pnas.1711842115.sapp.pdf

The biomass distribution on Earth

Y. M. Bar

On, R. Phillips, R. Milo

upplementary Information Appendix

Contents

Overview

................................

.........................

Plants

................................

...............................

Biomass

................................

.......................

Biomass of roots and leaves

................................

........................

Comparison of plant and bacterial biomass

................................

Bacteria and Archaea (Prokaryotes)

................................

...............

Marine

................................

.........................

Soil

................................

............................

Marine deep subsurface sediment

................................

.............

Terrestrial deep subs

urface

................................

.......................

Fungi

................................

.............................

Soil fungi

................................

...................

Ectomycorrhiza

................................

.........

Arbuscular mycorrhiza

................................

..............................

Marine fungi

................................

..............

Deep subsurface fungi

................................

...............................

Total fungal biomass

................................

Active and inactive microbial biomass

................................

.........

Annelids

................................

........................

Nematodes

................................

.....................

Litter microbes and fauna

................................

.............................

Chordates

................................

......................

Fish

................................

............................

Humans and livestock

................................

...............................

Wild mammals

................................

..........

Wild birds

................................

..................

Reptiles

................................

.....................

Amphibians

................................

...............

Arthropods

................................

....................

Previous estimates

................................

.....

Terrestrial arthropods

................................

Marine arthropods

................................

.....

Cnidarians

................................

.....................

Molluscs

................................

........................

Protists

................................

...........................

Marine protists

................................

..........

Terrestrial protists

................................

.....

Deep subsurface

................................

........

Sanity checks on the MAREDAT dataset

................................

.....

Comparison of MAREDAT and the

Tara

oceans dataset

................................

........

Cyanobacteria

................................

...........

Comparison of phytoplankton to remote s

ensing measurements

.............................

Comparison of the biomass of producers and consumers

................................

.............

Viruses

................................

..........................

Other Animal Phyla

................................

......

Benthic phyla biomass

................................

..............................

Pre

human biomass

................................

.......

Microbiomes

................................

.................

Inland water

................................

..................

Abundance estimates

................................

....

General

................................

......................

Archaea and Bacteria (Prokaryotes)

................................

.........

Fungi

................................

.........................

Molluscs

................................

....................

Fish

................................

............................

Arthropods

................................

................

Cnidaria

................................

.....................

Protists

................................

.......................

Usage of various estimators of the mean

................................

......

Supplementary References

................................

............................

Supplementary Figures and Tables

Fig. S1.

A graphical

representation of the global living biomass distribution.

............

Fig. S2. The relation between abundance and biomass of different taxa

....................

Fig. S3. The relation between species richness and biomass of different taxa.

............

Fig. S4. Comparison of graphical representation of the global biomass distribution

using pie

charts versus Voronoi diagrams.

................................

...

Fig. S5. The impact of human civilization on the biomass of mammals.

.....................

Table S1. Summary of estimated total biomass and abundance of various abundant

taxonomic groups.

................................

.........

Table S2. Methodology used to estimate the global biomass of plants.

.......................

Table S3.

Methodology used to estimate the

global biomass of marine prokaryotes.

Table S4.

Methodology for estimating the global biomass of soil prokaryotes.

..........

Table S5.

Methodology for estimating the global biomass of prokaryotes in the marine

deep subsurface.

................................

............

Table S6.

Methodology for estimating the global biomass of prokaryotes in the

terrestrial deep subsurface.

................................

............................

Table S7.

Methodology for estimating the fraction of archaea out of the total biomass

of prokaryotes.

................................

..............

Table S8.

Methodology for estimating the global biomass of fungi.

............................

Table S9.

Methodology for estimating the global biomass of humans.

........................

Table S10.

Methodology for estimating the global biomass of livestock.

....................

Table S11.

Methodology for estimating the global biomass of wild mammals.

..........

Table S12.

Methodology for estimating the global biomass of wild birds.

................

100

Table S13.

Methodology for estimating the global biomass of marin

e arthropods.

...

101

Table S14.

Methodology for estimating the global biomass of terrestrial arthropods.

102

Table S15.

Methodology for estimating the global biomass of fish.

..........................

103

Table S16.

Methodology for estimating the global biomass of annelids.

...................

104

Table S17.

Methodology for estimating the global biomass of molluscs.

..................

105

Table S18.

Methodology for estimating the global biomass of cnidarians.

................

106

Table S19.

Methodology for estimating

the global biomass of nematodes.

...............

107

Table S20.

Methodology for estimating the global biomass of marine protists.

........

108

Table S21.

Methodology for estimating the global biomass of terrestrial protists.

....

110

Table S22.

Methodology for estimating the global biomass of viruses.

.....................

111

Table S23. The global biomass concentrated in the terrestrial, marine or deep

subsurface environments.

................................

............................

112

Taxon

specific detailed description of data

sources and

procedures for estimating biomass

Overview

Here we provide a detailed description of the data and procedures for arriving at the final

estimates presented in the paper. The description is divided into the different groups of

organisms. For some

organisms, there is further division into the different environments they

reside in. We used

bold font

to highlight the concluding values derived for each taxonomic

group.

