by a warming planet as solar luminosity increases and the subsequent evaporation of the oceans will lead to different atmospheric composition and biosignatures when compared to the present. In the dry, microbial world left behind as Earth loses its water, observed spectra would once again change, as the spectra of a desert world with little or no ocean cover would exhibit light curves with much greater reflectivity than ocean and vegetation covered planets (Ford et al. Reference Ford, Seager and Turner2003).
Acknowledgements
The authors wish to thank Stefanie Lutz and Dirk Schulze-Makuch for helpful discussions and an anonymous reviewer for comments that improved the manuscript. The University of Dundee is a registered Scottish charity No. SC015096. JTO acknowledges an STFC Aurora grant.
Appendix
Equations for silicate weathering and carbonate deposition:
Weathering on land:
After movement down rivers in ocean:
After subduction (and after much delay) volcanism:
If the bicarbonate stayed in solution in the ocean more CO2 would be drawn down in weathering but it would eventually precipitate (even in the absence of calcifying organisms) as CO3 with production of half of the CO2 take up in the initial weathering.