All of the data used to generate our estimates, along with the code for analyzing t

he data,

are open

source and available at

https://github.com/milo

lab/biomass_distribution

The

SI Appendix

and all the accompanying files are made accessible so that researchers can see in

a completely transparent manner how each value was derived from the many literature sources

and be able to update the analysis using extra data or a different data analysis approach

. We rely

on hundreds of studies from the literature to support

the data pr

esented in the

SI Appendix

. To

generate our estimates of biomass, we extract values from the literature into

spreadsheet

files.

Our analysis pipeline is comprised of about 50 different Jupyter notebooks which use the data

extracted from the literature as

input and generate our estimates. The result of our analysis is

summarized in a summary table located

https://github.com/milo

lab/biomass_distribution

use the

results of our analysis

hen reporting the values in the manuscript

and

SI Appendix

including all the associated figures and tables.

Plants

We provide a fully detailed calculation for the following analysis, including all of the data as well

as the steps taken to achieve these results, in the following

link

iomass

There are several estimates for the biomass of specific components of plant biomass, mainly

forests, and they are based on both direct field observations (termed inventory data;

1, 2)

and on

emote

sensing data

–

. These studies, however, have mainly focused on forests, and have not

quantified the contribution of other components of the global plant biomass, such as shrubs,

gra

sses, or locations with trees not defined as forests (because of an average tree cover of less

than 10%). Alternatively, well

established studies

–

provide estimates for the carbon density

nd extent in plants for all biomes, which takes into account also biomes that are not dominated

by trees (such as grasslands, shrublands, and savannas). These studies, however, provide

estimates for the biomass density of plants in undisturbed ecosystems,

and do not take into

account the very broad land

use changes which have affected many natural ecosystems. These

changes include transformation of natural ecosystems to cropland, as well as the impact of

forestry and grazing on natural ecosystems. This is o

ne of the reasons that the global estimates for

forest biomass in Saugier et al. or Ajtay et al. diverge from more recent inventory

based or

remote

sensing

based estimates

–

. The best resource we were able to find as a basis for our

estimate of the global biomass of plants is Erb et al.

(9)

. Their study assesses the biomass of both

forest and non

forest ecosystems and takes

into account impacts of land

use changes. We briefly

describe the methodology Erb et al. used for estimating the global biomass of plants, for full

details we refer the reader to the original paper

(9)

. E

rb et al. generated seven maps of plant

biomass stocks. Their best estimate for the global biomass of plants is based on the mean of those

seven different maps. The first six maps could be divided to two broad categories: two inventory

based maps, and four

remote sensing

based maps. We now describe those estimates in more

detail.

For the inventory

based maps, Erb et al. consider five main land cover categories: forests,

cropland, artificial grassland (forest area converted to pastures or meadows), other woo

ded land

10% tree cover, including savannas and shrubland) and natural grasslands (<5% tree cover).

To each land

use unit, Erb et al. assign typical biomass stock density values from the literature or

census statistics. In the first map, Erb et al. use

national

level data from the global Forest

Resource Assessment for the forest land

cover. The second map uses data on forest inventories

from Pan et al.

(1)

. Erb et al. set the biomass density of grasslan

tree mosaics (other wooded

land and natural grasslands which contain trees), at 50% of the biomass density of the

neighboring forests (forests which are in the same country as the grassland

tree mosaics). Erb et

al. based this assumption on data from the

FRA on the ratio between other wooded land biomass

and forest biomass in ≈70 countries

(10)

. For herbaceous vegetation units (artificial grassland,

cropland and natural grassland without trees), Erb et a

l. assumed biomass stocks to equal the

annual amount of net primary production. For permanent cropland, Erb et al. added 3 kg C m

for

tree

bearing systems and 1.5 kg C m

for shrub bearing systems to account for woody above

and

belowground compartments

For the remote sensing

based maps, Erb et al. combined independent remote

sensing products for

tree vegetation and expanded them to account for below

ground and herbaceous compartments

where necessary. The main data sources used for the construction of

the maps are Thurner et al.

(5)

for the northern boreal and temperate forests, and Baccini et al. and Saatchi et al. for tropical

forests

(3, 4

)

. For some regions (the southernmost part of Australia, and parts of Oceania), no

remote

sensing data are available. In these regions, Erb et al. used the inventory

based map to fill

missing values. The first two remote

sensing maps use either Saatchi et

al. or Baccini et al. for the

tropical forest component of the map. The additional two remote sensing

based maps use

the

minimal or maximal biomass densities at each grid cell of the map, respectively. The last map

Erb et al. use is the map constructed by

Ruesch & Gibbs based on the IPCC tier

1 globally

consistent default biomass values

(8)

. To estimate the total biomass of plants from those seven

different estimates, we use the best estimate reported by E

rb et al., which is ≈450 Gt C (

link

full calculation).

The analysis of Erb et al. does not include halophytic flowering plants

(plants living in water with

high salinity) and bryophytes (liverworts, hornworts and mosses). Biomass of halophytic

flowering plants is dominated by mangroves and seagrass. An estimate for the total biomass of

mangroves is roughly ≈4 Gt C

(11)

. This value is consistent with estimates for the global storage

of organic carbon in mangroves, and the fraction of this carbon that is living mangroves (and not

dead biomass or soil organic carbon;

12)

. For seagrass, Fourqurean et al.

(13)

, estimate a global

biomass of ≈0.1 Gt

C. As for bryophytes, there seems to be no global data on bryophytes alone,

but Elbert e

t al.

(14)

give values of ≈5 Gt C for ‘cryptogamic covers’, i.e. varying mixtures of

bryophytes, lichens, eukaryotic algae, cyanobacteria and fungi growing as epiphytes, as crusts on

arid lands, and in bo

gs. This translates into an organic C content of about 1% of the total for

terrestrial plants. Supporting these general values, Edwards et al.

(15)

analyzed the global carbon

content of ‘cryptogamic cover

s’ in the Ordovician

Silurian period (about 400

500 million years

ago) and suggested it was similar to today’s values, i.e. a total of about ≈5 Gt C.

For the sum of

marine macroalgae (green and brown algae), De Vooys

(16)

gives a total global biomass of of

0.0075 Gt C, based on annual productivity and assuming one turnover of the standing crop each

year. Cherpy

Roubaud & Sournia

(17)

cite global annual pr

oductivity of all taxa of marine

macroalgae of 2.55 Gt C yr

, or a global standing crop of 2.55 Gt C assuming one turnover of the

standing crop each year

(18)

. This 2

3 orders of magnitude range of value

s (0.0075

2.55 Gt C)

means that more work is needed to obtain tightly

constrained estimates.

Plant biomass is dominated by land plants (embryophytes), and more specifically by vascular

plants (tracheophytes), with only a minor contribution from bryophyte biomass

(14)

. Overall, we

estimate that th

e global plant biomass, including contributions from land plants as well as other

groups such as bryophytes and all marine plant contributions is

≈450 Gt C

We now analyze the associated uncertainty of the estimate for the total biomass of plants. We

rep

ort the uncertainty as a fold change factor from the mean, representing a range akin to a 95%

confidence interval of the estimate. One approach to assess the uncertainty associated with the

estimate of the total biomass of plants is to calculate the 95% co

nfidence interval around the

geometric mean of the seven estimates from the seven different maps generated by Erb et al.

(9)

This yields a rather small uncertainty of ≈1.1

fold (

link

to full calculation). However, this

procedure only includes uncertainty stemming from the variation between different

estimates and

does not include the systematic unc

ertainty stemming from assumptions made to produce each

one of the seven estimates. The main type of uncertainty we believe is not accounted for by this

procedure is the uncertainty in the biomass contribution from other wooded land (such as

savannas). In

the inventory

based estimates, the biomass density of other wooded land is assumed

to be ≈50% of the national average biomass density of forests. This assumption has uncertainty

associated with it which is not easily quantified. Similarly, the remote sensi

based estimates are

mainly designed to quantify the biomass of forests, and therefore there is significant uncertainty

regarding the measurement of non

forest tree and shrub plant forms by remote sensing. Providing

a rigorous quantification of this type

of uncertainty is hard. To account, even if partially, for this

uncertainty, we use the multiplicative ratio between the upper (and lower) most estimate relative

to our best estimate, which is ≈1.2

fold, as our best projection of the uncertainty associate

d with

our estimate of the biomass of plants (

link

to full calculation).

iomass

of roots and leaves

Plant tissues are composed of an extracellular scaffold made out of cell wall (mainly cellulose

and lignin), supporting a network of cytoplasmic space, termed the protoplasm. Conceptually,

plants are not different from other organisms, which also contain s

upporting tissues such as endo

or exoskeletons, or even the extracellular matrix of microbial biofilms. The ratio between

protoplasm and cell wall varies between plant compartments, with leaves containing the least

amount of supporting tissues and stems o

f woody plants (such as trees) mainly composed of

supporting tissues. To estimate the total biomass of non

woody tissues, we chose to remove the

biomass of stems tissue, as it is dominated by cell wall. Roots also have a non

negligible fraction

of cell wal

l biomass, but the ratio between cell wall and protoplasm in roots is harder to estimate

globally. Therefore, to calculate the non

woody plant biomass fraction, we consider only the

biomass of leaves and roots. In order to estimate the total biomass of roo

ts and leaves, we rely on

two independent methods. The first relies on a meta

analysis of the biomass allocation to

different plant compartments across biomes

(19)

. Using these data, we calculated the ave

rage

biomass fraction of leaves and roots out of the total biomass to be ≈30% by taking into account

the contribution of each biome to the global plant biomass

(20

;

link

to full calculation). This

means that out of the 450 Gt C of plant biomass, 30%, or ≈150 Gt C is concentrated in

metabolically active pl

ant tissues.

Our second method for estimating the non

woody plant biomass combines estimates for the

global biomass of leaves and roots. For the global root biomass, we rely on the estimate from

Jackson et al.

(21)

, of ≈150 Gt C. For the global leaf biomass, we calculate the total leaf area of

forests, which dominate plant biomass, by using a combination of biome level estimates for the

leaf area index (LAI;

22)

and the total forest area in each biome

(23)

. We then convert the total

area of leaves to an estimate of the total dry mass of leaves by using the Glopnet database

(24)

which measured leaf mass per area (LMA) for ≈2500 plant species. This independent procedure

yields an estimate of ≈30 Gt dry weight, or ≈15 Gt C assuming 50% carbon content (

link

to full

calculation). Summing our estimate for the total biomass of roots and leaves, we get ≈160 Gt C.

We use the geometric mean of both methods,

which is ≈150 Gt C, as our best estimate of the total

non

woody plant biomass.

To estimate the total belowground plant biomass, we rely on the same two methods as for the

total non

woody biomass. We now consider only root biomass. The first method yields

estimate of ≈120 Gt C (

link

to full calculation). Our second source is the estimate of the global

root biomas

s from Jackson et al.

(21)

. We use the geometric mean of both estimates, which is

≈130 Gt C, as our best estimate for the belowground plant biomass. This puts the aboveground

plant biomass at ≈320 Gt C.

Comparison of plant and bacterial biomass

Our best estimates for the total biomass of plants and bacteria suggests that these taxa account for

≈80% and ≈10% of the global biomass of the biosphere, respectively. The uncertainty associated

with our estimate

of the biomass of plants is relatively small (≈1.2

fold), and the uncertainty

associated with our estimate of the biomass of bacteria is much larger (≈9

fold). In order to

quantify the certainty in the statement that plants are more dominant in terms of bi

omass than

bacteria, we use a bootstrapping approach. We randomly sample from the distribution of our

estimates for the total biomass of plants and bacteria (with the width of the distribution

corresponding to the level of our uncertainty). We find that in

≈90% of cases plant biomass is

higher than the biomass of bacteria (

link

to full calculation).

Bacteria and Archaea (Prokaryotes)

We provide a fully det

ailed calculation for the following analysis, including all of the data as well

as the steps taken to achieve these results, in the following

link

One of the be

known estimates of global biomass of prokaryotes (bacteria and archaea) is the

meta

analysis study by Whitman et al.

(25)

, which analyzed data of cell abundance densities for

various environments and e

xtrapolated the total prokaryotic biomass using estimates of cell mass

in each environment. We used this study as a baseline estimate on which various corrections and

updates were imposed for the different environments, as highlighted below. An estimate fo

r the

distribution of biomass between bacteria and archaea is not available currently. We generate

estimates for the biomass of bacteria and archaea in a two

step process. First, we estimate the

total biomass of prokaryotes in each environment, and then we

estimate the fraction of bacteria

and archaea out of the total biomass of this environment. Combining the contributions from each

environment, we estimate that the global biomass of bacteria is ≈70 Gt C, which is dominated by

≈60 Gt C of terrestrial deep

subsurface bacteria. We estimate the global biomass of archaea at ≈7

Gt C, with ≈4 Gt C and ≈3 Gt C contributed by the terrestrial and marine deep subsurface,

respectively. In addition to estimating the biomass of prokaryotes in each environment, we also

resent in detail the uncertainties associated with each estimate. When combining the

uncertainties for the biomass of bacteria and archaea in different environments we estimate the

uncertainty of the global biomass of bacteria and archaea to be about 10

ld and 13

fold,

respectively, dominated by the uncertainty of the biomass of terrestrial deep subsurface bacteria

and archaea.

Marine

To generate an estimate of the biomass of marine bacteria and archaea (prokaryotes), we first

estimate the total number of marine prokaryotes, and the characteristic carbon content of a single

marine prokaryote. We generate our estimate for the biomass of

marine prokaryotes by

multiplying the total number of cells by the characteristic carbon content per cell. Our estimate

for the total number of marine prokaryotes relies on three recent papers

(26

–

28)

. The first is a

review by Arístegui et al. which included a meta

analysis of values from the literature on the

concentration of cells of pelagic prokaryotes at epipelagic (<200 m), mesopelagic (200

1000 m)

and bathypelagic (1000

4000 m) depths. Ar

ístegui et al. used many samples of cell density per

volume in each depth zone to generate an average cell concentration for each depth zone.

Arístegui et al. then generate estimates for the total number of cells per unit area for each depth

zone by applyi

ng the average cell concentration per unit volume across the entire depth of the

zone (≈200 m for epipelagic, ≈800 m for mesopelagic and ≈3000 m for bathypelagic). We used

the depth integrated estimates of cell concentration per unit area for each zone in

Arístegui et al.

(29)

, and multiplied them by the surface area of the oceans (3.6×10

) to give a total estimate

of 1.7×10

cells (

link

to full calculation). As a second source for estimating the total number of

marine prokaryotes, we use Buitenhuis et al.

(30)

. Buitenhuis et al. compiled a database of 39,766

data points consisting of flow cytometric and microscopic measurements of the abundance of

marine prokaryotes with observations covering depths even belo

w 4 km. We binned the data

along the water column with bins every 100 meters. We calculated the average cell

concentrations at each depth bin and multiplied the average cell concentration by the total volume

of water in each depth bin to generate an estima

te for the total number of marine prokaryotes. We

estimate estimate a total of ≈1.3×10

cells based on the data from Buitenhuis et al. (

link

to full

calculation). The third resource we used for estimating the total number of marine prokaryotes is

a recent meta

analysis by Lloyd et al.

(28)

, which gathered ≈20 studies measuring bacterial and

archaeal abundance using fluorescent in situ hybridization (FISH). Lloyd et al. fit an equation to

predict the concentration of bacteria and archaea based on the depth at which

the sample was

collected. We used the fits to extrapolate the number of bacteria and archaea across the entire

depth of the water column. We then estimated the total number of bacterial and archaeal cells by

multiplying the concentration at each depth wit

h the total volume of water at that depth, and

integrating across all depths. Lloyd et al. also report the fraction of cells that typically get labeled

by the FISH signal. We used the geometric mean of this fraction, which is ≈0.8, to extend the

estimate o

f total bacterial and archaeal cells labelled by FISH to an estimate for the total number

of bacterial and archaeal cell in the ocean (

link

to full calculation). Based on the data from Lloyd

et al. we estimate a total of ≈8×10

cells. As our best guess for the total number of marine

prokaryotes, we take the geomet

ric mean of the estimates from Arístegui et al., Buitenhuis et al.

and Lloyd et al., which is ≈1.2×10

cells (

link

to full calculation). This number corresponds well

with the estimate of ≈1.2×10

cells made by Whitman et al.

(25)

. To estimate the characteristic

carbon content of a marine prokaryote, we rely on several studies from the literature

(31

–

35)

. We

use the geometric mean of the values reported, which is ≈11 fg C cell

as our best estimate for

the carbon content of marine prokaryotes (

link

to full calculation). To generate our best estimate

for the total biomass of marine prokaryotes we multiply our best estimate for the total number of

marine prokaryotes by our best estimate for the carbon content of marine prokaryotes. We

thus

arrive at a total biomass of ≈1.3 Gt C of marine prokaryotes, which is on par with the 2.2 Gt C

estimate of Whitman.

We note that these estimates referred to planktonic organisms and did not refer to prokaryotes

attached to large particulate organi

c matter (POM) In general, POM can be divided into two main

size fractions

microaggregates (5

500 μm in diameter) and macroaggregates (>500 μm in

diameter;

36)

. Most of the studies on the abundance and

relative contribution of particle

attached

bacteria and archaea in the oceans are focused on a small number of sampling sites, and thus a

robust global accounting of the contribution of particle

attached bacteria and archaea is not easily

attainable. We pr

oceed to make a crude estimate of the biomass contribution of bacteria and

archaea on microaggregates and macroaggregates. For macroaggregates, we rely on studies which

measured the relative fraction of cells attached to macroaggregates out of the total po

pulation of

cells

(37

–

44)

. We calculate the average value of those studies, and thus estimate that bacterial and

archaeal cells attached to macroaggregates account for ≈

3% of the total number of bacterial and

archaeal cells in the marine environment (

link

to full calculation).

ur samples of the

concentration of bacterial and archaeal cells attached to macroaggregates cover depths up to 1000

meters. We could not find samples of the abundance of particle

attached cells in the bathypelagic

realm. We assume that measurements in the

epipelagic and mesopelagic realms are characteristic

of the bathypelagic realm. To support this assumption, we compare the concentration of

macroaggregates measured in the deep

sea

(45

–

47)

. We f

ind that the concentration of

macroaggregates in the deep

sea is similar to the concentration of macroaggregates reported in

the studies on which we rely (

link

to full calculation).

For microaggregates, we have a more limited set of observations

(48, 49)

, but they suggest that

bacteria and archaea on micro

aggregates account for ≈4% of the total number of cells (

link

to full

calculation). It is important to note th

at from the available data

(37, 40, 42, 43, 50)

the

characteristic volume of bacterial and archaeal cells on aggregates is typically larger than that of

free

living bacteria and archaea in the marine environment. The data on the specific factor by

which particle

attached cells are larger than free

livi

ng cells is very limited. We rely on several

studies

(37, 40, 42, 43, 50)

, which suggest that, on average, particle

attached cells contain ≈3

fold

more carbon than free

living cells (

link

to full calculation). This means that even though in terms

of cell abundance par

ticle

attached cells account for ≈7% of the total population of cells in the

marine environment, in terms of carbon content they account for ≈20% of the total carbon (

link

full calculation). In order to account for the additional biomass contributed by particle

attached

bacteria and arc

haea, we use ≈1.6 Gt C (instead of the ≈1.3 Gt C calculated above) as our best

estimate for the total biomass of marine bacteria and archaea. We note that depending on the

method used to estimate the total number of bacterial cells in the ocean, some cells

attached to

microaggregates might be counted as free

living (the microaggregates are transparent, so in case

cells are counted by microscopy, they might be counted as free

living). Due to the scarcity of

data, we assume the distribution of biomass between

bacteria and archaea on particles is similar

to that of the surrounding water. Our estimates on the total abundance and carbon content of

particle

attached cells in the oceans is much less robust that our estimates for free

living cells,

and much more com

prehensive data is needed to probe the global importance of particle

attached

cells in the ocean.

The abundance of archaea in the deep ocean (mesopelagic realm, 200

1000m depth, and

bathypelagic realm, 1000

4000m depth) has been reported by Karner et al.

(51)

, which estimated

that archaea constitute about half the prokaryotic cells below 1000m in the Hawai'i Ocean Time

series station. Karner et al.

(51)

used fluorescent

situ

hybridization (FISH) with specific 16S

probes to estimate the fraction of bacteria and archaea out of the total population of prokaryotes.

In shallower waters, bacteria dominate the population

(51)

. Similar findings were found using

FISH in the Atlantic Ocean

(52, 53)

. To rigorously estimate the fraction of archaea out of the total

biomass of marine prokaryotes, we rely on t

wo independent methods

FISH and 16S rDNA

sequencing. A recent paper by Lloyd et al.

(28)

collected studies reporting on the number of

archaea and bacteria at different depths, based on FISH. As part of

the procedure to estimate the

total number of marine prokaryotes from the data in Lloyd, we calculate the total number of

archaeal and bacterial cells. The fraction of archaeal cells out of the total number of bacterial and

archaeal cells is ≈20%. In the e

pipelagic, mesopelagic and bathypelagic zone, archaea represent

≈6%, 24% and 35% of the the total number of cells, respectively (

link

to full calculation). An

alternative methodology to quantify the fraction of archaea out of the marine prokaryote

population is by using 16S rDNA sequencing. We use studies which have measured the fraction

of archaea in the epip

elagic, mesopelagic and bathypelagic realms. For the epipelagic and

mesopelagic realms, we use data from the Tara Oceans campaign

(54)

, which is based on ≈250

samples worldwide. The fraction of archaeal 1

6S rDNA sequences out of the total pool of 16S

sequences in the epipelagic and mesopelagic realms is ≈4% and ≈14%, respectively (

link

to full

calculation). For the bathypelagic realm, we rely on data from the recent Malaspina campaign

(55)

, which is based on 30 samples ranging in depth from ≈2 km to ≈4 km.

The average fraction

of archaeal 16S rDNA sequences out of the total pool of 16S sequences is ≈15% (

link

o full

calculation). Estimates based on 16S rDNA sequencing data are lower by about 2

fold than FISH

based estimates for the fraction of archaea across depths. This might be caused by the fact that the

copy number of rRNA operons in bacterial genomes is on

average ≈2

fold larger than that of

archaeal genomes

(56)

. We use the geometric mean of estimates from the two methodologies as

our best guess for the fraction of archaea out of the total population of m

arine prokaryotes in the

different layers of the ocean. Our best estimates for the epipelagic, mesopelagic and bathypelagic

realms are thus ≈7%, 26%, and 32%, respectively (

link

to full calculation). From the data in

Arístegui et al. and in Lloyd et al. we generate estimates for the fraction of prokaryotes that are

fou

nd in the epipelagic, mesopelagic, and bathypelagic realms. Applying the fractions of archaea

across the different environments, we estimate that archaeal cells represent ≈20% of the total cells

of marine prokaryotes (

link

to full calculation). As the average cell sizes of bacteria and archaea

don’t seem to vary consid

erably

(57)

, we estimate that the biomass of marine archaea represents

≈20% of the total biomass of marine prokaryotes, which is

≈0.3 Gt C

. This puts the estimate of

the biomass of marine bacteria at

≈1.3

Gt C

(

link

to full calculation).

We now analyze the associated uncertainty of the estimate for the total biomass of marine

bacteria

and archaea, which we report as a fold

change factor from the mean representing a range

akin to a 95% confidence interval of the estimate. In this analysis, we consider the following

factors. First, we assess the uncertainty associated with the estimate of

the total number of marine

prokaryotes. We then estimate the uncertainty associated with the estimate of the characteristic

carbon content of a single marine prokaryote. Finally, we assess the uncertainty associated with

the estimate of the fraction of ar

chaea out of the total population of marine prokaryotes. The

intra

study uncertainty reported in Arístegui et al. for the estimate of the cell concentration is

≈10% of the mean. Buitenhuis et al. and Lloyd et al. do not report uncertainty ranges for the

timate of the total number of cells of marine prokaryotes. The inter

study uncertainty between

these three studies is ≈1.5

fold (

link

to full calculation). For estimating the characteristic carbon

content of a single marine prokaryote, we used 9 independent studies which measure carbon

content in the open ocean. The maximum reported intra

study unc

ertainty is ≈1.4

fold, and the

inter

study uncertainty between the five studies is ≈1.4

fold (

ink

to full calculation). We thus

chose to use ≈1.4

fold to project the uncertainty associated with the carbon content of a single

marine prokaryote. Combining the uncertainty associated with our estimate of total number and

the uncertainty associated with

our estimate of the carbon content of a single marine prokaryote,

we arrive at an uncertainty of ≈1.7

fold for the total biomass of marine prokaryotes. We also

analyze the uncertainty associated with our estimate of the fraction of the total biomass of ma

rine

bacteria and archaea which is particle

attached. Our estimate relies on two main factors

the

fraction of the total number of cells which is particle

attached, and the carbon content of particle

attached bacteria and archaea relative to free

living b

acteria and archaea. Our projection for the

uncertainty associated with our estimate of the fraction of the total number of cells which is

particle

attached is based on collecting the intra

study and inter

study uncertainty associated with

the estimate of

the fraction of the total number of cells which are particle

attached, and using the

maximum of the uncertainties values, which is ≈3

fold (

link

to full calculation). We repeat the

same procedure for the estimate of the carbon content of particle

attached cells relative to free

living cells, and project an uncertainty of ≈3

fold (

link

to full calculation). We combine the

uncertainties associated with the two factors and project an uncertainty of ≈5

fold associated w

ith

our estimate of the total biomass of marine bacteria and archaea which are attached to particles

(

lin

to full calculation). We then combine the uncertainty of the total biomass of free

living

bacteria and archaea and particle

attached bacteria and archaea, and project an uncertainty of

≈1.8

fold associated with our estimate of the total biomass of marine

bacteria and archaea.

For the fraction of archaea out of the population of marine prokaryotes, the intra

study

uncertainty is ≈1.2

fold for Salazar et al. and ≈1.1

fold for Lloyd et al. (

link

to full calculation).

The inter

study uncertainty between FISH

based studies and 16S rDNA sequencing

based studies

is around ≈2.3

fold for archaea and ≈1.3

fold for b

acteria (

link

to full calculation). We use the

higher inter

study uncertainties for projecting the unce

rtainty of the fraction of marine archaea

and bacteria out of the total marine prokaryote population. Combining the uncertainties for the

biomass of marine prokaryotes with the uncertainties associated with the estimate of the fraction

of archaeal and bact

erial cells out of the total population of marine prokaryotes, we project an

uncertainty of ≈2

fold for the biomass of marine bacteria, and ≈3

fold for marine archaea (

link

full calculation).

Soil

To estimate the total biomass of soil bacteria and archaea, we rely on the estimate of the total

biomass of soil microbes we derived in the soil fungi section. We estimate a total microbial

biomass of 20 Gt C, of which ≈12 Gt C are fungal. This leaves us w

ith ≈8 Gt C of bacterial and

archaeal

biomass. In order to derive the respective fractions of archaea and bacteria out of this

total biomass, we rely on four independent methods which estimate the fraction of archaea out of

the total biomass of bacteria an

d archaea. The methods we rely upon are 16S rDNA sequencing,

16S rDNA quantitative PCR (qPCR), fluorescent

situ

hybridization (FISH), and catalyzed

reporter deposition FISH (CARD

FISH). We calculated the mean estimate of the fraction of

archaea out of t

he total biomass of soil bacteria and archaea for each method. We then used the

geometric mean of values from the different methods as our best estimate for the fraction of

archaea out of the total biomass of soil bacteria and archaea. For our 16S rDNA seq

uencing

based

estimate, we rely on a study which reported values for the fraction of archaea out of the total

population of soil bacteria and archaea in 146 soils from across the globe

(58)

. The mean frac

tion

of archaea reported by Bates et al.

(58)

is ≈2%. We account for the lower rRNA operon copy

number in archaea

(56)

by multiplying the measured fractions by a factor of 2. This procedure

does not affect our results significantly. For our 16S qPCR

based estimate, we rely on a recent

study which reported the fraction of archaea out of the total population of soil bacteri

a and

archaea in grasslands, forests and croplands in Korea

(59)

. The mean fraction of archaea reported

by Hong & Cho

(59)

is ≈3%. For our FISH

based es

timate, we assembled data from about 10

studies

(60

–

68)

. We calculate the mean fraction of archaea across these studies, which is ≈20% as

our best FISH

based estima

te for the fraction of archaea out of the total population of soil bacteria

and archaea (

link

to full calculation).

For

our CARD

FISH

based estimate, we assembled data

from four studies

(69

–

72)

. We calculate the mean fraction of archaea across these studies, which

is ≈20% as our best CARD

FISH

based estimate

for the fraction of archaea out of the total

population of soil bacteria and archaea (

link

to full calculation). T

o generate our best estimate for

the fraction of archaea out of the total biomass of soil bacteria and archaea, we use the geometric

mean of these four estimates based on the four different methodologies. We thus estimate that

archaea represent ≈7% of the

total biomass of soil bacteria and archaea (

link

to full calc

ulation).

We combine this estimate with our estimate for the total biomass of soil bacteria and archaea,

which is 8 Gt C, and arrive at an estimate of

≈0.5 Gt C

of soil archaea, and

≈7 Gt C

of soil

bacteria (

link

to full calculation).

The different methods we rely on to estimate the fraction of archaea out of the population of soil

bacteria and archaea have various caveats associated with them. In

general

for methods based on

quantifying relative abundances of 16S rDNA sequences, we rely on the assumption that the

abundance of 16S rDNA sequences is proportional to the biomass of bacteria and archaea.

Relying on 16S sequence abundance as a proxy for biomass

is not a well

established practice. For

qPCR, a recent meta

analysis claimed relative fractions of archaea and bacteria calculated using

this methodology are reliable for the marine environment and in subseafloor sediments

(28)

However, we only have a limited amount of data which is based on qPCR measurements. The

same meta

analysis by Lloyd et al. also states that the use of FISH or CARD

FISH for

quantifying the abundance of archaea and bacteria is very

sensitive to the details of the

experimental

protocol and

was found to reliably represent the community structure of subseafloor

sediments only under a specific set of protocol parameters (e.g. CARD

FISH with cell

permeabilization using proteinase K). It

is thus not obvious that the estimates based on FISH or

CARD

FISH represent reliably the fraction of archaea out of the total population of soil bacteria

and archaea. Due to lack of better options, we chose to use the geometric mean of these four

methods,

each with its own caveats as our best estimate. We hope that further research will shed

light on the appropriate methodology to quantitatively describe biomass distribution of soil

microbes.

We now analyze the associated uncertainty of the estimate for th

e total biomass of soil bacteria

and archaea, which we report as a fold

change factor from the mean, representing a range akin to

a 95% confidence interval of the estimate. We first assess the uncertainty associated with our

estimate of the fraction of arc

haea out of the total biomass of soil bacteria and archaea. As a

measure of the uncertainty associated with our estimate of the fraction of archaea out of the total

biomass of soil bacteria and archaea, we collect the intra

study, inter

habita

t and

inter

method uncertainties within and between each of the four methods we rely on to estimate

the fraction of archaea out of the total biomass of soil bacteria and archaea. We use the maximal

uncertainty among this collection as our best projection o

f the uncertainty associated with our

estimate of the fraction of archaea out of the total biomass of soil bacteria and archaea. We thus

project that ≈4

fold uncertainty associated with our estimate of the fraction of archaea and ≈1.5

fold uncertainty asso

ciated with our estimate of the fraction of bacteria out of the total biomass of

soil bacteria and archaea (

link

to ful

l calculation). We combine this uncertainty with the

uncertainty associated with our estimate of the total biomass of soil prokaryotes (bacteria and

archaea), which we derive in the soil fungi section. The uncertainty we project as associated with

the tota

l biomass of soil prokaryotes is ≈4

fold. We thus project an uncertainty of ≈6

fold for our

estimate of the total biomass of soil archaea, and ≈4

fold for our estimate of soil bacteria (

link

full calculation).

Marine deep subsurface sediment

The two major habitats of microbes in the marine deep subsurface are subseafloor sediments and

the oceanic crust

(73, 74)

. We first focus on subseafloor sediments, as much more data is

available on this environment. We then turn to look at the oceanic crust.

Whitman et al.

(25)

originally estimated the global biomass of

bacteria and archaea in subseafloor

sediments to be around 300 Gt C residing within ≈3×10

cells. A later study

(75)

revealed that

the original estimates b

y Whitman were based on extrapolation from samples which were not

representative of the different levels of productivity in the marine environment. Using additional

sampling, the study by Kallmeyer et al. updated the estimate for the total number of prokar

yotes

in the subseafloor sediments to around ≈3×10

cells, an order of magnitude lower than

Whitman’s original estimate. The estimate by Kallmeyer et al. is based on sampling of cell

densities worldwide at different depths. Kallmeyer et al. built a model

to predict cell

concentration as a function of location and depth below seafloor. The model uses distance from

shore and sedimentation rate as the primary explanatory variables. The specific parameters of the

model are given in detail in

(75)

. Kallmeyer et al. plugged into the model global data on

sedimentation rates and the distance from shore, and used the model to predict cell concentrations

at each location in the marine subsurface. Kallmeyer et al. th

en integrated the predicted cell

concentrations across the entire volume subseafloor sediments the generate an estimate for the

total number of cells in subseafloor sediments. A later study

(76)

, gives a

slightly higher estimate

for the total number of prokaryotic cells in subseafloor sediments. Parkes et al. calculated the

total number of cells in the subseafloor sediments by using regression of cell concentrations